The Practical Use of Fracture Mechanics by DAVID BROEK FractuREsearch Ine., Galena, OH, USA. a nw Kluwer Academic Publishers Dre Boston /Condoncriteria. The treatment is necessarily compromising rigidity, but it serves the purpose of providing the insight without which Fracture mechanics cannot be ‘understood. (My research work covered fundamental fracture and fatigue mechanisms, experimental evaluation of criteria for Fatigue, fracture, and combined mode loading, the development of engineering procedures for arrest analysis in stiffened panels, collapse conditions. and damage tolerance analysis in general ‘My engineering background however, has always prevailed and forced me to consider the practicality of procedures, This book reflects a lifetime of ‘experience in research and practical apples, No subject is discussed on the basis of he=ceay Instead the tasis is “hands-on” experience with virtually every. issue from the fundamental to the practical. Tam aware of my shortcomings. prejudices and opinionations, but believe to be entitled to these on the basis sf my enéneeting experience. This text reflects them, and I do not apologize. Too many “refinements” in engineering solutions pertain to secondary errors; they increase the complexity, but do not improve the solution. One does not improve the strength of a chain hy improving the strong links, The weak links in the frncture mechanics analysis are the unknowns, not the procedures. This book is for engineering stuxlents and for engineers, who must solve urgent problcnss yesterday. Engineering solutions are always approximative, no matter wat the subject is. Such is the nature of engineering. Necessary assumptions are far more influential than those due to limitations of fracture mechanics. “The textis intended for the education of engincers™ AL the sume time i serves 1s a reference. For this reason there is sone duplication and extensive cross- references are provided. This may be objectionable to the reader going through the text from A to Z, but it will be of help to those who read sections here and there. It is not perfect as no human effort ever is, and I shall welcome construc- tive criticism with regard to the engineering applications. My haste in accom: plishing things (enforced by the unfortunate situation that I have to make a living, while writing a book is an extraneous effort which is not very profitable) may be reflected in the text. Again, 1 am not apologizing, just explaining. Tam grateful to my wife, Betty, for putting up with my preoccupations and ‘moods while writing this text, and for submitting all writing to a word: (oF. Tam also thankful tomy son Titus, who spent numerous hours in producing. ‘solutions to exercises and in drawing figures. dedicate this book to the memory of my father. Harm Brock. Many sons see their father as the ultimate example. So do I. Hix always been a driver of my amt Galena, Ohio, February 1988 oe Notice Extensive computer sofware for acre mechanics analysis was develo the author of this book. This software is capable of " ae = performing resid strength aly in atordace with Chapes a both for LEFM antt EPFM. er) and o revforming fatigue crack growth analysis for constant amplitude, random Ieatingand ser anon tang in accordance with Chapters Sand I, wth or without retardation There ar options for various retardation mode rte one and tabular sate data (Chapter 7) aly generating semirandom sites histories on the basis of ce diagrams (Chaper 6) and performing clipping and Function spon command = termining inspection intervals an cumulative probaly of dion in secontance with the proce dissed in Chapter Is wing theese crack growth curves and aecunting for speci and aces providing potesina pos ui xe hen of tris data isin wl a an ctenive rary of geometry factor. Bees pre-processor can generat Bem factors. wing most ofthe procedures dicused m Chapler "The shove softwares avalabe for personal computes Because ofthe lage Sze of the sofvare. i split up in seven module, cach of which sa Peon computer The medules communicate trough he te at me fereaed automaticaly. The ware canbe elaine rom FrateREsedrh the, 149 Cupstone Drive, Galea, 43021, USA ‘A much simpiied version ef the same software (ao by the author) i ‘valle fom the American Soi of Metals (ASM Metal Park GH 4007, USA. This imped version hat no data Hvar. no prprocesor for roel factor, cannot do retardation, and does not generate semi-random stress histories,Contents Preface Notice Chapter 1. INTRODUCTION LL Fra nieol| 12 13. Crack growth and fracture 14, Damage tolerance and fractu 16. Pxereises Chapter 2. EFFECTS OF CRACKS AND NOTCHES: COLLAPSE 24. Scape 22. Am interrupted load path 2.2. Stress concentration factor 24, 25. Yielding at a notch 26. Plastic collapse at a notch wge tolerance analysis mechanics 1S. The need for analysis: purpose of this book late of stress at a stress concentration 27. Fracture at notches: brittle behavior 28. Measurement of collapse strength 29. Frercises (Chapter 3. LINEAR ELASTIC FRACTURE MECHANICS. 3. Scope 32. Stress at a crack tip AX Ger 34. Toughness 3:5. Plastic zone and stresses in plane siress and plane strain ral form of the stress intensity factor 16. Thickness dependence of toughness 3.7. Measurement of toughness 38. Competition 49. ‘The energy eriterio 3.10, The energy release rate 311. The LIX. Exercises th plastic collapse eaning of the energy crite 3.42, The vise in fracture resistance: redefinition of toughnessChapter 4, ELASTIC-PLASTIC FRACTURE MECHANICS 4.1, Scope 42. The energy crit 43. The fracture 44. The rising fracture energy 43. The residual strength diagram in EPFM: collapse 46. The measurement of the toughness in EPFM 4.7. The parameters of the stress-itain curve 48. The i-functions 49. Accuracy 4.10. Historical development of J 4.11, Limitations of EPPM 4.12, CTOD measurements 4.13, Exercises jon for plastic Fracture Chapter 5. CRACK GROWTH ANALYSIS CONCEPTS 5.1, -Seope 5.2. The concept underlying fatigue crack growth 53, Measurement of the rate function 54. Rate equations 5.5. Constant amplitude crack grow 56. Load interaction: Retardation 5.7. Retardation models 58. Crack growth analysis for variable amplitude loading 59. Parameters affecting fatigue crack growth rates 5.10. Stress corrosion cracking SII, Exercises in a structure Chapter 6. LOAD SPECTRA AND STRESS HISTORIES 6. 62. of stress histories 6.4, Exceedance diagram 65. Stress history generation 66. Clipping 67. Truncation 63. Manipulation of stress history 69. Environmental effects 6.10, Standard spectra GUL. Exercises kk RR RR 93 7 98 02 106 109) nz N6 UR Ri 13 1s 123 126 10 ay 56 4s 149 137 163 165 8.10, oD a2, Net 7. DATA INTERPRETATION AND USE ‘Scope Plane strain fracture toughness Plane stress and transitional toughness, R-curve ‘Toughness in terms of J and Jy Estimates of toughness General remarks on fatigue rate data : Fitting the da/dN data Dealing with seater in rate data ting for the environmental effect Obvaiving retardation parameters 8, GEOMETRY FACTORS Scope The reference stress ‘Compounding ‘Superposition A simple method for asymmetric loading cases Some easy guesses Simple solutions for holes and stress concer Simple solutions for irregular stress distributions Finite element analysis ‘Simple solutions for crack arresters and multiple elements Geometry factors for elas Exercises Chanter 9. SPECIAL SUBJECTS 32 93. 9a Scope Behavior of surface flaws and corner cracks Break through: leak-before-break Fracture arrest Multiple elements, multiple cracks, changing geometry ‘Stop holes. cold worked holes and interference fasteners Residual stresses in general Other loading modes: mixed mode loading Composites Exercises inter 10. ANALYSIS PROCEDURES, ‘Scope Ingredients and ical locations 208 208 212 214 21s 218 22 232 238 241 3 23 24 247 249 255 258 260 267 an 28 28 219 282 282 282 290 293 3 316 319 327 329 332 332 32103. Critical locations and flaw assumptions 334 Chapter 14. APPLICATIONS load tereancaeere nme “ fot ae eats] scooters ia foam Seaedasd eeeoa mee he te Gates . 43 See tencrccorsa) the a Nat ee 33 ‘canper 1s. SOLUTIONS To Exencises = Naa ne ceca ecceenee et - idl he nease samme cc ras 3 sumecr mpex sis Natl ecee ees : a "ie eel oY at ner ene s nlaateesaes a Caper {2, DAMAGE TOLERANCESURSTANTIATION 38 heise = tome > as Sern mame meet os Re Ober aaretan as fe eae oe By Peele eer ime atnamsea my rial one ie fines ies Salde ANAL es rn os ies ai eeeeee ee ies alia ase es ne HX6 Possible actions based on failure analysic 40 137. Exercises “40CHAPTER | Introduction Fracture control control of structures isthe concerted effort by designers, metallurgists, ‘production and maintenance engineers, and inspectors to ensure safe operations. Without catastrophic fracture failures. Of the various structural failure modes (buckling, fracture. excessive plastic deformation) fracture is only one. Very seldom does a fracture occur due to an unforeseen overload on the undamaged structure. Usually, itis eaused by a structural flaw or a crack: due to repeated, ‘oF sustained “normal” service loads a crack may develop (starting from a flaw or stress concentration) and grow slowly in size, due (o the service loading. Cracks and defects impair the strength. Thus, during the continuing develop- J the crack. the structural strength decreases until it becomes $0 low that the service toads cannot be carried any more, and fracture ensues, Fracture control is intended to prevent fracture due to defects and cracks at the (maximum) loads experienced during operational service, If fracture isto be prevented, the strength should not drop below a certain safe value, This means that cracks must be prevented from growing toasize at which the strength would drop below the acceptable limit. In order to determine which size of cack is admissible, one must be able to calculate how the structural strength is affected by cracks (as @ function of their size); and in order (o determine the safe operational life, one must be able to caleulate the time in which a crack grows tothe permissible size. For this, one must frst identify the “tions where cracks could develop. Analysis then must provide information wck growth times and on structural strength as a function of crack size. ‘This type of analysis is called damage tolerance analysis, Damage tolerance is the property of a structure to sustain defects or cracks safely. until such time that action is (or can be) taken to eliminate the cracks, Flimination can he affected by repair or by replacing the eracked structure oF2 Se eee eee pela insted then in the proof test. A ich cracks would grow it ignores plastic deformation. Nevertheless it has been used successfully for ‘many years in design, Similarly, fracture mechanics can be used successfully. ‘Acclaimed inaccuracies are due to inaccurate inputs much more than due to inadequacy of the concepts, which will become abundantly clear in the course ‘of this book, Naturally, the results of damage tolerance analysis must be used judiciously, but this can be said of any other engineering analysis as wel. Although further improvements of fracture mechanics concepts may well be desirable from a fundamental point of view, itis unlikely that damage tolerance analysis can be much improved, as its accuracy is determined mestly by the accuracy of material data and predicted loads and stresses. Fracture mechanics can give useful answers to questions that hitherto could not be answered at al. The answers may not be perfect but a reasonable anwser is better than none. Unfortunately it is rather easy (o obtain a wrong answer: Pitfalls are numerous. This book is intended to explain the engineering usage of fracture mechanics, and to point out the pitfalls in detail, Some ofthe pitfalls are so treacherous that one sees things more often done wrong than tight. These tnd other things have led o some myths about fracture mechanics and its uses, which are hard to eliminate. They will be addressed in this book. Although the basis of fracture mechanics concepts willbe discussed, this book Focuses on how the analysis should he performed, on how to solve practical problems and on how to avoid errors. Attainable accursey, and the factors Meeting uccuracy, are discussed in detail. A brief recapitulation is given of those concepts of fracture mechanics that are actually used in practical damage tolerance analysis. For more in-depth treatment of those subjects, the reader is referred to other text-books on the matter (1, 2,3}. Yet, the present text provides Sufficient background for a proper understanding of the practical methods discussed, AAs this book deals with the practical use of established fracture mechanics concepts, references have been kept to a minimum. They can be found in more extensive texts. It is not necessary to reference established concepts, e.g references to Hooke’s law are superfluous, and so are references to accepted and éstablished fracture mechanics concepts. Hence, references are provided only in those cases where relevance or extent do not warrant complete treatment in this fext, s0 that use of the original publication might be desirable. 1.2. The two objectives of damage tolerance analysis Festablishment of a fracture control plan requires knowledge of the structural slrength as it is affected by cracks. and of the time involved for cracks to grow toa dangerous sire. Thus, damage tolerance analysis has two objectives, namely to determine |. the effoct of cracks on strength (margin against fracture) 2. the ersck growth as Fanetion of time.4 ‘These two objectives are discussed below. Figure 1.1 shows diagrammatically the effect of crack size om strength. In fracture mechanics crack size is generally denoted asa, fn Figure I.1 the strength is expressed in terms of the load, che structure can carry before fracture ‘occurs (fracture load). Supposing for the moment that a new structure has no significant defects (a = 0), then the strength of the new structure is P,. the (ultimate) design strength (load) It should be emphasised that the strength of the new. crack-free siructure is finite. Fracture will, and must occur when i i subjected to a load P,, otherwise the structure was over-designed. In every design a safety factor (ignorance factor) is used. This factor may he applied in different ways, but the result is always the same. In some arcas of technology the safety factor is applied to load. For example if the maximum anticipated service load is P,, the structure is actually designed to sustain iP, = 7, where jis the safety factor. The designer sizes the structure in such a manner that the stress is equal to or slightly less than the ultimate tensile strength when the load is P, (checks against plastic deformation are usually necessary as wel) ely. the safely factor is pte! the allowable stress: if the actu ‘material strength (ultimate tonsil sis: is iy, the structure is size in such ‘way that the stress atthe highest service load, P,, és less than or exit 6 Ful A cmex oe Te) 5 where / is again the safety factor. Hence, since load and stress are usually Proportional, the structure is actually capable of carrying jP, = P,. Plasticity may «°*! jwohibit proportionality, but since plasticity is generally limited (0, small areas at notches and stress concentrations, the above is approximately correct, But. even if i is not correct in actual numbers, itis true in spirit: the structure is designed fad by a Factor carry a foad higher than the highest anticipated service strength is P, = jP,. The value of jis ig most of the time. For example ‘cranes, bridges. off-shore structures, ships and airplanes are usually ‘much fess than #", Only in exceptional circumstances (e.g. storms) does the load reach P,. AL oti times the load! may be only a fraetion of P,, so that the margin ‘agains Fracture is much larger than j. except in extreme situations. The loads on some structures. .g. pipelines, pressure vessels, rotating machinery are reaching ‘more nearly always the same level (P,), as shown in Figure 1.2 The new structure has a strength P, with safety Factor js strength is finite, ‘so that the probability of fracture is not entirely zero. I'the load should reach 1. (€8. in a storm) the structure fails. The probability of this occurring is nnon-7ero. but experience has shown that it is acceptably low. If cracks are present the strength is less than P.. This remaining strength under the presence of cracks is generally eferred to as the ‘residual strength’, Pa; the diagram in Figure 1.1 is called the residual strength diagram. With a residual strength Pax < Py the safety factory has decreased: j = Pa/P, which is less than i = P.JP.. In concert, the probability of fracture failure has become higher, Fracture is the catastrophic break-up of the structure into two or more picees. With a crack of size a, the residual strength is P,,, Should a load P = P., occur then fracture takes place. The fracture process may be slow and stable intially, the crack extending (by fracture), but the structure stil hanging together. Eventually. the fracture becomes unstable and the structure breaks into two oF iiore pieces. The whole process of stable-unstable Fracture may take place in @ fraction of a second. If the load P= Pq does not occur, service loading Continuing at loads at or below P., the crack will continue to grow, not by fracture but by cracking mechanisins such as fatigue, stess-corrosion er creep, ‘Due to continual growth the crack becomes longer, the residual strength less, the safety factor lower, and the probability of fracture higher. If nothing is done ‘and the structure remains in service, the residual strength evenutally willbecome ‘equal to P. (or even equal tothe average service load P, in Figure 12). Then the safety factor is reduced to | and fracture occurs already at P,.i.¢ atthe (highest) service load. or even at P,. This is what must be prevented: the crack sheuld not Fiore 11, Residual strength in he pcience of cack: tength of vew srr a ~ Oi, = be allowed from becoming so large that fractive occurs at the serves lone i.mie wemnenete— ton rhe ae wenen—e wan Fiona Hoe/arane Beara © Figure 12 Schematic example of lad Noi. a) Types ang lho ucts ships ‘planes: (typi aang of rotating machinery. Hence, the structure or component must be replaced before the crack becomes dangerous, of the crack must be detected and repaired before such time. ‘The above implies that the limit should be set somewhat above P,. For example, one may require that the residual strength never be less than P,, where g is the remaining safety factor, and P, the minimum permissible strength. The design engineer or user does not decide what should be the safety factor j. This factor is prescribed by rules and regulations issued by engineering societies (eg. ASME) or Government authorities. Similarly, these rules or requirements should prescribe g. This has not heen done for all types of structures yet, while e.g. the ASME rules approach the problem somewhat differently (rules and regulations are discussed in Chapter 12). However, some rule or goal must be estabished, some decision made, to set the minimum permissible residual strength, so that the m: permissible crack size, a, can be determined from the residual strength Provided the shape of the residual strength diagram is known, and Py 7 Prescribed. the maximum permissible crack size follows from the diagram. In ‘order for damage tolerance analysis o determine the largest allowable crack, the {ist objective must be the calculation ofthe residual strength diagram of Figure 1.1. 1f a, can be caleulated directly from P, it may not be necessary to calculate the entte residual strength diagram, but enly the point (a,, P,). However, this is seldom possible and rarely time saving. In general, the calculation of the entire diagram is far preferable. The maximum permissible crack size follows from the calculated residual strength diagram. and from the prescribed minimum permiss- ible residual strength, P,. The residual strength diagram will be different for different components ofa structure and for different crack locations; permissible crack sizes will be different as well. The permissible erack sizeis sometimes called the critical crack sire. However, the objective of fracture control is to prevent ‘critical” cracks. A critical crack is cone that would cause fracture in service. Cracks are not allowed lo grow that long, Instead, they are permitted to grow only to the permissible size o,. They ‘would be critical only in the event that a load as high as P, would occur. Knowing that the crack may not exceed a, is of ltt help, unless itis known when the crack might reach a,. The second objective of the damage tolerance analysis is then the calculation f the crack growth curve, shown diagrammatic ally in Figure 1.3. Under the action of normal service loading the eracks grow by fatigue, stress corrosion or creep, at an ever faster rate leading tothe convex curve shown in Figure 1.3 Starting at some crack size a the crack grows in size during time. The Permissible crack a, following from Figure 1.1 can be plotted on the curve in Figure 1.3, Provided one can calculate the curve in Figure 1.3 one obtains the ‘Heme 13. Crack prom carve chemi)time H of safe operation (until a, is reached). If ay is For example an (assumed oF real) initial defect, then the component or structure must be replaced afler a time H. Alternatively a, may be the Fst of erack detection by inspection. This crack a will atime HF Since erack growth is not allowed beyond a,, the crack must be detected and repaired or otherwise eliminated before the time 17 has expired. Theretre, the time between inspections must be Jess than 17. Ai an inspection at time », the crack will be missed, because a is the detection fimit. Ifthe next inspection were to take place 1 hours later, the ‘rack would have reached a, already, which isnot permitted: ie. the inspection interval must be less than H; i is often taken as H/2. In any case, the time of safe operation by whatever means of fracture control follows from H. In turn, HF emerges from the damage tolerarcr analysis, provided both the residual strength diagram (a,) and the crack ¢r0:9th curve can be calculated to obtain H. 1.3. Crack growth and fracture ‘The residual strength and crack growth diagrams are essentially different, not ‘only in shape but also i we» Fracture is the final event, often taking place very rapidly, and resuiting in a breaking-in-two. Crack growth on the other hand occurs slowly during normal service loading. Also the mechanisms of crack growth and fracture are different Crack growth takes place by one uf five mechanisms a. Fatigue due to cyclic loading, b. Stress corrosion due to sustained loading ©. Creep. 4d. Hydrogen induced cracking, ©. Liquid metal induced cracking, OF these, the first wo, and combinations thereof are the mi ysevalent, while the last is hardly of interest for load-bearing structures. Crack growth is sometimes referred to as ‘sub-critical crack growth’, a pleonasm. Fracture is critical. and fracture isnot the same as crack growth AA crack may cause a fracture. There ate only two mechanisms by which fracture can occur, namely a. Cleavage. b. Rupture. A third ‘mechanism’, namely intergranular fracture, requires operation of some form of either cleavage or rupture. ‘A mechanism for fatigue erack growth is shuwn in Figure 1.4, Other mechanisms are possible hut not essentially different (4,5, 6). Even at very low loads there is still plastic deformation at the erack tip because ofthe high stress concentration, Plastic deformation is slip (due to shear stresses; see Chapter 2) of ' | PE Leone pm age coy he a10 atomic planes, depicted in Figure 1.4, s ary planes results in a blunted crack tip (Figure 1.4, stages B-D). The very first slip step in stage 2 has already caused a very small crack extension a. Upon unloading (or if necessary compressive loading) the crack tip again becomes sharp. Mechanisticaly, the whole process of slip could be reversed $0 that the ‘end result after unloading would again be as stage A. However, because of oxidation of the freshly exposed material along the slip steps, and the general disorder due tothe slip, the process is irreversible in practice; the crack extension ‘Sa remains. In the next load cycle the process is repeated: the crack grows again by Aa, Growth per cycle, Aa, is extremely small as can be judged immediately from * the mechanism in Figure 1.4. Typically, the growth ison the order of 10 *=10 inches (10°10. * mm}; however if the lond is eycled for 10*-10" cycles, the Crack will have grown by an inch. The repeated blunting and sharpening gives rise (o marks on the crack surface (often etroncously referred to a8 fracture surface). which can be made visible at high magnification in an eleciron microscope, as shown in Figure 1.5. The marks, called fatigue striations, Fepresent the successive positions ofthe crack front. i. the blunting/sharpening, steps. If the crack grows by 10 “inch, then the stration spacing is 1) "inch (ie 0.05 inch at a magnification of $000 x). Not all materials exhibit striations as regular as in Figure 1.5. (4, 5, 6] (Chapter 1). ‘Crack growth by sitess corrosion is a slow process as well. The crack extends due to corrosive action (often along the grain boundaries) facilitated by the high stretch and consequent atomic disarray at the crack tip. A common mechanism of creep cracking is the diffusion of vacancies (open atomic places). a con- elomerate of vacancies forming a hole, which subsequently joins up with the crack tp, AA crack by itself is only a partial failure, but it can induce a total failure by fracture. Fracture occurs by cither of two mechanisms, cleava (Figure 1.6). The facets by themselves are lat and, therefore, good reflectors of ident light. This causes the cleavage fracture to sparkle when fresh, but the alter may soon fade due to oxidation. ‘The alternative fracture mechanism of ductile rupture is shown in Figure 1.7. All structural materials contain particles and inclusions. These particles are generally complex compounds of the alloying elements, Some aloying elements are used to improve castability and machineability; others are specifically included to improve the alloy's strength. First the large particles lel loose or break, forming widely spaced holes close to the crack ip. In the final phase, holes are formed at myriads of smaller particles; these holes or voids join up to complete the fracture. Because of its regularity the Fracture surface diffuses the a Gs FPiure 13. Fatigue salons on rac scfce of slain alloy. Spacing of sation coincides ith eye londing (ise), Magaieation 12000 light and looks dl rey. The hoe ofthe lange parc and those ofthe small pattles (Figue 17) are visble at hgh magniaton in the electron Incroscone. as shown in gue 18 Aste factor surface shows the halves of the oles, the fracture i referred to as dimple rupture or ust rpte. El thon ofthe large patcles (by sketivelloying and heat treme can improve the factreresintance ofa ally [4] On the other hand to provide strength inthe fist place so that not al pat That they ply aoe the nal factor proces becorftvater of secondary importance; without the pa Both cage lractore and rupure are fast process, A cenvage facture may rum as fasta leer (100 ma dimple rupore a stn 150 see (Son although maybe slower. Fracturing is sometimes sable. The creckFire 16. Cleavage facto starting at (Sunted) cack ip, Botom: Aa facets gliter due to 7 reletion of ince Fight (Fracture) then extends by one of the fracturing processes (cleavage or rupture) instead of by one of the cracking mechani, Usually stable fracture is immediately followed by the final unstable ts : ro re ere clue of ste below the transition temperature costs gr 7. Rowan terete: cece shows hoe yg ad by cleavage. Cleavage is often referred to as brie fracture, while dle eprint rupture i described as ductile fracture, This may be adequate for melallurgnts but in general one shouldbe very careful using these terms, By far thc mejecly defects, a true overload fracture being very rare indeed (Chapter 13). 1 is of service fractures occur by dimple rupture, but most ofthese exhibit ver hl equally wrong (though not as confusing) to speak of fatigue overall plastic deformation, 0 thatthe are bile from an engineering pete ot corrosion fracture, The fracture may bea consequence off view (see Chapter 2). Reference to cleavage and rupture avens conison, A . siress corrosion crack, but it occurs by cleavage or rupture brittle fracture is one with little (overall) plastic deformation, leavage ‘There is a third fracture type, namely intergranular fracture. The distinction or rupture, here is on the basis of fracture path (along the grain boundaries). Nevertheless, . fractographers sometimes refer to rupture as the ‘overload fracture’ 7 intergranular fracture is not an altogether different type. Whether inter Bush it from the fatigue or stress corrosion crack surface (Chapter 13). fF transgranular, the mechanism of separation either resembles ch ‘This can also cause confusion, Distinction between ‘fracture surface’ and ‘crack rupture, be it that cleavage at @ grain boundary surface’ is more appropriate. Practically all service failures are due to cracks or trancersiar cleavage. i eeFrewe 18. Fracture sutice of ductile rapture showing a inteiconsectng dpe; Alumni sty oles wih cracked large ptices and Scanning letron microscope; a0" Baton Is 41.4. Damage tolerance and fracture mechanics [Pamage Colerance analysis must provide a capability forthe calculation of both the res iagram ({racture due to cracks) and of the crack growth carve. Fracture mechanics methods have been developed to analyee fracture te ‘obtain the fracture stress (residual strength), and to aralyze fatigue rack creep crack growth will not be discussed, ‘Aithoueh itis possible to deal with stress corrosion cracking in principle, the {Zack erowth times ae usually so short (rom ! to 100Dh) that crack growth ing from a practical point in view. Fracture control in the ase of stress corrosion is often aimed at crack growth prevention, Freciure ‘mechanics methods to accomplish this will be discussed In the treatment of erack growth emphasis will be on fatigue cracking. In ‘funy structures fatigue cannot be prevented altogether; eracks then will ern These ust be dealt with by means of fracture control ie. they must be liminated before they can cause a fracture, The time to the initiation ofa foxigue crack will not be considered for two reasons: frst because fractare with existing eracks, and second because itis not necessary to initial erack be assumed, Fracture mechanics (as all engineering mechanics) uses stresses, rather than toads Thus, the residual strength diagram of Figure 1.1 is normally based upon ‘rx the stress the structure can sustain (instead of load), before fracture occurs The residual strength diagram based upon stress is shown in Figure 1 9. For this ‘urpose the engineering sress is sed, asin the case ofthe uncracked structure, de sidual strength being given as o,, (Figure 1.9). Noe thet gn is a strength {ike the tensile strength or the yield strength) and not a stress. Fracture oxesey Rihen the stress equals 0: fracture if ¢ = oy). Residual stength should not bh confused with residual stress, (Residual tress ia stress residing ina stuctore While there are no loads applied) ae Stress can be used a8 basis for the analysis i thee is relationship between Ihe applied stress and the processes taking place atthe crack tip Avahe enact {in events are governed by the local siresses at the crack tip. required ned taal crack tip stress be described as a function ofthe applied stress, Such ‘clavionships ean be derived, provided the problem is defined clearly. For th the 3 modes of loading shown in Figure 1.10Fire 1.9. Residual svength diagram on the fas of nominal enginerng sem Hey ‘Pere 110. The mode oatng. Mode opening mde Mode seat med, Mode tating ‘These modes of loading (no! modes of cracking), are usually refered to simply bby Roman numerals I, 1 and 111. Other descriptions used are opening mode or tension mode for mode I in-plane) shear mode for mode Il. and out-of-plane) shear mode or tearing mode for mode Ill {turns out that the crack-tip stress equations ste very si modes. As a matter of fact the format of the equation Consequently, the fracture and crack growth a ‘modes individually, turk out to be identical, be it that different numbers apply Fone knows how to deal with mode 1, one knows essentially how to ansiyse ‘modes 11 ani wode HHL individually. This is one reason why thir book dents primarily wits code 1. However, there are other more compelling reasons, {In practice, by far the majority of cracks result from mode | loading. The ether ia foreach of the ‘exactly the same, ysis procedures for each ofthe "7 {wo modes do not occur individually, but they may occur in combination with ‘mode 1, ic, I-Il, HII or 1-If-I11. However, if the loading of these modes isin Phase, cracks will rapidly choose a direction of growth in which they are subjected to mode f only (Chapter 9). Thus, the majority of apparent combined ‘mode cases are reduced to mode I by nature itself. There are few eases left then| which cannot be treated as pure mode 1. These occur when for example in a ‘mode 1-11 combination (where the crack would normally select a mode I path), ‘here isa direction in the material where the resistance to eracking and fracture ‘s substantially lower. A case in point isa circumferential weld in a torque tube (Chapter 9). Normally, a crack in a torque tube will develop and grow under 45 degrees (tension only) and thus be in mode I. But if circumferential weld ‘confines the crack (o the circumferential direction (shear plane) mode 11 must bbe considered. Another case occurs when mode I and e.g. mode II loading are ‘out of phase (bending and torsion eycles with different frequencies; see Chapter 9 h individual loading modes are easy to deal with, » combined motie uling case is rather more dilfcull, Several combined mode (or mixed made) analysis concepis have been proposed. A practical approach will be presented in Chapter 9for the Few instances of "true" mixed mode loading described above. [LS. The need for analysis: purpose of this book During the first half of the industrial era structural failures were numero lroad accidents elevstor accidents, boiler explosions, ec., were a common ‘occurrence; accidents due to structural fracture were reporied weekly if not daily. The first powered flight was postponed because of a broken propellor shaft. Orville Wright returned from Kitty Hawk to Dayton to machine & new sha, which caused a 2-week delay of the ist fight. Due to improved materials, id refinements of design procedures, and not in the least, due to the enforce: ment of design safety factors and quality control measures, the number of strucutral failures has abated, but not been reduced to zero. ‘More than two dozen major bridges have collapsed during this century. Since ‘World-War It in excess of 200 civil airplanes had fatal acciderts due to fatigue cracks: the number of cracks discovered in commercially operated jets is estimated at well over 35 000. Many cracks have been reported in nuclear power structures. Although less frequent, catastrophic structural failutes have far more serious consequences now than in the past. At the same tiThesociety has become fess tolerant and abundantly more litigious, Thus, no manufacturer, nor ‘operator of larger structures, can afford to ignore the issue of fracture control Rational fracture cont‘commercial airplanes. Damage tolerance analysis of an airplane subject to these ules, typically requires 20000-60000 man hours. (An equal or larger effort is in the associated testing for material's data and analysis substantiation). The [ASME boiler and pressure vessel code requires damage tolerance analysis, at least for nuclear power structures. Before long, requirements are likely to be implemented for railroads, ships. bridges, pipelines and other structures, but ‘many manufacturers are already using some form of damage tolerance analysis. Fracture mechanics concepts and damage tolerance analysis may not be idea! (they never will be), but they provide answers where none were available before. Judicious use of these can certainly reduce the risk of fracture. (No tisk in life is ever reduced to zero). The purpose of this text is (o present the procedures of ‘practical damage tolerance analysis as they are (or can be) applied to almost any structure, including ships, pipelines, aircraft, pressure vessels, cranes, bridges and rotating machinery. Each of these structures presents its own specific problems, but there is enough common ground fora universal procedure, which can be modified, simplified or extended to serve a certain purpose. The various ‘components of the universal procedure will be discussed. Specific use and interpretation for various types of structures will be reviewed; examples of applications given AA theoretical text this is not. The emphasis is on how damage tolerance analysis is and should be performed, on how the necessary information is ‘obtained and used, on how the results of the analysis are employed to exercise Fracture control. The reliability of the analysis and the major sources of error are discussed. Engineering approaches and approximations are given ample ‘pace: in view ofthe general inaccuracies caused by material data and projected loads and stresses, approximative solutions are quite adequate as the errors so introduced are secondary. In many instances, approximative solutions are the ‘only feasible ones from an economic point of view. It may well be reasonable {o spend large sums on the analysis of one crack in a nuclear pressure vessel, the cost of a power plant being extreme and the consequential cost of a failure beyond imagination. However, the analysis of a crack in a hand-tool must be ‘cheap and the best engineering methods are indicated, For an aircraft structure with hundreds of potential crack locations in very complex details, so many ‘cases must be considered that approximations and engineering judgement must bee used. The same holds for ships, offshore structures, bridges. etc. shown in this text why and where appro. 8 are useful and st order of magnitude of the error in the final solution willbe the guideline. ‘The ‘how’ cannot be understood without some knowledge of the ‘why’ It would be a disservice to the reader to refer her ut him to other texts for the “why's in particular where theae texts are often intended for theoreticians rather ners, Thus the present text provides the concepts and theoretical background in an abridged and simplified form. adequate for the understanding 19 of applications and use. The background is contained in Chapters 2-5 which [resent the concepts of fracture mechanics for the analysis of fracture (elastic as well as elastic-plastic) and for the analysis of crack growth. These chapters, although providing sufficient basis for what follows, can serve as an introduc. tion to more comprehensive texts eg. 1, 2, 3] for readers desiring more theoreti- cal information, Chapter 6 is devoted to the analysis and interpretation of load and stress Fistories (the sequence of cyctic stresses causing fatigue crack growth). The stress history is of eminent importance in damage tolerance analysis, and yet, this subject is rather ignored in theoretical texts Mathematics have been kept to the barest minimum. Lengthy derivations do ‘not appear if they are irrelevant to the understanding of the ‘why’. Instea dimensional arguments are used (o show why an equation must be ofthe form in which it appears. Nevertheless, as in any analysis, a certtin amount of ‘mathematics is unavoidable and necessary Following the background are chapters dealing with obtaining the ingredients for the analysis. and the performance of the analysis, Chapters 7, 8 and 9 respectively provide information on the interpretation and use ef the materials data needed for input, simple methods to obtain geomeiry factors for complen Structures and cracks, and the effects of special conditions such as mixed mode loading, cold work of holes, interference fits, residual stresses, els. Approxima tions and the assessment of their errors are discussed in extenso. The analysis procedure is illustrated on the basis of examples of applications to different {types of structures, cracks and load spectra (Chapters 10 and 14). Chapters 11. {12 and 13 show how damage tolerance analysis is used for the implementation, of fracture control measures, while addressing sources of error, pilfalls, common misconceptions, rules and regulations, and the establishment. of fracture control plans. ‘The reader may find that certain subjects and problems are touched upon at several places in this book. Cross references are then made to the pertinent chapter where the subject is discussed in detail. This was done on purpose, at the risk of offending those readers who go through the entire text in sequence. 1 is anticipated that many readers will use this as a reference text; cross references to related subjects and problems are included for theit convenience. For those readers and students desiring to sharpen their skill, exercise problems {are provided at the end of each chapter; Solutions are given in Chapter 15. ‘A final remark about the use of units is appropriate. Although metric units become more and more accepted, English units and kg-force are sill used at ‘many places, Since the unit system used is of no consequence whatsoever for the «essence of the engineering procedures, various unit systems are being used in this {ext. No attempt is made to provide other units between parentheses, as this Fequiresa lot of space and subiracts more than it adds. The only units of interest2» are those of length, stress, force, and stress intensity. Should the reader want to do s0, she or he can easily make the conversions. Material properties such as tensile strength, yickl strength and colapse strength are consistently referred to as strength and denoted as F.., F, and Fog respectively, Although they are expressed as sivess, denoted us a, they are critical values (materi 16. Exercises |. Design a tension member with circular cross section to carty a load of 11000000 Ibs (1000 kips) with a safety factor of 2 apninst yield, and a safety factor of 3 against static fracture. Consuler tws ussterias: Material A with F, = SOksi and F, = ksi; Material & with F, = SOksi and FF, = BOkSi, What i the ulate design lon! in cach ease 2. Design the same member as in problem | with: “ fety factor on load of 2.5 for static fracture, while no yielding is allowed at 1.3 times the maximum service load, 3. Assume that for the members designed in problems | and 2a damage tolerance rule applies requiring that the residual strength may never drop. ‘below 1.2 times the maximum service stress. For all cases calculate a, and P,, How large is the remaining safety faciur in eac 4. Define the difference between cracking and fracture, By which mechanisms do cracks develop and grow? 5. Assume Material A and the design of problem 1. and assume the damage tolerance requirement of problem 3 is in effect Furtics assume that the maximum permissible crack size following from the damage tolerance re- uirement is 2 inch. Sketch the residual strength diagram in terms of stress and load, and identify all known points (novsclature and number) x the diagram, 6. Using the result of problem 5, and assuming that cracking aceurs by fatigue, estimate graphically the fatigue crack size that would cause a fracture if a {oad equal to 1.5 times the maximum anticipated service load would occur (is this likely 19 happen’). Also estimate the crack size which woul eaure a fracture at the maximum anticipated service stress (under which circum stances might such a fracture occur”). References UTD. Broek, Flewewtryemeinering facture mechani, Fon, Ni (1985) a 2] MF Konninenand CH Poplar. Adtancedracrre mechan, Oxford Univesity Pes (198). LAT Kooi, Fidanenats of acre mechan, Baterworths (193), HID. thock, Some contsbations uf ckectronTacigraphy to the theory of facture, Int. Met Reviews, Revi 184.9 (1974), pp 15-181, 19,0 Mowkesand D-Brck, On the Format p78 45, 161 eins, Modes of frie, Metals Handbook, 12 (1987), pp 12-7 mo ative striations En. Frac. Mech, 8 (1972),‘CHAPTER 2 Effects of cracks and notches: collapse 2.1. Seope well ni late. tn Linear Elastic Fracture Mechanics (LEFMs Chapter 3) a8 well Plate Fractre Mechanics (EPFM; Chaper 4) the analy of fractre is based ons parameter representing the crack lip ses el while the quanti sd a EEPFM is actualy the sin energy release rate, this can be shown to be caivalent to srs eld parameter neither LEPM nor EPEM the possi iy of socalled plastic collapse i inpicily evaluated. Fracture mechanics fnalyis may provide a Gace ess rendu strength) higher than the ates Tor pli collape:ice the manimur ond eareying capacity reached a he time of colapas the fracture see eaclted wih raclre methane ay Be imcrror(ooigh, Siniary sucha situation, the calculated eral crack ie would be to lage. Fractire parameters messed in test where fracture Cart dato claps would be Loo low No fracture analysis is compete, without the evaluation of collapse conditions Cola and racrenecometing coins an te ne aii fist wl prevail Before he ractre mechanics analysis procedures are acu im Chapters 3,4-and the coodiions for collapse wil be reviewed ln tht chapter, 2.2. An interrupted load path commits, fl an oss, and crackin pic gv is oases concentaton eal egon vie teases we hgh tan the wo or everest. The flowing i 8 it snd Seed dasuon of eat CSncetatons defo geomet te etre, Ses cm ans ply a deci len vir al sacral caching probe Comer wo paral sof the sue Sado he se ae an vigure bach cars Ba ie atl los tesa nears equa tating a longton 21. thee ar cat in wo the gf ml have to 2 2B Pure 2.1, Elect of eating one of to parte bace. carry ll the load: the stress the strain and the elongation will be twice as high as before. Consequently, the gap hetween the two halves of the let bar will be 2OL. The left bar carries no sigess at all (Figure 2.1). Next consider the case where the left and right bar are altached (e.g. welded) as depicted in Figure 2.2.1 the bars are intact, the situation is identical to the Previous, However ifthe left bari cut a diferent situation develops. I the op ball of the right bar is strained, the top half of the left bar must necessarily tundergo approximately the same strain. Both bars being of the same material (same modulus of elasticity, £). equal strain in the bars dictates equal stress (a = «£). Thus, each bar sill carries the same stress and therefore each bar ‘carries half the load. However, since the left bar is cut, the right bar alone must carry all toad across the cut. Relow the cut the two bars ate again attached and ‘must strain equally. Consequently the bottom halves of the bar again share the toad equally ‘The attachment ses the condition for approximately equal strain and equal stress in both bars, almost all the way to the cut, Close to the sit, the load of the left bar must be transferred tothe right bar which, overa shor distance, wll then carry the total load. The nominal stress in the right har ifthe section of the slit is then P/4. However, the load from the left must be transferred to the Fight and back over such a small distance that the additional load cannot be buted evenly. Instead, most of the extra load will be carried by a small ht half cross section, so that higher stresses occur dose to the ure 2.2cy: there is a stress concentration at the cut. It does not make am4 ome of two wee bars obs one (ier inter ba difference whether the bars are welded together or whether they are one piece, cut until midway. Transfer of the load from the cut hal! © ive place by shear. This can be appreciated by imagining how. sai the left deforms in the area of the slit as depicted in figure 2.2. ‘As long as the bar is intact, the strain is uniform aod the total elongation is ‘AL. W the bar is half-way cut the stress and strain will still be uniform in the ‘extreme upper and lower portions; they will be the same as in the bar without the cut. Only in the area ofthe slit, where the total load must be carried by half of the bar, the stresses and strains are higher. Hence, the total elon bbe somewhat larger than AL, but much less than 2A/ (compare Figures 2.1 and 22), Itis helpful to consider load-path (load-flow) lines: imaginary lines indicating €. how one unit of load is transferred from one loading point to the other (Figure 2.). For uniform load the flow lines are straight and equally spaced, idicating that the load is evenly distributed (uniform stress). IF the load path is interrupted by a cut, the flow lines must go around this slit within a short distance as shown in Figure 2.3. Were the load lines rubber hoses and the cut a wedge, a i iern would develop. ‘Atte tp ofthe ett the low ines are sey spaced indicating that more load is owing through a smaller area, which means higher stress. Load flow lines are alto useful for obtaining a rough indication af the direction of steese_ tn bby-passing the cut the flow lines are bent, ie. the load changes direction. The direction of the load flow line is an indication ofthe direction ofthe local tensile 2s o Oo) cy Fite 23. "Losd ow ines. slress, as indicated in Figure 2.3.. It appears that the direction of load around. the notch is not the same as in the uniformly loaded part: the local stress has a vertical ts well asa horizontal component, and it must be conchided that in the vicinity of the cut the stresfield is biaxial (2, and a,) while the applied load is ‘uniaxial. tn the absence of the cut the stress field is uniform, the state of stress uniaxial throughout. Due to the slit biaxial stress field develops locally. Not only does the slit cause a stress concentration, it also gives rise lo a transverse stress, 2.3. Stress concentration factor In the case of a notch (instead of a sharp cut) the situation is verysimilar. Every discontinuity forms an interruption of the load path, will therefore deviate the load flow fines and, hence, cause a siress concentration (Figure 2.4). Ifthe notch is blunt, eg. a round hold, its dimension in the direction of load is larger, causing an earlier deviation of the load-flow lines. Redistribution of this load ‘can then take place over a larger distance. As a consequence, the area of stress «concentration is more extended than inthe cas of cu, an the highest stresses are less, ie. the stress concentration is lower. ‘Asa general rule, blunt notches produce lower local stresses, sharp notches ‘cause higher local stresses. The highest local stress gis a numberof times higher than the nominal stFe55 Ojqe: The ratio between local stress and nominal stress is called the theoretical stress concentration factor, denoted as k, (clastic esses). The local stress is6 Fy 0 OD B Pare 24, “Lond. fines around otches and contequent stress concentaions 61 = Ka ‘The nominal stress is not always defined in the same manner. Someti we i] 18€5 the (uniform) stress in the full section away from the notch, sometimes the . average stress in the section through the notch. As the local stress is the same in either case, the use of a different definition for the nominal stress leads to different values of the stress concentration factor. This presents no problem as {ong as the proper combinations are used to obtain the local stress. Toran cpa oh igure 2) hse scr acer he tead 22) where @ and b are defined as in Figure 2.5. The radius of curvature, g, of the ba ellipse at the end of the transverse a = ah, so that Equation (2.2) can tho be won a b lee sat ey In the cae of a circle, b @ = R; being the radius ofthe hole ie £0 thatthe stress concentration factor ofa circular hoe is equal tok, = 3 aor 25. Eick High ste concent) Low ses cet Sues Sirs concentration factors for many wotch shapes have been determined bad pestering pled adel aaa They canbe found in handbooks, the bet known being the one by Peterson [I] Equation (2.3) is a more or less general form for the stress concentration factor sharper the notch (smaller 9), the larger is k,. For an ellipse with bia = 3(o = ithis interpreted as a relevant geometrical dimension («depends Upon notch 7) the stress concentration factor ish, = J, while for an elie witha & 3, scometry). The equation shows the large effect ofthe notch-rot radius g; he the sess concentration is only = 1678 2.4. State of stress at m stress concentration ‘As was shown in Section 2.2 a stress concentration also causes a change in the state of stress; even if the stress is uniaxial throughout the remainder of the body, the state of st-. = the area of the notch will be atleast biaxial (Figure 2.3), At the free surfs ere no external loads are acting the state of stress will be plane stress (therg are no stresses on a fee surfnce). Since the free surface ‘caries no shear either it sa principal plane with a principal stress equal to zero. A state of stress in which one of the principal stresses is zero isa state of plane ‘The face of the notch (Figure 2.5) isa free surface. it carries no stress (plane stress). The root of the notch (if there isa radius) is also a free surface. This means that at the free surface of the notch root a, must be zero, because there is no stress on (perpendicular to) that free surface. Shghtly inwards from the ‘notch root, however, a, will be non-zero. The faces ofthe plate are stil stress free, s0 that 6, = 0. The state of stress is (plane stress). At some small distance from the notch root, the state of stress may he triaxial as explained below. Because the surface (face of plate) is a principe! nline. snd because the three principal planes are mutually perpendicular, it follows thet the stresses 0, and , at the notch root in the plane of the notch are the principal stresses 6, and. 0. Due to the stress concentration the local values of a, and a; are very high and so are the strains e, and ¢,. According to 11 this will lead toa ea assuming that the stress ¢, = 0, This nege the plate. ‘The stresses are high only in the vicinity of the notch root. Further away 2, vanishes and o, is much lower. Hence further away e, would become very smal. The faces of the notch are stress free, so that e, = along the faces. N appears that there will only be a small amount of material at the notch root for which «should be very large, while around tc, is either 7ero (notched faces). oF small, (Carther inwards). Assuming that the material wanting to undergo large f. is approximately a cylinder (roll) the situation is as shown in Figure 2.6, I the roll is very long (large thickness) and thus relatively thin, such a large «cannot take place. The surrounding material, being attached to this roll, will ‘ot permit the contraction to occur, apart from litle bit at the face ofthe plate ‘Imagine that the roll of material isa stel bar cast in concrete, where the concrete represents the surrounding m: cooled, it wants to contract, However, contraction will be prevented by the concrete, which would lead 10 indicates a thinning of a Pare 26. Contraction a1 woth, (Uhcrmal) tension stresses in the bar. The same happens to the roll of material in Figure 26: contraction is prevented by the surrounding material and a tensile stress develops in the roll Apparently, constraint of the contraction causes a tension a, in the Z- direction. Should the contraction be completely constrained then the straint, is ‘zero, Writing the complete equation for F, provides: a= G-rE- eke == es and then; + = ra, 44) = 0300, +4,) 26 Equation (26) shows that a high tension develops in thickness direction when ‘no contraction is permitted. This stress is exerted on the roll by surrounding30 material cannot exist, but it builds up rapidly, going inward (Fi ‘constrain the contraction. At the free surface this stress re 2.7). Due to the absence of a6, at the surface, an ¢, occurs there, so that a small dimple develops at the surface It follows then that atthe surface, with stresses only in X and Y-ditetion, there is a biaxial state of stress (plane stress). Further inward there isa triaxial sate of stress. Should there be complete constraint then thie tr sressis plane strain, because ¢, = 0 for complete constraint. (A state of stress where one of the principal strains is zero is called plane strain) [Now considera thin plate with a notch as in figure 2.7b. In this case the roll ‘Fiure 27. Conia of long. thin ender (ee thickres) and fee conection of short, thick ‘inde (ml thickness) a Fieare 2%. Length of contacting ylindr bear no reaion to hick ein as of snc noch of material wanting to undergo contraction is short and relatively thick. con traction can occur freely and will be in accordance with Equation (2.4). The stress in thickness direction will be zero (a, = 0; plane stress) {In cases between those of Figure 2.7a and b, there will be some, but not complete constraint. These transitional cases have a triaxial state of stress, but not one of plane stra Apparently the state of stress at the notch root depends upon the length of the roll. As the length of the roll is equal to the thickness, one can argue that the state of stress depends upon thickness. However, this dependence upon thickness is a coincidence. Ifthe notch is of the type of Figure 28, the length of the roll has no relation to the thickness; in such a case the thickness is irrelevant for the state of stress. Although this may seem trivial, its importance is emphasized; in the use of fracture mechanics serious errors are possible if it is assumed that the state of stress is always dictated by thickness, 258. Yielding at « notch The stress required for plastic deformation depends strongly upon the state of stress. In plane stress yielding occurs when the highest principal stress is ap-2 proximately equal to the yield strength, but much higher stresses are required in the case of a triaxial state of stress. For readers not familiar (any mote) with yield i summary may be of help. Yielding is plastic deformation which takes place by slip; it is therefore cused by shear stresses. Plastic deformation will not take place unless the shear stress is sullicient to cause ship. In tensile hit (uniaxial {ension) the state of stress is as shown in Figure ? 94, Yielding occurs when a= F,, where F,, is the uniaxial yield strength as messured in a tensile test. Since plastic deformation takes place in the case wt trate 29a, the shear stress ia, the following brief | ld bh o | | L wl mars wh oof pee oe @ ‘Figure 29. Posiponement of plas deformation due to combine ses (a) Unt ong ‘al seat flowed by si (atic deformation: (b) Uninal traverse thea fallowe hy ip (6) Basia no shen. Ts Past deformation () Rawal ay = #2: plas deformation “ess tee a8 igh a in ce 2 3 ‘equited for slip must have been exceeded. The maximum shear stressacting on ‘plane at 45 degrees is « = 6/2. Apparently then the shear stress required for slip (yield) equals t,, = F.2 Figure 2.9b shows a similar case of uniaxial loading: again yielding occurs when the shear stress equals F,,/2. Next consider Figure 2.9¢ assuming for the ‘movment that the material is two-dimensional (existing in the plane ofthe figure only). Since there is equal tension in both directions the two shear stresses of igures 2.94 and b cancel, so that there is no shear stress in the case of Figure 2. Without shear strgss there can be no slip and no plastic deformation. In order for stip to occur one of the stresses must be larger than the other by F,. in which case the same shear stress will be present as in Figure 2.93. For example, i the horizontal stress is F,. the vertical stress must be 2F, for yielding {to.occur, so that tv. hes sess must be twice the yield strength before plastic deformation begin 1 she horizontal stress were 3F,, the vertical stress would have to be 4F, belore the shear of Figure 2.98 would be restored. In both of these examples the difference hetween the two principal stresses is F, Naturally, a real matcral is three dimensional as in Figure 2.9, and out of plane shear siresses are possible But it is easy to see now that there stil will be tno shear when all three principal stresses ate equal (Mohs circle becoming a point), and hence, there will be ne plastic deformation. Apparently, yielding ‘equites that the difference between the largest and smallest principal stress is ‘equal to F,. because only in that case will there be a shear stress ¢ = F,/2 a8 Fequmted for yield. Thus, yielding (slip) will take place when (x = F,/2) (=a) = Fy or 4) = Fe en depending upon which is lower 0: or a. Were for example a, = 0; = 08 a) then it would follow from Equation (2.7) that the stress required for yie orm = SF, ic. the stress would have to be 5 times the yield strength for plastic deformation tooceur. Ifa, = a; = a4. the shear stressis zero, and yielding will never occur ‘egardless how high the stress, The above is the Tresca yield criterion. Instead, the Von Mises yield criterion is more generally used, but its results difler very little from those of Tresca. As it makes no difference for the essence of the discussion. the Tresca criterion is used here, because it is the easiest to understand. The above has important repercussions for the yielding at a Rote, since the Sate of stress may be triaxial. For example, consider a case of plane strain, where the third stress «. is given by Equation (2.6). Note again that for a material element in the section th pal stresses 0,23 ‘ayate a... respectively. Consider a case in which «, = 6, as would occur ata very sharp notch. Given that » = 0.33, the lowest principal stress a,M would be ¢, = 0.33 (9, + 0,) = 0.66 o,. Plastic deforma stress 0, = 3F,, according to Equation (2.7). igure 2.10 represent the notch and the distribution of @,, on the section through the notch. At the notch root the local stress so, . At low loads the local stress is less than 3F,,and therefore the strains remain elastic, Further increase 3F,,, upon which would then "Next consider a notch in a thin plate with plane stress as in Figure 2.11. In this case plastic deformation occurs when 6, = F,, because (¢, ~ @.) = (a, — 0) = @,. During further increase of the load the stress distribution develops. similar manner asin the previous case. Eventually the entire remaining section will yield, untess fracture occurs earlier. plane steain than in plane stress. n the latter case they ae limited to Fin the former case 10 3F,, ‘Thus a plane strain condition is more severe and can more easily lead to fracture and cracks as will be shown in detail in later chapters. I the notch is blunt, the stress @,(¢3) cannot exist at the noteh root (Figure 2.5), Although there could be a g, (3), the state of stress would still be plane stress, because of a; = 0. Hence, at the root of the notch the stress 6, would be limited to the yield strength F,,. Further inwards a, will exist: in a thick plate with constraint, the state of stress will jal and possibly plane Strain, Further inward therefore the stress might reach as high as 3F, . as shown in Figure 2.12, before yielding occurs Fiee 210 sharp noch in plane stain as ‘Pigre 211, Yielding at notch in plane tes, Assuming the material does not exhibit strain hardening (horizontal stress strain curve beyond yield), the stress cannot increase further after yielding ‘occurs, Away from the notch there is again a uniaxial state of stress(a, = 6, () so that the stress will again be limited to F,,. Hence, by the time the entire section is yielding (Fracture may occur before this can happen) the stress distri bution in the seetion is as shown in Figure 2.12. Whether there is plane stress of plane strain at the notch, the final stress distribution will be about the same (compare Figures 2.11 and 2.12 and note that triaxial stress area is very small), part from a small area where the stress reaches 3F, in the plane strain case. Note that these stress distributions may not be reached if fracture occurs before the entire ligament yields, 26. Plastic collapse at oth ‘Not only may the high local stresses cause cracks by fatigue, stresscorrosion (or creep) which may eventually lead to @ fracture, they may cause fracture to proceed immediately from the notch, in particular when the notch is sharp (eg, crack). Alternatively, faiture can occur by plastic collapse which is always followed by fracture. Ifthe stress distributions of figure 2.11 or 2.12 can be reached before fracture then plastic collapse could occur. Obviously, with the given stress-strain curve without work hardening, the cross section with the notch cannot carry any more load once the entire cross section is yielding. ‘because the yielding will continue unihibited until fracture results, This is called plastic collapse. Thus, in plane stress where the stress in the entice cross section sequal to yield strength at the time of collapse, (Figure 2.11) the maximum load% is called the collapse toad or Figure 2.12. Progress fyi tbat notch Collapse: Pan = BW - avF,, ‘ais the notch depth, W’the (otal width and B the thickness. This failure load it load. The novainst sires it 8) the full width part is @ = P/BW. Hence, the part in Figure 2.11 fails when the nominal stress is: Wa “This isthe eg Collapse: a4 = Pau! = of a straight Fine (as 4, failure will occur already when the nomi F 9) oF the mote depth a. er = This failure sires is plotted as a function of notch depth in Figure 2.13. If fracture indeed 37 Pure 215. Net section collapse. occurs as a consequence of collapse, the strength, oj, would be the residual strength as defined in Chapter 1 IF the material work hardens, the notched eross section can carry a higher load. In general however, it cannot reach a situation where the entire cross section carries a stress equal to the tensile strength. This can be understood if itis realized thatthe strains inthe cross section are not uniform. Yielding starts at the notch root (Figure 2.11) and proceeds through the ligamest. Hence, the strains at the notch root are always much higher than elsewhere. Ths is depicted in Figure 2.14. Clearly, even though after there is sill a strain concentration at the notch. ‘As long as the stress is elastic there is a stress concentration given by o, = Katy I the elastic e886, 6 will be: Ce (2.10) ‘where the strain concentration, k,,is equal tothe theoretical stressconcentratin, 4k, = k,. When the whole ligament is yielding, the stress concentration has disappeared (k, = 1 in Figure 2.14b), but there is sila strain concentration. Neuber [2] writes forthe stress concentration factor k,, and for the strain concentrativn factor k,, and postulates that: kak = aay In this equation &, stands for the theoretical stress concentration factor, as discussed. In the elastic case k, = k,, 30 that both are then equal to k,. The3 (©) Stee sith ion factor, ky. is reduced when the material yields, but the strain concentration increases. Taking the case of Figure 2.14b where k, has become k, = I the strain concentration factor has become k, = K2/k, =] 12 A In the case of eg. k, = 3, the stress concentration decreases from Ak, =, = Bim the elastic case tok, = Vin the fally plastic case: the strain concentration on the other hand increases from k, = k, = 3in the elastic case tok, = KP = 9in the plastic ease Now consider a work hardening material in which the stres in the ligament ‘could be raised somewhat above F,,. Some stress concentration will remain (Figute 2.14) but the strain concentration wll become very high. As the whole nent is plat, even testi at te ae eg is dove yl bt he ‘notch the (plastic) strain is many times higher. Eventually the Sin ican ctr long ttre the test rer away be reached ales much above F,. Once fracturing has begun a much sharper notch has formed, so that the situation becomes worse and fracture continues. Given the stain distribution and the stress-strain curve, the stress distribu- tion canbe skeched as shown in Figure 12.14. When tearing or plastic collapse commences atthe notch tip, the stresses in most of the ligament are sill close » {to F., because the strain gradient is very steep. Thus, even ina strain hardening ‘material and plane stress, the (average) stress in the cross section cannot reach F... where F, is the tensile strength (see stress-strain curve in Figure 12.14) Apparently, collapse will occur at an average ligament stress somewhat higher than F,, but less than F,. The average ligament stress at which collapse occurs is called the collapse strength F4. Note that for a non-waskhardening material Fa = F,.and that al best Fay = Fy. Ibis not very well possible to determine the values of F, and F. other than by a (tensile) test; similarly, it isnot very well possible to determine F,y other than by a test on @ notched sample. However, its value depends upon the severity of the notch (k, and k,). For circular holes with k, = 3, the collapse strength is often close to the tensile strength (Fa © F.), bul for a sharp crack, the value is very close to the yield strength (Fa = Fy) 1K follows that for a work hardening material Equation (2.9) changes into: 2.12) ion of notch depth (Figure 2.13). If there is plane sirain, or in general non-plane-stress, the stress distribution afte yield is not uniform (Figure 2.12) Since the stress peak is local, the average stress in the section cannot become much higher than in the case of plane stress. In most cases — depending upon notch acuity - Equation (2-12) sil applies. 11 should be noted that the above discussion was strictly for the case of ‘uniform applied loading. If there is bending (or other stress gradients in the applied stresses), the conditions for collapse are lighily more complicated. The fully plastic stress distribution in the ligament forthe case of bending is shown in Figure 2.15e. The maximum bending moment occurs when the net setion sress is equal to F.4. Taking the moment around point A one obtains: tras = mre SE) = bean oF ay Using the collapse strength of eg. S0ksiand W’ = 2in, B = 0.5in(Fi and a = 0.Sin, the collapse moment is: Man = 1% 50 x 05 (2 — 0.57 = 14.1 inkips = 14100 ints. One can establish the entire collapse curve by solving Equation (2.13) for various crack sizes, as shown in Figure 2.15d, ‘Analysis of combined bending and tension requires determination of the point of stress reversal, D, in Figure 2.1Se. Since the stresses in thecross section hhave to be in equilibrium with the external load, one can establish 2 equilibriuma Tj iP 2a tind ee DUD | LO) = HF | ft Mo amd @) ‘eum? . , . “) 7 Sf. o Fire 215, Colpse analysis (a) Center crack: () Cola tet: (6) Sess dsb for 1M. (4) Collapse strength: (e} Combined beng 41 ‘equations, one for P and one for M. For a given P one finds Dand then M by solving the equilibrium equations. This is shown in Table 2.1 ‘Generally speaking, in engineering all solutions must be obtained in terms of the nominal applied stress, as was done going from Equation (28) to (2.9). For the bending case the same can be done, because the nominal bending stress at Myo: i8 = 6 Myy/BH*. Substitution in Equation (2.13) leads to: eae oe) a9 [Note that in contrast to Equations (2.9) and (2.12), the Equation (2.14) is not ‘straight line, but a curve (Figure 215d). For complete derivation see Table 2.1. 2.7. Fracture at notches: brittle behavior ‘The collapse fond or limit load is the highest load that can ever be reached, -vaximum load carrying capability is defined by the collapse load. ' ‘ualows automatically when the collapse conditions are reached, However, fracture may occur already before the collapse conditions are ‘altuined, (This is the concern of fracture mechanics analysis as discussed in ‘Chapters 3 and 4). In cases where fracture occurs due to collapse, the residual strength (Chapter 1) is determined by Equation (2.9) of (2.14): 04 = 6. The residual strength can never be higher. If fracture occurs before collapse, the residual strength is lower than the nominal stress at collapse (@,q < 64). Strains and, up to point, stresses are higher at the notch than anywhere else. ts the ease of very high k, this may lead to local fracture long before the remainder ofthe section reaches collapse. As a mater of fact, with very sharp notches (cracks) the condition for fracture at the notch root may already be met while virtually the entire notched section is stil elastic. Once fracturing start the notch has beome longer and sharper, so that the fracture proceeds throughout the ligament. Cracks due to fatigue or stress corrosion constitute very sharp notches: in some cases a cracked part will be capable of reaching the collapse load, but in many cases fracture occurs at much lower loads (stresses). This problem will be discussed in detail in Chapters 3 and 4. Nevertheless, collapse, iit occurs before the fracture condition is reached, will cause failure ‘According to Equation (2.9) collapse occurs when the stress in the full section hhas not yet reached the yield strength, Thus all plastic defogmation is confined to through the notch. If the collapse strength is higher than the yield stcgth Equetion (212) applies. In that case the stress inthe ill ection could reach or exceed the yield ifthe collapse stress is high and the notch size small In general however, the siress at failure in the full section will stil be below yield. Thus, in the vast majority of fractures occuring at notches and cracks, plastic ion takes place in the notched section only, in particular if fracture owigy ‘foregeipsecyrtions gh ataje=t Chayere Sangsd2 Mow o From Moment quia around 4 we Cope a = 10 oF M = 0250 = of ha Sette fr A = 6 ton = 13(% 54) © Not qudaccuao ey = 13 fre = coup caesintnne = LS Flat sin tension. Combined bending and tension (ending de to excentiity: unit thickness, in comer; hence ecentricty with respect to cracked section. Point (i. %) i unknown orizonal eqiitiom: oad in P= a OW = W~a~ sha o Moment equilibrium point 8 w " orem =o © Solve (8) and (©) to obtsin ay = 05. W — 025 GH = KAW a), and more important, oy = (al + 03 JA ~ Ra a) hg “a ‘The limitation of yielding tothe notched section has considerable-ffect on the {otal elongation at the time of failure. In a normal tensile test on an unnotched sample the strain at the time of necking is often on the order of 10% of more. Thus the bar has become 10% longer (Figure 2.16). Di 8 tll higher strains occur. In a notched sample plastic deformation occurs only in the notched section. Even if this small section would deform 10% before fracture, the total elongation would be much smaller than of the unniotched bar (Figure 2.46). Consequently, elongation after fracture is hardly noticeable. If litle overall plastic deformation occurs, the fracture is called “brittle” from an engi- neering point of view. The text in this book strictly adheres to the engineering, use of the word. In that sense virtually all service fractures are bitte: fractures ‘occur almost always at notches or cracks. Thus plastic deformation is always restricted to the notched section and the fractures are brittle in the engineering sense. The fracture mechanism may be either by dimple rupture (aften called ductile) or by cleavage (often called brittle: see chapter 1), but the great majority of fractures occurs by rupture; from an engineering point of view they are britle because of litle overall deformation in the cracked section (Figures 2-16), {os | Bel a Figure 2.18, Apparent “iene do notches aed cracks as x consequence offend pase ‘eformaton in are of high set,2.8, Measurement of collapse strength ‘The casiest way to measure the collapse strength, Fa, is by means of a center cracked panel. In view of the fact that constraint ‘will still play a role, the specimen should have a thickness close to the thickness used in the structure of lerest. As this may lead (0 unwieldy specimens. a compact tension specimen (Chapter 3) can be used where ciccumstances so dictate, provided itis analyzed for combined bending and tension (Table 2.1) 1 should be noted here, that in the case of center cracks, the definition of crack size is 2a instead of a. This is a convention, by which all cracks with one tip are defined asa, all cracks with two tips as 2a (see uso Chapter 3). Note that with this definition Equations (2.9) and (2.12) become for the center eracked panet: W — 20 %% r Fa Qs) ‘The specimen may be fatigue cracked, but blunting before collapse usually is 30 extensive that a sharp saw cut and a fatigue crack arc undistinguishable at the onset of fracture. The load at fracture (maximum load) is recorded. From this, the net section stress can be calculated directly as: ky = Pine We Note that the measured Fy should be greater than or equal tothe yield strength, “otherwise fracture occurred before collapse was reached and the measured net section stress is not the collapse stress. In that event the test should be analyzed using fracture mechanics as discussed in the following chapters, Instead of calculating the net section stresses directly, one may calculate the remote stress, o = P/A, atthe time of fracture, which isthe residual strength. By plotting the latter in a diagram such as in Figure 2.17, and by drawing a straight Tine through the data yt nnd the point (W, 0), the collapse strength Fu, is found at the intercept of this straight line with the vertical axis, as can be seen from Equation (2.15). The advantage of this procedure is that results obtained from specimens with different crack sizes can he plotted as well, thus providing ‘a check of the applicability of the collapse crit straight line as per Equation (2.15) then collapse occu not, the tests below the line represent cases where fracture oct collapse and these tess shoitld be analyzed using fracture mechanics concepts (Chapters 3 and 4) Figure 2,17 presents test data (3] for 304 stainless stcetshurwing « eollapne strength of 67ksi. (The yield strength wis s0-vial is 30Ksi and the ultimate strength is 90ksi). Indeed, the collapse slicagth is considerably higher than e . Feat o we "ts 4s46 yield. For all eracks smaller than S-in the residual strength is more than 30ksi ‘This means that the entire panel was above yield (e, is the nominal remote stress) and not just the net section. Data points at various crack sizes all fall on (or close to the straight line, thus confirming that collapse occurred in all tsts. Other metal alloys usually have a collapse strength much closer to, or equal to the yield strength; in this respect the example of 304 stainless steel is not representative 29, Exercises 1. Calculate the theoretical stress concentration factor of an elliptical notch witha semi major axis of 2 inch perpendicular to the applied load, and a semi minor axis of 0.4 inch, What is the stress concentration factor; what is the strain concentration factor. What are these factors after the notched section has fully yielded in plane stress, assuming no strain hardening? 2, For the case of Exercise | calculate the nominal stress inthe full section at the time of collapse ifthe yield strength is SOksi. Do the same for a material witha collapse strength of 75ksi. Calculate the fracture load in the two cases if the notch is a center notch and W = 12 inch, B = 0.5 inch, 3. Cuta strip of paper 4 inches wide and 10 inches long. At three inches from Cone end cut a circular hole of 1.6 inch diameter in the center. At three inches from the other end cut a shit of 1.6 in long and (wo holes at both ends of the slit of0.2 inch diameter. After cutting the holes cut the notch into an elliptical shape (semi minor axis 0.2in). Calculate the stress concentration factors of the two notches and predict where the strip wil fail. Roll the ends around 2 pencils and pull the strip (failure to prove your prediction, 4. Ia crack appears in service one sometimes drils a 50 called stop hole at the ‘cack tip as a temporary repair. Suppose a crack has started at the edge of a strip. Ils size isa. A stop hole is drilled with diameter a, its center exactly ‘on the crack tip. The erack tip root radius is almost zero. Assume the crack th the stop hole to be an ellipse (why is this permissible). Calculate the ‘theoretical stress concentration factor before and after stop drilling for the ‘general case. Ifthe crack is I inch long calculate which size stop hole is needed to give a theoretical stress concentration factor of 5: and 7? 5. What is the stress at yield at a notch in plane strain if F, = 30ksi? What is it atthe free surface of the notch root? Sketch the stress distribution, 6. Using the solution of Exercise 5 calculate the nominal stress the full section at which yielding begins at the free surface of the notch. Sketch the stress distribution in these two cases. Assume k, = 4 a 1. Consider a cylindrical bar with a circumferential groove of hal circular shape. In order to reduce the stress concentration one sometimes machines relief grooves: grooves but of smaller depth at short disiance on both sides of the main groove. Demonstrate by means of load-flow lines that this indeed gives a lower stress concentration factor. 8. Using load-flow lines demonstrate the high stress concentration factor (as high as 10) at the fillet radius of the head of a tension bolt. 9. Calculate and construct the diagram for fracture due to collapse for a center cracked panel as welt as for an edge cracked panel. W = 600mm; Fas = 350MPa. At which stress will collapse occur ifthe panels contsin a erack of 150mm? For the edge cracked panel assume that the side edges are con strained (kept straight, so that no bending occurs). References 10) RE Peenon, Sr coment desig fcr Sohn PI Neer Tey teens flr shea soles pst aes wih ai non-linear stress-strain law of J. App. Mech. 28 (1961) pp. 544-550 " 1 SUF Kaien, a. Towards an caine ate mci opty fo ree Nr En wD) S08CHAPTER 3 Linear elastic fracture mechanics BL. Scope In this and the following chapter, the fracture sechanies concepts for the analysis of fracture will be discussed. With theye eon =p ill be prcsite 10 ‘obtain the residual strength diagram and the maximum permussible crack size (Chapter 1). Following the arguments used in Chapter 1, the discussion will be limited to mode I loading (mixed mode +s considered in Chapter 9). Materials with relatively low fracti Tail below theit collapse ‘strength and can be analysed on the bass of elastic concepts through the use of Linear Elastic Fracture Mechanics (LEFM). Such materials are, among others, practically all high strength materials used in the aerospace industry, high- strength-low-alloy steels, cold worked stainless steels, etc. For the fracture analysis of many other materials, the use of Elastc-Ptastic Fracture Mechanics (EPEM) as discussed in Chapter 4 is often indicated. Understanding of EPFM ‘concepts requires familiarity with LEFM concepts. 3.2. Stress at crack tip ‘Consider (Figure 3.1) a body of arbitrary shape with a crack of arbitrary size, subjected (0 arbitrary tension, bending, or both, as long as the loading is mode 1 (Chapter 1), The material will be considered elastic, following Hooke's law. For such a case the theory of elasticity can be used to calculate the stress field ‘The stresses, c, and o, can be obtained as well a: the shear stress ¢,,. Details of the derivation can be found in more extensive texts (1, 2, 3]. Inthe following ty the solution will be given, and it will be shown that the result is in accordance ‘with what would be expected, ‘As was explained already in Chapter 2, the crack tip stress field is at least biaxial (load-flow fi if contraction in thi inection is constrained. Hence, there will be stresses in at least X and ¥ direction, @, and ¢,. From the stress field solution it appears thatthe stresses on 0 ‘a material element as shown Figure 3.1 can be described by (in the absence of constraint): «= font sndan = Beot(iande) | Beers on coe) fu = gs e08§ sin 5 08 5 Indeed, both a, and a, exist (as should be the case as discussed in Chapter 2). For the case that 0 = O(plane through the cracked section), the shear stress, t, is zer0, as should be expected for a plane of symmetry. Although it does not make any difference for the following discussion, itis convenient to confine the considerations to the plane through the crack with @ = 0; in that case the functions of @ will be ether 0 or 1, s0 that they essentially disappear: (note also that x = r for 0 = Oy: a. ce (3.2) ek 2 ie appears that, atleast along the plane ¥ = 0 for which Equations (3.2) hotd, the transverse stress, is equal in magnitude to the longitudinal tess, ¢,. The streses depend upon the distance x from the crack tip; note that at grealer distances (larger x) the stresses are lower. The stresses also depend upon Parameter K, which is a8 yet undefined ‘The expressions forthe stresses appear to be remarkably simple. Even more remarkable that a tes feld solution could be obained at all, since the body, crack, and the loading are arbitrary. Because of this arbitrariness i is not surprising that the equations contain an unknown parameter K. As Equations (2.2) are for an arbitrary case, they must describe the stress at each and every crack in each and every elastic body in the universe. Apparengly, the equations are universl and can be used forall crack problems. Thersfore, also the Parameter K wll appear in all crack problems, reason to give it 4 name: K is Called thesreaintenity factor, no to be confused with the stress concentration factor. ki. discussed in Chapter 2. Simi ined for other modes of loading (Chapter 9), beFie 3 mode Uoadiog. that the goniometric functions are different for different modes. A stress intensity factor appears in all solutions, but the definition of K is slightly different for different modes. For this reason, the siress intensity factors are ‘often labeled in accordance with the mode of loading (Chapter 1), namely Kj, Ky, and Ky. In the above equations K should then be labeled as K,. However, when there can be no confusion about which mode is meant, the subscript is ‘often omitted; this is what will be done in the following. Should more than one loading mode be operable, the labels must be carried to avoid confusion, as will ‘be done in Chapter 9. ‘Since Equations (3.2) are forall crack problems, there is no objection ¢ geometry. Let this be a very large (infinite) panel, subjected iform unaxial loading with a nominal sess ¢, and a central erack, a8 in Figure 3.2a. The size of the crack will be called 2a. This is a convention: In fracture mechanics all cracks with 2 tips are called 2a, all eracks with one tip ‘There if no reason why it could not be done differenily, but this convention is ‘necessary to compare notes. The convention should be adhered to if one wants, {o use data generated elsewhere, otherwise erroneous results will be obtained. Equations (3.2) apply to the problem of Figure 3.2a, and it is now possible to examine the significance of the stress intensity factor. K. It should he noted then that the stresses everywhere in an elastic body are proportional to the applied load. If the load is increased by a factor 2. all stresses everywhere will ed = cy Fine 42. Cte rack with uniform loading (Infinit plate: () Brite pt. increase by a factor 2. Thus, it can be concluded that the crack tip stress must be proportional to the applied stress, ¢, ie. (see Equation (32)) 3) I stands to reason that the crack tp stress wl alo depend upon crack size The sresses wil cetinly he higher when a lager, Rene the cack stag, Ist pea in he numerator in Equation (3.3), Tees oly anc vay ine thiscan happen. Hoth sides of the equation must have the dineason of ene but in the denominator tthe right there appears a squat oo feng te Aistanc x. In order o cancel he square 10 of lengih nthe enone, the crack size mus appear inthe mumeator square fot of G4) Equation (3.4) tll contains a proportionality sign instead of an equal ( an equal sign, because dimensional analysis does not show whether there isa dimensionless ‘number involved. Calling this dimensionless number C, one finaly arrives equation: es) Clearly. itis simple 19 find the format of Equation (3.5), but a formal solution UL, 2, 3] would be necessary to obtain the actual value of C. Tt turns out that C= J, for the case depicted in Figure 3.2a. Hence:32 36) Comparison of Equations (3.6) and (3.2), leads to the conclusion that for the configuration of Figure 3.2a: K = ora en 3.3. General form of the stress intensity factor ‘The manner used in the previous section to demonstrate the significance of Kis. not limited to the case shown in Figure 3.2a. For example consider (Figure 3.2b): the plate of finite width W. From the same arguments as used abeo= it follows that: Cola GR . = ae {In this case it must be expected that the size ofthe plate also will affect the crack tip stresses. It must be anticipated that the stresses will increase when W- ‘becomes smaller. The only manner in which the effect of W can appear isin the factor C. Hence, C must bea function of (depen! yas she width W. However, C must be dimensionless, and hence, it can dejend upon W only through dependence upon a dimensionless parameter such as W/a or a/W. For the ‘configuration of Figure 3.2b the expression for C°appents to be [I] Ir see *2 as ,- Vs G.10) a Er 0 that and ICV is very large, or a very small, the value of Jace a7i7) accordance with anti vin, because for small a/1V the configuration of| 3.2b is identical to than of Figure 3.2a, Indeed, in that case Equation (3.11) reduces to Equation (3.7). "Now it becomes obvious that for ANY configuration the crack tip stress will a stopte DN em ‘ Vine Vinx Ke (a ox wr sa aie) enh dinenson dering he rot ote racked {In the practical use ofthese equations all C's are divided by /n, and a is stituted for vt to compensate. The function C(a/L//m is then renamed f. the geometry factor (often erroneously called correction factor}: o, = Befee KL 1 Jie Sine = (2) ei Note that Equations (3.14) are identical to the previous equations. The F.quations (3.14) present the stress and the stress intensity factor in comparison to those for the infinite panel: f= 1 for the infinite panel, but for the finite idth panel G14 Gus) He must be emphasized that Equations (3.14) represent the erack tip stresses and stress intensity for all crack problems, the ing been derived from the general solution for an arbitrary crack in an arbitrary body with arbitrary ‘mode | loading. For any crack in any practical problem only the function B, oF its functional value, need be derived. For many configurations the function, has been calculated already: the results can be found in handbooks [4, 5, 6) Examples of 8 functions for a few commion crack cases are showr in Figure 3.3. Note that in the Equations (3.1) through (3.14) the stress, ais the nominal stress in the uncracked section. The fact that the stresses are higher in the cracked section when eg, W becomes smaller, is wholly accounted for by fl. Although hhandbooks are a source for f! expressions for many generic configurations, ‘obtaining f for a practical problem may be fare many simple ways in which it can be obtained with good accuracy. Such Simple procedures to derive f for practical cases are discussed in Chapter 8, At{1ttt TTTTT i BV I. L/We2 ony: 12-023.$ 41086(8)" 217 Basie vos $4.70 8)" + 15.6 ‘Figure 3.3. Examples of Munctions 55 this point it is sufficient to note that can always be obtained and that Equations (3.14) are universal. 34. Toughness Several difficulties arise with the direct use of Equations (3.14), and these may have serious consequences, as discussed in subsequent sections. Nevertheless, although somewhat prematurely, the practical use of the equations for the analysis of fracture problems, will be considered first, be it that some of the conclusions will have to be modified Fracture will occur when the stresses at the crack tip become t00 high for the ‘material to bear. As the stress intensity factor determines the entie crack tip stress field, the above statement is equivalent to: fracture will occur when K igh the stress intensity can be depends it must be determined from a test. For example let test be performed on a 30 inches (762mm) wide plate of a certain steel X, with a central crack 2a = 4 inches (101,6mm), Let the plate hhavern thickness of .2 inch (5.08 mm). The plate is pulled to fracture in a tensile lest machine. Suppose that fracture occurs at a load of 180 kips or 180.000 Ibs (800.000 N). The nominal stress in the uncracked section atthe time of fracture was then 180/(0.2 x 30) = 30ksi (800000(5.08 x 762) = 207N/mm? = 207MPa), This information permits calculation of the value of the stress intensity at fracture. Note from equation (3.15) that f= 1 for the small a/W’used, and that ‘a = 2in (0.0508 m), Hence: Ke 1x 30x fexd = 15ksi vin Kom 1x 207 x Je OOH = ASMP a Jim Kom 1x 207 fe xT = 2620Nimm™ tly, when K reached the value of 75 ksi yin (83 MPa Hquations (3.2) and i is the ‘same in plane stress and plane strain, because further away from the erack tip ‘constraint is not a problem (Chapter 2). Apparently, within some distance r,, from the crack tip the yield conditions are met, so that there is always a small area at the crack tip where plastic deformation takes place (Note that o, infinite according to Equation 3.14). This area is called the plastic zone. 1 possible to estimate the distance from the crack tip over which plastic deforma- o ‘igre 25, Ses Disbation (,) a crack tp (e Plane tes (6) Pane sein. tion occurs, i¢, the size of the plastic zone. For plane stress this follows from the Tresea yield erterion (Chapler 2) and the condition 0, 1nd for plane strain from the condition a, = 3F,. Using Equation (3.2) for 2, one oblains:“ K Seo oon Fer, a 02» for plane stain one Plane strain: Fo 3Fy OF te ™ care {mn realty, the plastic zones are about twice as large as in Equations (3.23). This, is due to the fact that actually a much higher stress should have been carried by the plastic zone. This extra stress must be bypassed around the plastic zone thus increasing the stress there and causing more material to yield {1}, By limiting the lip stresses because of plastic deformation, the dashed part of the sess tion in Figure 3.5 was essentially eliminated. However, the total stress mm including the dashed part, was calculated from the theory of elasticity on the basis of equilibrium. Cutting off the dashed part is violating equilibrium. This must be mended by restoring equilibrium, The dashed areas in Figure 2.5 represent a load which cannot be carried by the ‘material in the plastic zone because the load carrying capacity of that materia is limited by yielding. However, the cut load must sill be carried through (cquilibrium) and it must bypass the crack, or it must bypass 1, a8 the plastic zone cannot carry it. This implies that yielding actually extends beyond r,. (Note thatthe high stresses were due in the frst place due to a load path interruption by the crack, so that the load had to bypass the crack as shown in Chapter 2. [Now it appears that the load represented by the shaded areas in Figure 3.5 mist bypass further away, because once itis loaded \« yield, the ares r, is no longer 1 viable load path either). The consequence is that the plastic zones are actually larger than those represented by Equations (3.23). Restoring equilibrium will lead to plastic zones exactly twice as large (I), as those in Equations (3.23). For ‘other reasons {I}, the plastic zones are usually taken as oe Oa G24 with a = 2 for plane stress and a = 6 for plane strain 1 appears from Figure 3.5 and Equations (3.23) and (3.24) that the plastic zone in plane strain is much smaller than in plane steess. Also, the crack tip, stresses are higher in plane strain than in plane stress. The longitudinal stress, 0,,dilfers by a factor of 3, while, = 0.660, in plane strain anda, = 0 in plane slress (see also Chapter 2) In the example of the hypothetical test given in the previous section, fracture {ook place ata value of K of 75ksi Jin. Supposing that the material had a yield strength of F, = 100k, the size of the plastic zone at the time of Fracture would have been according to Equation (3.24) o 18 = Sree ~ 2 inch 1 = Baap 7 008 inch “These are very small plastic zones indeed. At this point it cennot be decided whether there was plane stressor plane strain inthe exainple inSection 3.4. This question is discussed in the following section, plane stress: 1, plane strain 3.6. Thickness dependence of toughness ‘The true size of the plastic zone is not important for the following discussion, bbut the format of Equation (3.24) is relevant; it shows that the size of the plastic zone depends upon K only. Thus forall cracks in any configuration, and at the ‘same K, the stress distributions atthe crack tip will still be completely dictated by K and by K only. It follows that the fracture criterion then still holds regardless of the plasticity; if fracture occurs ina test specimen ata certain value (of K (the toughness) then it will do so in any other configuration at the same K, ‘because in that other configuration the stress distribution as well asthe size of the plastic zone will stil be the same as in the test specimen, provided there is equal constraint ‘One problem has arisen however. The stress distributions are not the same plane stress and plane strain; as shown in Figure 3.5. Also a, dilfers in the two ceases. Asa consequence, the above fracture criterion cannot he used indiscrimi- ately. If fracture in plane strain occurs ata certain value of K, then it will not in plane stress at that same value of K, because then the sires distributions will be different (different, and different ¢, atthe crack tip due to different plasticity as shown in Figure 3.5). I follows then that the toughness will depend upon the state of stress. As the latter depends upon thickness, the toughness must be ‘expected to depend upon thickness Ii should be noted here that Equation (3.24) was derived using only the first {crm of the stress distribution in Equation (3.19). Should the plastic zone be so targe thatthe other terms come into play then it size will deperd not only upon but also upon C, Detc. In such a situation equal K is not enough to guarantee equal stress distribution and fracture is not likely to occur at always the same value of K. Apparently. for the fracture criterion to be useful the plastic zone ‘must be very small. Since r, depends upon K, this implies thatthe stress intensity al fracture. i. the toughress, must be low. As r, also depends upon the yield strength, the fracture criterion will be better applicable to materials of high yield strength and low toughness. Although the criterion does not work when the yield strength is very low and the toughness very high, satisfactory results can bbe obtained for many such cases, provided one realizes that plastic collapsea (Chapter 2) may prevail. This will be discussed in detail ina later section. In the "we of extremely high toughness andjor low yield strength, EPFM may have to bbe used (Chapter 4), Consider two plates, a thin one with plane stress, and a thick one with plane strain. Both have cracks and are loaded to have equal stress intensity factors, In the case that the prevailing stress intensity is equal to the toughness of the thick plate fracture ensues in the thick plate. The stress distributions in the thin plate will be much more benign (see Figure 3 § snd note the difference in.) at this same K, so that it must be anticipated that the stress intensity in the thin plate can be further increased; ie. the expectation is that the toughness is higher in plane stress thar: plane strain (Figure 3.6). In plates of intermediate thickness the state of stress gradually changes from, plane stress, via more and more triaxilty, to plane strain in thick plates, ‘Therefore, the toughness will not suddenly. «i ysrticular thickness, drop from its high plane stress value to its low(er) ys.» strain value, but the change will be gradual; ie. the toughness decreases wil: sickness as depicted in Vigure 3.6b. Beyond a certain thickness there will heptane strain. As the state of stress ‘cannot become worse than plane strain (complete constraint of contraction) the toughness will not decrease further For plane strain cases the toughness is called the plane strain fracture toughness. It is generally denoted as K, i thc vsaual value of the mode 1 stress intensity, Ky, at which fracture occurs. (Compare: Fis the critical value Of the stress at which yield occurs). Similarly one speaks of the plane slress fracture toughness, and of the transitional acture toughness for intermediate situations, Plane «iss od transitional toughness are generally denoted as K, ‘oF X;,.. Unfortunately, this introduces a confusing inconsistency. In plane ress or in transitional cases, the loading is sill mode 1, the toughness is still the critical value of the mode I stress intensity. Thus, the general denotation of toughness should be Ky, regardle-s af the state of stress, One cowl? k of THE | a Wor the value is for plane si. plane strain, oF tas wt as it may, the above der become the ‘accepted’ ones. Once the user gets used to these denotations, they are, in a way. helpful. Ifthe toughness is given as Kit is immediately clear that this toughness is for plane strain; while a toughness K, ot K,, is obviously for a thickness in which full constraint does not occur (plane stress or transitional), The toughness as a function of thickness can be measured in tests of the type discussed in Section 3.4 on plates of different thickness, I can then be used 10 calculate the residual strength of eracked structures: Fractureif K = K, or A = Ay ie, Practueif: foJaa = K, (or Ky) 0.258) a K x @ > Thickness Prewe 24 Dependence of ougiess upon thickness. (2) fect of thickness om plastic zone and Mateo sec (8 et of thickness on tourhoessCy from which the fracture strength follows as (3.250) For the configuration at hand can be obtained from a handbook or ‘otherwise (Chapter 8), and the residual strength calculated provided one uses the toughness for the thickness at hand, in the same manner as discussed Section 3.4, tis of interest to know at which thickness full constraint and plane strain ‘occur. As discussed in Chapter 2, plane strain develops when the roll of material at high stress at the crack tip is long and thin (Figure 3.7). Hence, L/D, where Lis the roll’ length and D its thickness, must be large e.8 larger than a certain value Q. When the crack is all the way through the thickness, the length of the roll is equal to the thickness, B, ie. L = B. Taking D equal to the size of the plastic zone, D = +, from Equation (3.24) one obtains as the condition for pane stain L 8 3° Wap? 626) In this equation a and Q are numbers. Rewriting Qj = g, the condition can be Ki a> ait am ‘The value of g cannot be derived mathematically. tis obtained from toughness {ests on plates of different thicknesses as described, and by finding the thickness ‘a which the curve levels off. Such experiments have shown (7] that this happens for g = 2.5, but the results are not very conclusive as ean be seen in Figure 1.8. “The scatter is considerable and the most that can be said from Figure 3.8 is that the leveling off occurs somewhere between g = 2and q = 4. Nevertheless the number 2.5 i often considered sacrosanct. The reader is advised that itis but rough indication (not: instead of ¢ the symbol is used in Figure 4. I should be emphasized that the thickness is only of relevance for through: the-thickness cracks. Constraint is determined by the length of the roll of highly stressed material, In the case of a part through crack such as in Figure 3.9, contraction is always fully constrained and the thickness has no relevance to the problem (note thatthe thickness came into the problem because the length L of the coll happened to be equal tothe thickness in the case ofa through-the-thick: rest crack, and in thal caze only; thir was discussed already in Chapter 2) Tigre 37, Thicker and Sap of Sires. 3) Aterpt fo om thickest upon contraction tio tea Yet of = o 65{epparent) Ric (Vi) 120] 4} 109} - PSL ie eo ‘aesih ol Frere 38. Ect of eicknes on measured Koa Maraeing Set 7 Figure 3.9. Constant condition for pa hough cracks ‘Therefore part-through cracks (surface flaws and corner cracks) are in plane strain, at least in the center. For such cracks, one must use the plane strain fracture toughness, K,., regardless of the thickness. o 3.7. Measurement of toughness ‘Hhprinciple the toughness can be measured on any kind of cracked specimen. Fracture occurs when Equation (3.25) is satisfied, With knowledge of f for the specimen,at hand, and the measured fracture stresg (load) the siress intensity at fracture can be calculated; the result isthe toughness, Inthe reverse operation, Fi follows from Equation @.25b), Although any kind of specimen can be used afew simpletest specimens have been standardizedjby ASTM [7,8] forthe measurement of plane stain rectors toughness. The most universally used isthe so-aled compact tension (CT) Specimen depicted in Figure 3.10, It contains a notch in the form of a chevron is a 28 Peasossw { totgue ie Pare 310, Sanne , pce: eons: gh: ath, oe 8 et ' o86 ‘and is loaded through two pin’s ina tensile machine. Cyclic loading is applied to introduce a fatigue crack. When the erack is atthe desired length the cycling js stopped, and the load is raised until fracture occurs. The stress intensity at fracture can then be calculated; i. the toughness obtained factor for the compact specimen is usually written as: x= aol) — ws(Q)" +0069) = 1017()" + 9(Z)" | 6%) ‘This expression seems to be rather diferent from Equation (3.14), but it is essentially the same. Iti waitten in this manner because the nominal stress is hhard to define. However, there is no objection against defining a (somewhat hypothetical) nominal stress as@ = P/BW. Then Equation (3.28) indeed reverts to: re [er 14e() + a +361) Jov = pasa. om uring the test the so-called crack mouth opening displacement is measured by inserting a gage in the notch (Figure 3.11), Since there is hardly any strain inthe material between the loading pins. the crack mouth opening displacement is ‘essential equal to the displacement ofthe louding pins. Load and displacement are plotted in a lond-displacement diagram (Figure 3.11). The maximum load is substituted in Equation (3.28) to find the toughness, ‘The test standard prescribes that the load-displacement line must be straight cor nearly so. When extensive plastic deformation occurs the diagram wil become non-linear. However, yielding is often confined to such a small plastic, zone that non-linearity is hardly perceivable. Should the diagram be significant. ly non-linear then the plastic zone is large: a discussed in the previous sections the fracture criterion based on K is then no longer valid and “the” toughness ‘cannot be defined. Hence, the standard sets a limitation on now-lineaity for the measurement of ‘valid’ toughness numbers. Should the entire ligament yield then collapse may occur as discussed in Chapter 2 ‘The standard test is specifically for the measurement of the plane strain fracture toughness K,. This means that there must be plane strain. Therefore the thickness of the specimen must be suffi ‘sity Equation (2.27) with q = 25. The toughness is not known in advance (otherwise measurement ‘would not be necessary), so that one cannot determine how thick the specimen should be, Its possible that the specimen was made too thin, in which ease the o Soe oy au Pon 11 Standard ky Cervatre de 1. (0) Clip gage mounted in specimen: () Singh — 8 record (e) asym growth measured toughness would not be the plane strain fracture toughness. For this son the toughness obtained from Equation (3.28) is initially called the widate value. Ky. ofthe pane sean fracture toughness. Ke must check xB > 25KS/ES. I this the case then indeed K,, new test must be done om a thicker specimen if one insists on knowing the plane perfectly useful also Chapter 7), ‘As was pointed out already, the number q = 2.5 is rather dubious, and therefore rigorous application of Equation (3.27) somewhat ludicrous. The ‘number 2.5 was obtained from an interpretation of a limited numberof test data (Figure 1.8). It was then agreed upon by a committee that the number was reasonably representative, and as a consequence it appears in the test stand: ighness number fr he actual thickness ofthe specimen. (See [8]. However, a committee decision, nor an ASTM standard, is @ sufficient guarantee that the number is indisputable. If there is general concurrence that the number is useful then there are no objections against its use, but it sil emains an arbitrary number. Also the uncommon significance attached to the standard testis largely exaggerated. Toughness can be measured on any cracked specimen, provided the plasti zone is small. fit could not, the result ‘could not be used to calculate residual strength of any other structural crack (based upon the generality of Equations (3.2)), which is the purpose of the ‘measurement. In that case the standard test would have no use in the fist place. {See also Chapter 7).nm Although the compact specimen is cheap and simple, it is totally unsuitable for the measurement of plane stress and al toughness. For such toughness measurements a center cracked pane! is the most advisable. In order to obtain useful numbers, the plastic zone aé fracture must be kept smal. Thi often requires very large panels, a8 will appesr is the Following section, Due to the fict that stable fracture often ;ecec's icacture instability, « revised definition of toughness is needed as weil (Section 3.12). 3.8. Competition with plastic collapse In Chapter 2 the conditions for collapse were discussed. It turned out that collapse occurs in a center cracked panel when: W 20 % wre 3.30) In this equation Fo the collapse strength which may be low as the yield sirength and almost as high a the tensile strength. Recall from Chapter 2 that Equation (3.3) is straight line asa function of erack size a. ‘Also discussed in Chapter 2 is the fact that Equation (3.30) represents the absolute highest load carrying capability: at collapse plastic deformation becomes unbounded and fracture follows, regardess ofthe toughness. Hence, (fracture) feilure by collapse may occur before K reaches the toughness. Naturally, Kis the result of a semi-lastic concept while at collapse the entire acked section is yielding. Therefore, the two failure conditions cannot be direcly compared; but they can be in an indirect way. IF the toughness is high, Equations (3.25b) will predict avery high fracture stress (residual strength). Ths may be as high or higher than the ses for failure by collapse. Then failure by collapse will prevail, because, of two competing failure modes, the one which frst becomes possible wil prevail. This may hhappen when the toughness is high, but under other conditions as wel. Should the fracture stress a, be higher than the stress causing failure by collapse, then collapse wil prevail. This is the case when the result of Equation (3.256) s more than the resll of Equation (1.30), i. when (fora center crack: ap, Klor Ki) ean ‘The lowest of the two isthe actual residual strength discussed in Chapter 1 From Equation (3.31) it appears that there are 3 situations in which a collapse failure could prevail, namely when or © n ~ the toughness is very high or; = the crack is very small (a + = the width Wis very small These situations are illustrated in Figure 3.12. For the material with Ky = 160ksi Jin a panel, of 12 inch width always fails by collapse, but a 60 inch wide panel of the same material fails by K,. Apparently, for a particular structural configuration failure may occur by collapse (unizonditionally leading to fracture) in certain instances and by fracture in other instances, Collapse always prevails if the cracks are small, regardless of how low the toughness. At any (oughness Equation (3.25b) puts the fracture stress at infinite when the crack size approaches zero. Thus the residual strength curve will always rise e (2) Residual stent of 1m wide panel (©) Residual seenth of in wide pence,n asympotcallytinfnityfor sal this means that there willalvays be a point ata certain small where coll fallaes wil prevail regardless how low the toughness, ora crack size a ~ 0, the fracture equations peti that the facture sess wil become lite. Cleary, when @'=O the Factor sess equa 10 the imate tense srength For les. Thus, the mos! le-hand prt of the carve represented hy Equation (3.25) wl always e in eto, whether the toughness isigh or tow. For a = the strength is (or le). while for large the Equation (2238) apis (which sth curve in Figure 12). Obviously. the behavior Setween a= 0 witha = Fyvan the cue fo age a cannot be as vror Bin igure 12a, Nene, one assumes the eyeball curve C (angen tothe curve the asmpton cannot be far from te ths cet wil be adequate for enginerng analysis Tis important that of two plates of dierent sex - bt ofthe sme materia ~ one ma fail clase te cterby facture. Thai depicted aso i Figure 112 as mel asin Figure 13, From this ican fe conclded hat specimen fr the mensurement of toughness mst feof scent size Fin the ete of Fig 5.138 plac ofthe smaller size, would be used ina tes the faire woukl occur by collapse: Hence the vale ofthe tes intensity atte ime offre woul Ale lower than the toughness, beeuse Khas nye reached the oughnes "This problem also exist for plane stnin toughness testing with the CT specimen The specimen ay fail by colle i the toughens ery high of the specimen foo smal (specimen size rather ipnoed inthe test standard) Tone insisted tp cleat Re fom such atest by sing Equation (3.28). the Ke Pgure 11. Ect of pate size fore moe n ‘would be LESS than the actual K,. Fortunately, the test would be invali because the standard requires a ‘straight’ load displacement diagram. If the specimen were o fail by collapse, the load displacement diagram would not be straight because of large scale plastic deformation in the ligament. ‘Apparently, Equation (3.2Sb) is not sufficient to determine the residual -ngth dliagram, nor is Equation (3.25a) enough to determine the toughness. It always necessary that the stress for collapse be determined as well, because the lower of the two prevails, and the residual strength, 67, i the lower of a, and 1%. For short cracks, the tangent approximation, a, may prevail iit is lower than a, and 4,,.. Determination of the collapse curve is simple in the case of uniform tension loading. Ifthe loading is non-uniform such as for the compact, tension specimen. the collapse condition is somewhat more dificalt to evaluate (sce Table 2.1). The calculation of residual strength diagrams is discussed further in Chapter 10. 3.9. The energy criterion ‘This and the following sections present another look at the fracture criterion, This is of interest for understanding LEFM, but it is essential for the understanding of EPFM as discussed in Chapter 4. Finally, it will appear that the definition of toughness may have to be revised (Section 3.12). Conservation of energy demands that the work, F. done by the load on a body. is not lost. Its conserved as strain energy, U, So that Fu =0. 32) ‘The work done by the load is F = J P 45 (where P isthe load and 6 the load displacement), which isthe area under the lond-displacement curve. As long as the material is elastic. the load-displacement diagram isa straight line, so that the work done by the load is 175 (see Figure 3.14). It follows from Equation (3.2) that U = 4°8 as well, But U can be determined in a different manner. ‘Consider a small material element of nit size, subjected (0 a uniaxial tension 1.28 shown in Figure 3.14b. The stress does work to deform the material over de the total work done is f¢ de. which fora linear elastic material is the area of the triangle shown in Figure 3.146, namely fae. By substituting ¢ = o/E from Hrooke's law, the work becomes Jo'/E. As this work must be equa tothe strain energy. the value of Jo'/E also represents the strain energy in the material element The stress on each volume element strain energy in the entice body or 3 over the volume of the body: not always the same. Therefore the total veture is obtained hy taking the integral Um {ffSacara: x” baees © ‘Fire 3.14. Strain every. (0) Load and load aplacement diagram (6) Stes om material cement. in which « may depend upon x, 1"and =. In the ease ofa simple tensile bar the Stessisthe same oncachand every value clement, In ha ase the otal stain ‘energy is simply equal to fo'/E times the volume. If the cross section of the bar is A and its length L, then the strain energy is: a oan v= Zu With knowledge of F ( 0.32) to obtain 1P6) one can substitute both F and U in Equation 0 835) 1 ' ‘ ' ' 475 — ' ' 18 ‘By noting that, in this case, 8 = cL = @L/E, and P = aL.A, it can be readily seen that Equation (3.35) is true, As already stated, the criterion provides U = 40%, Faquation (3.32) also holds when the body has a crack, 2a. In the case of limited plasticity the load-displacement diagram is still a straight line as i Figure 3.15. Ifa crack of somewhat larger size, a + da, exists then it will ake less load to cause the same displacement: the stints is less. This, the load. displacement line is lower as shown in Figure 3.15d Let the lond increase 10 Py, the displacement to 3, as in Figure 3.154. In the event that fracture takes place at this load, the crack extends over a small increment da from a to a + da. During this process there must be energy ‘conservation, but instead of two energy terms F and U there is now a third nergy term W, where IV is the work expended in fracturing material over da. ‘The energy conservation equation now covers only the changes of the energy. During fracture over da the load may do some work dF, the strain energy may ‘change somewhat. dU. and some energy, dW, will be required for fracture. How the energy conservation criterion reads: de-v-m no 01% i ww deem ox ‘The equality must hold when fracture occurs. Conversely when the equality cannot hold, fracture does not yet take place and Equation (3.32) remains valid Apparently then, Equation (3.37) is a fracture criterion, Fracture will occur when enough energy can be delivered to provide forthe fracture energy dW/da. Energy delivers sonst come from a surplus of 4(F — Uda. If this surplus is sulfcient to cover the energy required for fracture, dW/da, then fracture will if not Equation (3.32) continues to gover. ‘The fracture criterion of Equation (3.37) can be used if the energy delivery UF ~ Uyida can be quantified. This will be done in the next section, 3.10. The energy release rate Consider again the situation of Figure 3.15d with a load, P,, resulting in a displacement 5,. First examine the situation in which the displacement does not change when the fracture occurs over da, In that case the load decreases to Ps, Lvevause it is easier to maintain the displacement 8, with a longer crack. However. the load does not move and therefore it does no work: dF = 0. The strain energy does change from | P, 8, (area OAD) to 45, (aren OBD) after the6 Oo) ° " 6 o Fire 3.15. nergy rkaedcing constant dapacement and constant fd (a) Pate with cack ‘indr lod P deplacemet (0) Load displacement recon () Ps ecord for wall and Ig ‘sack (4) Fracturing at constant digplacement)Fractring at constant oad process. The strain energy decreases, so the change is negative. Hence, it turns rable energy is —dUda, and Equation (3.37) becomes: aw au 1 oH where 2 <0. 38) [Next examine the case in which the load remains constant during the fracture cover da. In that case the displacement increases from 8, to 8; as shown in Figure n 3.15e. Now the load does work, namely an amount P(6, — 6); note that the load remains constant during the process. The strain energy also changes, ramely from 4,4, to | P,d,. This time the strain energy increases, so its change is positive. Equation (3.37) becomes: Pads — hd — YPUds ~ Ay = SE ” 69) aw 1Pi5; ~ 3) = So In one case the deliverable energy is ~dU/da, in the other case} P(By ~ 8; Note that the latter amount is exactly equal to dU/da. It appears then that the deliverable energy is always the same, namely dU/da, regardkss of whether fracture oecurs under constant load or constant displacement. In either case the cliverable energy is equal to the change in strain energy dU/de, For constant displacement the deliverable energy is coming directly from a release of strain ‘energy. In the other case the strain energy increases, but the load does twice th ‘much in work. so that the surplus is still equal to the change in strain energy, be it with opposite sign ‘Clearly then, the deliverable energy is always equal to the change in strain energy regardless of its sign. Instead of d(F ~ U)/da one may then use the absolute change of the strain energy dU/da in Equation 3.37. Then the fracture criterion is du | aw da * “da The left hand side is called the strain energy release rate, the right hand side the fracture energy or fracture resistance. Fracture will occur if, due to the extended slit by da, sulficient strain energy is released to cover the energy (o create a fracture over da, The fracture criterion is somewhat simpler now than before, but it remains necessary to find an expression for dU/da. As the strain energy is affected by the ‘crack one can write: U = Uretyaen mens + Usa Gan Consider a very larg plate of length L, width Wand thickness 8 with a smal center crack of 2a, Ifthe loading is uniform tension then the sean energy of| cach element of the uncracked body is the same and equal to Jo/E, so that the slrain energy of the uncracked body becomes: 40) Ustrsineeons = $e LBW. Gayn ‘The stress due to the crack will depend upon the crack tip stresses. These crack tip stresses are proportional to the applied stress 6, Thus the strain energy due to the crack must be proportional to ¢’/E. The term will also be proportional to the thickness B (twice as thick a plate has twice the energy). Hence, ae 34) Vsamona + FB. aay ‘The contributions by Equations (3.42) and (3.43) must have the same dimension. They differ by a length squared (L"} Since the contribution due 10 the crack must depend upon crack size, a it ber=sves inescapable tha it depends "upon a’, otherwise the dimensions would not be correct. Therefore: = cS Be. A Usama = CF Ba ow ‘There can be a dimensionless coefficient Cin the equation, the value of which ‘cannot be obtained from simple dimensional analysis, Formal calculation of the strain energy will show that C. ing Equations (3.42) and (3.44) gives the total strain energy U. As everything is evaluated for unit thickness, one must divide by B to obt e xot@ ee ad 3 Um Sus G45) Now dU/da follows from differen aU _ 2na’a wu. be (3.45) This is for a crack with two tps. Since all consideration are for one erack tip, fone must use av aE oan pet crack tip and per unit thickness “This permits writing the fracture eiterion of Equation (2.40) as ap Rea de Fracture 222 = oat ‘Usually, the fracture energy per unit crack extrusion ABV Ma idenoted by R (For fracture Resistance), while the energy release rate WUjda is denoted as G. Then Equation (3.48) in shorthand i given a9 o GeR (49) Equg G4 thy shorg tom ” a ‘The meaning of the energy criterion Equation (3.49) shows that fracture occurs when (xo?a) reaches a certain value, namely ER. Now notice that oa is exactly the square of the stress intensity, factor: K. Hence, the equation pronounces that fracture occurs when K? reaches, certain value, which is the same as K reaching a certain value VER. G50) |W follows then that YER must be the toughness K,, and appareatly the fracture resistance is R= KS/E. Wimay be concluded that, fortunately, the fracture criterion derived from the ‘nergy conservation law appears tobe identical tothe one derived before on the basis of crack tip stress. Had this not been the ease, one of the two would have to be declared wrong. Apply ths criterion to the same test and example asin Section 34, From the test one can obtain the material's fracture energy R. Assuming a steel for which E = 30000ksi, and knowing that f= 1, a = 2 inch, and the fracture stress 30 ksi, Equation (3.49) provides the fracture energy: Fracture if. K nx Wx 2 ksiin Re S900 = 9 Fi in kip ~ 019242 — o9kipin = 190Ibsfn Note that indeed R = K?/E, because with K = 75ksi vin (Section 3.4) one ‘obtains R = 75'/30000 = 0.19kips/in. The fracture energy is an energy (in/lbs ‘or Nm) per unit crack extension (in or m) and per unit thickness (in oF m), 30 that its unit is in Ibsjin or New!/m = N/m. ‘One can apply this result to calculate the residual strength for the same case as in Section 3.4, The fracture stress follows from Equation (3.49) as: R = re asp Substituting the value of R = 0.19kips/in obtained above, f = 1.12, and 4 = 3in, one obtains oy = 22ksi, which is the same resull'as obtained in Section 3.4. The two fracture criteria are indeed equivalent. 3.12. The rise in fracture resistance: redefinition of toughness 1 would seem reasonable to assume that the fracture resistance R is constant and independent of how far fracture has progressed. This is astusning that it takes equally much energy to fracture the first da as the adjacent, next da. It would cause R to be a horizontal line as @ function of Aa, as shown in Figure 36a, NS iverag rereyg 40). @ vdeo as GH salt” 1 FoFigure 116, R-Corve (a) Horiontal Recurve: (b) Rising Rewrve the case that f= 1, G would be G = no*a/E, which isa straight fine as a function of a Wi pon @. Assume a erack of size a present. im by taking a tothe left. One cam then Grav the dashed line for G, the actual value of G for erack sire a being given by G, om the ordinate ‘Clearly G; is less than R, and fracture will not take place at the stress 0, Raising the stres to wll increase the value of G to G. Since G, = R. fracture will indeed occur. The ‘rack’ extends during fracture. so that G increases from G, to Gy, while R remains the same. Hence, G remains larger than R and fracturing will continue uncontrollably. Tn reality Ris nota horizontal line. The fracture energy appears to incre somewhat during fracture {1} If the increase is small such as in Figure 3.16, the fracture will sil proceed as above. This is generally the case when the toughness is tow orn plane strain. When the toughnessishigh, the Recurv rises more steeply, as shown in Figure 317. TAU the sites othe energy release Gis equal to for the fst tne Fracturing will commence but not continue: iis stable. If the fracture proceeds cover Aa, the value of G rises from G, to G;. But R rises steeper, and although G, was equal to R, the new G; sess than the new increased) R. Apparently. the fracture cannot proceed (itis stable) Farther inereve of the tess fo wil aus fractre to continue, AG he eliverable energy is agtin equal to R. Finally. ofthe strexs op, the fracture wi be unable, beesse Gy increasing to Gy remawns larger than R. Fracture will then proceed uncontrollably (instability). Pare 117. Rising Curve. () Stabe facture from G0 Gy: () Stable cure at cack of iene length During the stable phase of the fracture the crack extends from a, to a. This should not be interpreted to mean that a crack of size a, has a strength ay, Although this seems trivial, serious errors in fracture calculations often occur ‘due to this misinterpretation as shown later in this section, Extension of ay (0 «4, occurs by fracture and not by the process by which a, was formed as e.g. by fatigue or stress corrosion, A fatigue crack of length a, cannot sustain the same high stress. Fracturing will commence at a lower stress than a, namely 2. (Figure 3.17). Fracture instability occurs at of, which is lower than ay, Note that the smaller slope of the G-lines indicates lower stress ‘The interpretation is that a fatigue crack of size a, causes the strength to be pr. That the fracture proceeds to a; before it becomes unsiable is rather umaterial for the end result (Iwo half structures). Similarly. a crack of size a, hhas a lower strength 0,2. Given the presence of a certain fatigue crack say a), in eg. an airplane wing, the strength of the wing is o,,. That the fracture proceeds first to a, before the wing comes off is rather immaterial for the passenger (and for the engineer) 1s the strength that counts. Naturally, ifthe siress would rise above o, (Figure 3.17) but not reach o, then the wing would stay together and the crack would be somewhat longer than a, due to some stable fracture. However, this is trivial ifthe Fracture stress is not reached the structure always stays together, whether there is stable Fractute or not. Clearly fracture instability occurs when the G line is tangent to the R-curve, when the two have equal slopes (Figure 3.17). As the slope of the tine is described by the first derivative, the condition for instability is82 Ga)= Ra) and also 0.32) (Gide), only the first equation is satisfied fracture will occur, but in order for fracture to become unstable (uncontrollable), the second equation must be satisfied as well, ‘This has important consequences for the definition of toughness tw mrevious sections toughness was defined as the stress int: at which fvacis- asses, 1 now appears that (unstable) fracture occurs at diflerent vali. iy for different crack sizes. Figure 3.17 shows that with crack of size a fracture occurs at G,, while with a crack size ay, fracture occurs at Gy. Since K? = VEG it follows that the values of K are different atthe time of fracture inthe (wo cases. {In previous sections it was assumed that fracture always takes place at the same value of K, namely the toughness K,. Apparently thisis not true if there i stable Sracture first. This might cause serious complications For the analysis of fracture. but it wil be shown below that an engineering approach can solve the problem. It should first be mentioned that Figures 3.16 and 3.17 are for the case that B= 1, namely the infinite panel. For real structural racks lis not equal to one, so thal the G:tines are curved (note thal G = 7 ‘naj (the curvature being larger for large cia: bs. ihe values of G, an. sce of K, at fracture tend {o be more or less equal to dillrent crack sizes as illustrated in Figure 3.1Ra. From this one might conclude that the assumption of constant (namely K,) at fracture was not so objectionable from an engineering point of view. [Next consider the residual strength diagram as shown in Figure 3.18b. In the presence of a fatigue crack of say one inch, fracture starts at a stress indicated by A. and fracture instability occurs ata stress indicated by B (compare previous figures). This means that if a crack of 1 inch is present, the strength of the structure is defined by B. It does not mean that the residual strength with a crack size of 3 inches is defined by B. The figure shows that such a crack would cause fracture instability at D, and its residual would therefore be defined by D. ‘Although the one inch crack shows stable fracture until 3 inches, a fatigue crack of three inches would cause fracture at D, and not at ‘The two curves in Figure 3.18b can be defined by a “critical stres it ‘The top curve is determined by the critical G (or K) for fracture ins bottom curve by the value of G (or K) for the onset of fracture: Onset of fracture: flo,/ma, = K, } (053) Instability Bo, Jaa, = K, From this the stresses at onset of fracture, o,. and the stresses at fracture stability, ¢,, would follow from: eo Psa sey 8. wn o no ‘irre 3.18. Approximate Consaney of Gand K; (4) Carved ms opposed to ssight Cine: () Residual strength curves for plane ste aeeex) a a (3.54) ” * ina,84 “The curves represented by Equations (3.54) are shown as the top and bottom ccurve in Figure 3.19a, Clearly, the value of K, in the above equation is not a ‘constant (see previous figures) but the value of K, depends upon crack size. On the other hand, as shown in Figure 3.18a, the value of X, will not vary a great dealin actual structures. ‘Neither the top curve, nor the bottom curve in Figure 3.19 provides the actual residual strength, On the bass of Figure 3.18, it was shown that the actual strength of a structure with a crack of say one inch is B. Hence, to obtain the factual residual strength of this crack one should plot the strength B at the original crack size. This has been done in Figure 3.19. Pala oe 4 7 Yt o Fue 319. Une of Ka in Feature Analyst (s) Residual strength and iia crack se (bY Predicted curve and tet date om bit of Ky 8s ‘The residual strength in the presence of eracks is determined neither by the top curve, nor the bottom curve in Figure 3.19, but by the middle curve. The fatter gives the actual residual strength for the case that a crack of size a is, present. One could assign a critical K, (effective toughness value) to this curve, id describe the eurve by the following equations: Fracture i fo, fra, = Ka (55) ova Ka igre, In these equations, indeed a, is the length of the crack as caused by fatigue Or stress corrosion, the fracture siress, 6, is indeed the strength of the structure with this crack present. However, the Equations (3.55) combine a crack siress not occurring simultaneously. As such, the equations are physically in error. However, from an engineering point of view, they are perfectly acceptable, as long as Kis more or les a constant (material property). As was shown in Figure 3.18a this is indeed the case. ‘The upper curve in Figure 3.19a could be defined by: Boo, na, 0.56) This would be a good definit roximately constant, which is the ‘case as shown in Figure 3.18a. However, i this were used in a fracture analysis, an erroneous result would be obtained: kK, ie @sn “The fracture stress calculated by Equation (3.57) would be for eg. @ crack sizeof three inches in Figure 3.186, and therefore would be B, while the REAL fracture stress is at D. Cleatly, the use of K, as defined would lead to wrong answers. The use of Equation (3.55) would lead to the correct answer if an engineering approximation is accepted. ‘It must be concluded that if the toughness is defined as Ky, the engineering answer willbe correct, while erroneous answers are obtained with K,..Neverthe- less, im many instances reviewed by the author, K, was used instead of Ky. AS its somewhat cumbersome to use Kee for toughness, the discusion of residual strength analysis in Chapter 10 will make use of the denotation K,, but itis strongly emphasized that Kis meant; i. the numbers used for the toughness are Keg. and NOT K,86 It seems almost superfluous to mention that the above discussion is not restricted to plane stress or transitional cases. All arguments hold for plane strain, although ths is often not realized. As the R-curve in plane strain usually rises only moderately, the problem is not acute; it nevertheless exists. Hence, all Of the above applies to plain strain as well as to plaie stress Clearly; the above engineering approrimations need not be made if fracture analysis is based on the Recurve. The procedure wuld then be to solve the problem of Figures 3.15 through 3.18 either graphicsity or mathematically. Thi ‘can be done rather easily (Chapter 10). The question is whether the result will bbe any more reliable. From an academic point of view, it will be, but from a Practical point of view it is not. Such a solution requires knowledge of the ‘Reeurve. Certainly, the R-curve can be measured, but it requires the measure- ‘ment of the progress of fracture. Such measurencats by their nature are $0 inaccurate that the R-curve can be determined only in an approximate way. Its subsequent use in analysis does not provide more accurate answers than the ‘engineering solution discussed above. Academic rigor dots not necessarily lead to belter answer; it only complicates problems. This will become abundantly clear in the next chapter. 4.13. Bxeretses 1. Calculate the fracture toughness of a material for which a plate test ‘central crack gives the following information: W = 20%n, B = 0.75 inch, 2a = 2inch, failure load P = 300kip. The yield strength is F, = TOksi Is this plane strain? Check for collapse. How large isthe plastic zone at the time of fracture? 2. Using the result of problem 1, calculate the residual strength of a plate with an edge crick W = $ inch, a = 2 inch. Do the same for W = 6 inch and @ = 05 inch (See Figute 3.3 for f). Check for collapse, ven is a toughness of K = 70 ksi in and a collapse strength equal to the yield strength with F, = 75ksi, Determine the residual strength of «center cracked plate of 20 inch width with a crack of 2a = 3 inch, and ofa center cracked plate of two inch width with a crack of 2a = I inch; cheek for collapse. 4. Im a toughness test on center cracked plate one obtains the following results: Wi 6 inch, B= 02 Inch, 2a=2 inch, Poy = Alki F, = S0ksi, Cileulate the toughness. How large is the plastic 2on fracture? Can thid problem bé solved with Equation (3.25), and why not? the calculated toughness indeed the true toughness; why not? What caused the fracture? _ 0 5. Ima plane strain fracture toughness tet bd a compact tension specnien the ftilure load is Skip. W = 2in,a = lin, B = 1 inch, F, = 80k. Calculté the plane strain fracture toughness, using both Equations (3.28) and (3.29). 1s this @ valid number? What is the sizeof the plastic zoid al fracture? 6. Ina plane strain toughness test one obtainé a value of Ky = S0ksi Jin: Is this a valid number ifthe material's yield strength is 100kst and the spocimen thickness is 0.5 inch? What is the maximum toughisss this specimen can ‘measure? Ifthe toughness is not valid, thet! éélienate the bland StHaIA ftdcture toughness with an accuracy of 5%. How thick stiould Gat like the speinien {in order to measure the plane stra toughilelé Unde AMY ebfditlohet 7. Cut a strip of paper four inches wide. With a knife ot scissors cut a center ‘crack (slit) of length 2a = 1.5 inches in the top half, In the bottom half cut two edge cracks with a = 0.60 inch. Predict where the strip will fail in tension, Confirm your answer by rollig thé ends afdlind pehcllé and by poli ven 4 (a) Cateuite the complete residual strength diagram for & ctnter cracked {late of a material with F, = 7Oks), K, = 90ksi Jin thd W = 24 inch {() Do the same for a pancl of W = Ginch. (©) Do the same for a panel with a single edge crack assuming f of Figure 8.3 is applicable; W = 10 inch. Note: Think of tangent and collapse. For each case determine the permis ‘S0ksi, and a stress of 10ksi, by reading from your diagrams. ‘ References ort UD. Broek, Elementary engineering fracture methanct, 4th Editioa, Rtijhol (1585). th Rein pi esc ney Pen 4 (BIAF iet Paani 9 recone toes OO) I i iesy Pc PateeosG ei Tc seen chan or tobeo Ksach 9 Hee ‘Sih, Handbook of stress intensity factors, Lehigh Univertity Hrést (1943)."‘* IS1b6 neste DA Carwig, Gonen sey Her May 3 in IT} WEF. Brown and J.E. Srawley, Plane strain crack toughness testing of high sirengih metallic A mo AM SP 08 IB Anon., The standard K, test, ASTM Standard £-399.CHAPTER 4 Elastic-plastic fracture mechanics A. Scope If fracture is accompanied by considerable plastic deformation. a concept known as Flastic Plastic Fracture Mechanics (FPFM) is use. The literature on this subject is very confusing, The fracture parameter used is often referred to as the “integral. However, J is simply the strain energy release rate, and on this basis the equation for Jcan be readily obtained, without the lengthy derivations, from a generalization of the energy criterion discussed in the previous chapter. 8 the reader is assumed to he familia with the strain energy release criterion for LEFM. Instead of following the historical development of EPFM, it will be shown that the strain energy release criterion can be used directly for elastie plastic fracture by the simple expedient of an equation For the non-linear stress-strain curve. This leads directly to an expression for J. tts application and use will be discussed, In order to avoid confusion, the argu ‘mentation will bein small steps at a time. This leads to some repetition but will facilitate understanding. It also leads to a few more equations, but these can be followed easily because no steps are omitted. Ina ater section the integral form ‘of J will be considered. The use of the crack opening displacement (COD) criterion will be discussed as wel 4.2. The energy criterion for plastic fracture Whether there is pasiciy or not, the enerey conservation criterion must hol It was deponsrned in Chapter 3 that this lads to the following fracture criterion: www aa” a For elastic behavior it could be demonsrated that G=R a 9 (42) E), the energy release rate, dU/da, the above equation can be written in terms of stress and strain: Poa = Root G= R 43 ‘The strain energy release is usually denoted in shorthand by G, and the fracture energy by R. Although the expression for G was derived formally in the previous. chapter, itis important to notice that Equation (4.3) must be twe om the basis. of dimensional arguments. The factor f is dimensionless. Since strain energy is Jor de = Car (Chapter 3), the equation for G, the strain energy release rate, ‘must contain the factor at. Obviously, G depends also upon the crack si [ASG is an energy per unit thickness and per unit extension, its dimen: in-hsjin? = Ibsjin, The dimension of o i tbsfin® (cis dimensionless). Therefore, the erack size must appear in Hinear form as @ (nota? ofa” or otherwise). In that case the dimension of G becomes indeed Ibs/in? > in = Ibsfin. Thus is can be understood without analysis that Equation (4.3) must be of the format shown. Equation (4.3) can be changed into (4.2) because the stressstrain equation ‘a/E)is known, Expressing G in terms of stress only, permits the use of the criterion in engineering analysis where stresses are of intrest: miata E the fracture stress can be calculated from: o> fee & na) a Provided fis known for the geometry at hand, Equation (4,5) can be used for engineering fracture analysis to calculate the fracture stress of @ cracked structural component. In the case of plastic deformation the geometry factor may change, but there will still be a less geometry factor; let this factor be denoted as H. ‘Whether the stress-strain curve of the material is linear or not, te strain energy release rate is sil given by Equation (4.3) (same dimensional arguments) R 4.6) [Note again that the equation must contain a factor at, the dimension of which is bein, to that the crack size must appear in linear form (a) forthe dimension to become bsjin (strain energy per unit thickness and per unit crack extension). Indeed, Equations (4.3) and (4.6) are identical regardless of the form of the =k aay Hea90 strain-stress curve, be it that another dimensionless geometry factor may linear-lastic material dU/da is generally called G, but for a non-linear the shorthand denotation J is uses. is. litle confusing and for ‘which there is no good reason. Since the densi J is used throughout the literature it will be used here as well. Similarly, while the symbol R is used for the fracture energy of a linear-elastic material, the fracture energy for a non- linear material is denoted in shorthand as Jy. The fracture criterion is sill dUjda = dWjda, which is written in shorthand as G = R for elastic materials, but in accordance with the above new symbols, it becomes for a non-linear materia: Jah an or alternatively: Hora = Jp 4a) ‘The previous Equations (4.1) through (4.6) have not changed: only the symbols have. 1 was possible 1o use Equation (4.3) for fracto -sIysis hecause « = a, $0 that Equation (4.3) becomes (4.5), thus permitting calculation of the fr slress. By the same token, Equation (4.8) can be used for fracture analysi can be expressed in o, ie. ifan equation is available for the non-linear stress- stain curve, ‘Any form of stress-strain equation will be useful. provided it properly describes the material's stress-strain curve It will have to be a curve-fting ‘equation (empirical). There is no objection against this; also Hooke's taw is ‘empirical as itis no more than a linear curve fit, where E follows from the slope of the line. ‘The most convenient curve fit for a non-linear stress-st is a power function. The resulting stress-strain equation bberg-Osgood equation: $4 oe = tty For most materials, but not all, Equation (4.9) provides good fit of the stress-strain curve. The equation can be used in various forms, but Equation (4.9) the most convenient. Other forms of the equation will be discussed in Section 4.7 ‘curve (Figure 4.1) known as the Ram- « 49) 4.3. The fracture criterion ‘The Ramberg-Osgood stress-strain equation can be used to evaluate the fracture criterion of Equation (4.8). First note that the first part of Equation Fieme 41. Stes-Siia diagim. (4.9) represents the linear (Hooke’s) part of the stress-strain curve. For this part its already known that dU/da = xf%o'a/E. Therefore, only the effect of the plastic part needs consideration: wef (4.10) Note that this equation simplifies to Hooke’s law for the case that = = B) Substitution of Equation (4.10) in Equation (4.8) provides: Hata Je ain F 1 (F = £) the Equation (4.11) indeed reduces to Equation (4.4), as it id this implies that for m = 1:1 = xp Now dUida for the elastic part, G, and dU/da for the non-linear part, J are both known. The fracture condition is (combining Eqs. 4.11 and4.4): Pola Hone aw E Wa (412) Gtd=In has become generat practice to denote the Fracture energy dW/da a8 Jy. This is again confusing because actually it is R + J,, which is now shorthanded as In Because G is equivalent to J for n = 1, as pointed out on the basis of2 Equation (4.11), iis more consistent to cal it J,, and to deop the shorth denotation G altogether: Fracture ifs Jy + Jy = Je aay nota | Hoa ri rfrora E F ‘This equation indeed provides a useful fracture criterion for plications, because itis expressed in stress. Fracture will occur when the stress is sufficiently high to satisfy Equation (4.14). The fracture energy J, in the jon represents the materials fracture resistance; it is therefore essentially the material's toughness expressed in a somewhat different form. It can be ‘mengured in a test, ifthe geometry factors land H are both known for a certain specimen configuration, The modulus F, as well asm and Fean be derived from A stress-strain curve measured in a regular tensile test, By performin test om a cracked specimen, the fracture stress can he measured. Values forall parameters in the left hand side of Equation (4.14) are then available, so that the ‘toughness’ J, can be calculated. For example, assume that for a certain material = 20000ksi, n = 7 and F = 2 x 10°, Section 4.7 explains how m and F are obtained from a tensile {est), Also assume that for a certain specimen with a crack of a = 2 inch and the values of and H are 1.1 and 6 respectively. The specimen is tested and it is observed that fracture starts at a stress ¢ = 30ksi. Substitution ofall data in Equation (4.14) provides: LP x We 2 Gx Wl x? 3000 Te TOO Oe so that Jn = 023-4394 = 4.1Tin-ksi = 4170Kbs/in sclure energy now being known for the material at hand, Fquation (4.14) bbe used to calculate the fracture stress ofa cracked structure by solving for 2, Naturally, both geometry functions f and 7 must be known forthe structure al hand, Unfortunately, Equation (4.{4) cannot be solved directly for 0: am iteration procedure must be employed, but this is merely a tedium. As the solution usually is obtained by a computer, there is no serious difficulty. In the event that J, is small compared to Jy, the ist term in Equation (4.14) can be neglected: it reduces to: Horta F he (as) 8 Which can be solved directly for the fracture sires: (4.16) Once more, 1 (F = B) this equation reduces to Eq (4.9), with apparently 1 ‘Assume that for a certain structure with a three inch erack the geometry tors ate fl = I.2and H = 7. The structure is made ofthe material on which hove hypothetical test was done, so that E. Fand mare as given above. The material’ facture resistance was found to be 4.17 kips/in = 4170 than. Indeed the elastic part of J was small, so that Equation (4.16) can be used to calculate the fracture stress of the structure as: 2x 10" x ai7ye” wae) Clearly then EPFM is no more difficult to use than LEFM; the only compli tion isthe necessary iterative solution when Equation (4.14) must be used. (One other complication arises as discussed in Section 4.4.) Of course the ‘toughness Iq to be measured in test, but also the LEFM toughness must be measured. ‘The geometry factors f and H must be obtained for the structureat hand. The clastic geometry factor, f, has been calculated for many geometries and compiled in handbooks [1. 2, 3}, or can be derived using one of the simple procedures discussed in Chapter 8. Similarly, the plastic geometry factors, 1, have been calculated for a number of geometries. These are also available in handbooks [4,5]. One complication arises here because H depends upon n also: H = Holl, m). Obiaining H is more cumbersome than obtaining f, but once calculated H can be presented in handbooks for later use. 4A, The rising fracture energy EPFM deals with high toughness materials. For such materials the fracture energy. Je. tends to increase during the fracture process. This was discussed already on the basis of R in Chapter 3; its consequence is that fracture may be slow and stable initially, until at some point an instability occurs causing Fracture to become fast’ and uncontrollable. The procedure to distinguish between stable and unstable fracture in EPFM isthe same as in LEFM, and also in this respect EPFM presents no new problems. The LEFM procedure for Recurves as reviewed in Chapter 3 will be assumed known for the following discussion Figure 4.2 shows the rising Jy-curve. Fracture commences whenJ = J. For a certain existing crack, the curve for Jean be drawn for a number of values of the stress, on the basis of the left part of Equation (4.14). Several such curves4 Figure 42 ewe (or Siren ses a Jy are shown in Figure 4.2. At the stress o, the value of J(a) is given by point 4, which is lower than B. Hence J < J,, and fracture is .~ paesible. Increase of the stress to 0 raises J(a) to point B. Fracture now takes place because J = Jy But this fracture is stable; it cannot proceed if the stress remains equal to ¢,, because then J would increase to point C. while J, would increase to point D. so that J would be less than Jy. For fracture ( continue the stress must be increased to 0). which brings J (0 point D) During the increase of the stress from 29,10 0, the fracture proceeds in a stable manner (vom a toa + Ady. Further crease of the sltess causes Fracture to proceed . The fracture is controllable (stable); it can be stopped at any time simply by keeping the stress constant Finally, when the stress reaches 0. the fracture become unstable (point E), because rom there on J will emain larger thin J,. Fast. complete fete will, cause the structure to break in two. [fat any tine between oéand F the stevciare ‘would be unloaded, it would stay intact, although the damage would be larger by Aa. ‘The stable fracture is often referred to as stable crack growth, Ti misnomer, because the crack extends by fracture and not by one of the erack Chapter 1. Stable fracture is fracture in nugh itis slow and can be stony! if the stress is not further increased. After instability fracture is unconteol e, Also the tert ‘sub-critical crack growth’ is sometimes used forthe stable fi:scture, The description may be clear in the context, but in its general use it refers to cracking by fatigue and slress corrosion, The crack hecomee critical when fracture hegins, whether stable or not. The use of more consistent hingunge in fracture mechanics would Certainly avoid contusion. One may argue he preciseness of the terms 95 tused here, but at least different phenomena are identified consistently by different names. 1s apparent from Figure 4.2 that the condition for instability is: Ha,)=Igla,) (ai), (a), ; which signifies tangency between the J(a) curve and the Jy-curve. The instability, Point can be found graphically by plotting the Jy-curve and the J(a) curves for a number of stress values as in Figure 4.2, Algebraic solutions are discussed in Chapter 10. (See also Eqs. 3.52) ‘The procedure is valid when there is a condition of load control; i. if the stress does not drop once the fracture becomes unstable, If thee is displacement ‘control, the stress may drop. For example, when the loading is due to thermal stress, the end displacement is fixed. During fracturing the sifness decreases which may cause avery rapid decrease ofthe stress, because less sress is needed to maintain the fixed displacement. Since J depends upon 6, it may not be Possible o maintain J > J,. so that the instability is postponed. In that case the ‘drop in stress must be calculated from the decreased structural stiffness, and the calculation be continued until Equation 4,17) is satisfied (Chapter 10). Dott in load control as well asin displacement control, the sires will increase wo Instability in load control coincides with the maximum stress. In displacement control the fracture may remain stable after the ‘maximums is reached (Figure 4.3); instability may or may not ensue later, Up till the maximum stress there is no difference between load control and displace- tment control. Displacement control is governed by the stiffness of the system (loading + structure): instability is therefore system dependent and not a ‘material property’ (Figure 4.3). Most structures are under load control. For example, the wave load on a ship, the gust load on an airplane or the toad on a erane does not abate when the structure is fracturing. Hence, a load conteol analysis is usually applicable. In the event of displacement control, the only thing of interest often still isthe ‘maximum stress that can be sustained. In that ease no account for displacement contro! is needed either, because up til the maximum stress there is no distin a7 ction anyway. In * ages the difference between a, and the stress at instability is very small, «nt solving Equation (4.14) once for @, (point B in Figure 4.2) may be adequate (conservative). Examples of analysis are provided in Chapter 10, Finally it should be mentioned that Equation (4.17) has been modified by96 Figs 2, Load conto Uf) sermo dptcement contol (ight ‘multiplying both sides by £/F leading to: Ew, EW, de or in shorthand T= Tua (4.18) On ra Pare 44 Resa strength dingran in FPEML 7 ‘The quantities at both sides are then renamed. The left side in shorthand is called the tearing modulus, 7; the right side Tyy for lack of a descriptive word, Hence, the instability equation is written as T = Ty, but nothing has changed. To the uninitiated only the fog intensified 45. The residual strength diagram in EPFM; collapse Ry solving Equation (4.14) for a number of crack sizes one obtains the residual strength diagram, as shown in Figure 4.5. Solution of Equation (4.17), ether eraphically or otherwise (Chapter 10), provides the stresses for instability as well. Hence, there will bea line For the onset of fracture, and one for instability (compare Figures 3.19 and 3.20), For convenience only the lower of these (Figure 4.5) is taken into account in the following discussion, because in general there is litle difference between ¢, and ay Now consider again Equation (4.14). For the case of a -» 0, the calculated ‘residual strength will he infinite. This same problem was encountered in LEFM; it as not een solved by EPFM. An approximation is still necessary as in LEFM. For small crack sizes the calculated fracture streses will be very high; for very small cracks it will be infinite, ie. higher than F,, which is clearly wrong. Apparenily, as in LEFM, collapse as a competing Fracture condition ‘cannot be ignored. Collapse determines the highest load carrying capability of {Figne 15. FEM anal orcenercricked pane: W = I6inch,E = 28000Ks F MF, ost Colpee prevail (lower curve)98 the structure, regardless of any fracture criterion, whether LEFM or EPEM. ‘Thus collapse must be analysed as a separate condition, ns inthe eaee of FEM Chapters 2 and 31h cx mun satisfied first (at the lowest stress) determines the strength of the structure, For an ideally plastic material (horizontal stress-strain curve beyond F,). the -Osgood equation (Equation 49) is n = ac. Thus the left hand side of Equation (4.14) becomes indeterminate, J i undefined and attains any value necessary (collapse) to satisfy the rwation. This always will ‘occur at plastic instability. When plastic defirmati heceunes uncontrollable J altains the necessary value for collapse. Attempts have teen made to incor- porate the collapse condition in the EPFM criterion, but this cannot be done by jst considering J. It would require a separate evaluation of plastic i criteria. Although this might be possible in principle, engineering analysis, Concocted modifications of plastic zone corrections o K [6]. The practical solution isto treat collapse in the manner discussed in Chapters 2 and 3. and to assess both the collapse condition and Equation (4.14), The one leading to the lowest failure stress determines the tesidual strength. Then the problem is the same as in LEM, discussed exten: sively in the previous chapter. 46. The measurement of the toughness in EPEM ‘Any kind of specimen can be used to measure the K, or Ky: I this were not the case, applications to structures would not be possible either, There is no ‘objection against a standard specimen for which fis rnown very accurately, but any other specimen with known fis appropriate. In the same vein, Jy can be measured with any specimen for which fi and H are known. An example was given already in Section 4.3. In order to obtain the Jy «urve the stress should be measured during the stable fracture as well as the progress of fracture a= a + Aa. Equation (4.14) then provides J. There is definitely a specimen size requirement, because J becomes indeterminate when collape occurs, a8 discussed. {is unfortunate that a test procedure to measure the Jy-curve was standar ized {7) prematurely, ata time Equation (4.14) had nol evolved. Thus the test seems (0 bear no relation to what was discussed in the previous sections. Recall from Figure 3.12 that the energy relewe rote i equal (0 dU/da, the small area between two load-displacement curves for «and a + da, regardless ‘of whether the load or the displacement is constant over da. If the displacement is constant there is x decrease of strain energy (negative), so that the available energy is — dUjda (see Chapter 3). Inthe elsstic case the area between the curves das G = dU fda = ~ 425. Thisis sil the cave when the is easily deter ” ve is non-linear as in Figure 46; the area between the ive), s0 that J = —dUjda is: Joad-displacement curves is dU /da (1 (a Je - [2 19) {an equation forthe load displacement curve is avalable J can be obtained from Fquation (4.19). The standard for J measuremen's is based upon an ‘estimated equation for the load-displacement curve. ‘Consider a bend specimen in which the remaining ligament is Fully plastic as in Figure 47, The elastic deformations can then be neglected. The specimen will deform as two rigid bars connected by a plastic hinge. The bend angle y depends ipon the bending moment, which is equal to PL/2 (Figure 4.6). Ifthe steess- strain curve is assumed horizontal beyond F, the resistance of the plastic hinge is equal to F,,. The equation for y then must be as shown in Figure 4.7 Obviously. "is proportional to P and inversely proportional to F,. the resistance of the hinge. If the specimen is twice as thick, willbe half as large for the same load, so that must be inversely proportional to B. Hence, forthe {equation to be dimensionless the ligament must appear as in the denor ‘A dimensionless constant (will complete the equation. By noting that y he equation for the dependence of on P is obtained as in Figure 4.7 Apparently, this isthe sought equation for the load-displacement curve. Differentiation of the load displacement equation with respec! to b provides P06, and it follows from substitution of P that AP/Ab = 2P/b (Figure 4.7), ‘When the crack extends over da the ligament will decrease by db, which means that da = —dh, and therefore 9P/2a = —2P/b. Substitution of this in Equation (4.19), and noting that J must be determined for a unit thickness (hich requires division by B), the equation for J becomes: 24 2 Jom pgfP ds or sm oF (4.20) Fine 48, Stein cvcigy these 1 epacement GnGiitesina da. Ki: ret: rah ‘notinear Compare wih Figure 315(0) Band specimen with plastic hinge a + OR ae 1 Fat = ee Fab ar © Jor bend specimen Figure 47. Standard J, and Jy est (9) Specimen: (b) Anat: () Rest “The integral part ofthe equation is simply the area, 4. under the load displace iment diagram (Figore 4.7c). which can be measured - by using a planimeter or ‘otherwise - from the test record of P — 3. Surprisingly, the small area between the curves in Figure 4.6b is found by taking twice the (otal area, 4, under the curve and by dividing by b and B. [Naturally this is true only under the assumption made, namely that the entire ligament ig yielding and that the stressstrain curve is horizontal beyond Fi, (note that 1 does not appear in the equation), These are in fact the collapse conditions under which J becomes indeterminate as argued above. The latter means that Equation (4.20) is unsuitable for the imepensdent calculation of J. since J = J» during fracture, the equation can be ured to determine Jj. Figure 4.7 for ligament in bending: it applies toa compact tension specimen tol as well, be it that a small correction must be made (7 as there is some tension superposed on the bending, ‘AL the time of fracture J = Jq. Hence the fracture resistance Jy is derived From the load-displacement diagram measured in a test on a compact specimen hy applying Equation (4.20). The fracture resistance increases (Section 4.4). Iva test fracture is first stable. AC various stages during the test the amount of stable Fracture, Aa. is measured and marked on the load-displacement diagram (Figure 4.8). By the use of Equation (4.20), J can be obtained as a function of ‘Aa, yielding the sought J_-curve as in Figure 4.8b (compare with Figures 3.18, 3.19 and 46. The standard recommends use of a so-called ‘blunting’ line, which forces the Jy-curve ta go 7er0 for a = 0, Obviously, with a = 0 in Equation (4.14), the ‘ress for fracture would become indeterminate. Thus the blunting line cannot be used in a fracture analysis. I i¢ an artificiality, blunting is duc to plastic deformation (Figure 1.4), not fracture (Figure 1.7), It has no place in the “Tycurve, because it ignores the physics of the problem. ‘Constraint is essentially ignored in the standard. In LEFM the toughness K, ‘oF R depends upon constraint (thickness), and the standard for K, (ests puts {real emphasis om the constraint condition through the thickness requirement crs Faure 44 Movcuement of Jy (2) Load delacement diagram with indications of facture extension: (B) J curve102 Clearly, this problem exists equally in FPFM. The fracture resistance, Jp, ‘depends upon whether contraction in thickness direction can take place freely ‘oF whether itis constrained by the surrounding material (Chapters 2 and 3). “The standard for Ky tests requires the thicknes co be larger than B> 25 (K./E,,7. Since in the elastic case J = K?/s:, this tenaslates into: > 25 ERIFL aa ‘There is no reason why this requirement should be less severe in elastic plastic fracture, Instead, the following condition tas been proposed (8) B> 251 Wooo this | spplied in LEFM, 1, SOKSi aud Equation (4,21) (4.22) yields with F = 30000ksi, and eg. = 7 a 2b) > o0¢() (eam we 38) The condition of Obviously, this is an inadequate condition (F Equation (4.22a) puts C = 1500 rather than C= 25, in order for the requite- ment to be as stringent as the one for LEFM. These i no reason why it should bee any less for Jy tests. Naturally, regardless of constraint, the test will provide J for the thickness used, like an LEFM test would provide K, (Chapter 3), I seems strange that ASTM embarked! on standarstiring a tc: hile the subject was sil inthe research phase (a :sndard aay be useful for engineering applications; it has no place in research), and ata time Equation (4.14) was not available. Knowledge of J, is of no use but for calculation of the fracture stress ‘ofa structure. The latter became possible only sive 1 Fquation (4.14). Last but not least, Equation (4.14) has made the equation m the standard obsolete, A thorough revision of the standard for engineering rather than research seems ‘opportune. "4.7. The parameters of the stress-strain curve This section could be very short ifthe stress-strain equation used in EPFM had been given the simple form of Equation (49). Unfortunately, new confusion has been introduced as will appear in Section 4.8. Its for this reason that a longer discussion of the stress-strain equation is necessity here. Hooke’s law is a simple mathematical descripsion of tie experimental fact that elastic stress and strain are proportional: itis an empirical law. Inthe same vein one can use mathematical equations that fit the remainiler of the strest- strain curve. The most useful equation is the Ramberg-Osgood equation, which 103 o wet (423) Ms exponent » is called the strain hardening exponent. Further, Fis @ propor- tionality constant like Eis a proportionality constant. For lack ofa better word. F might be called the ‘plastic modulus’. For n = | and F = E the equation covers elastic behavior as wel ontinaton of i cas and paste sin provide he as: ent S42 aay ‘The modulus E is measured as the slope of the linear part of the stress-strain ‘curve. Values for m and F can be obtained rather easly as well. A tensile test vides the stress-strain curve in terms of strain, At any stress he plastic strain {s oblained from the measured total strain 28 fy = @. — o/E (Figure 4.1); for an example see the solution to Exercise 1). ing the logarithms in Equation (4.23) yields log ty = log a — log F. (425) Hence. in plot of log(siress) versus log{plastc strain), the data should fall on a straight line as shown in Figure 4.9a. The slope of the line ism, the intercept with the abscissa is - log(F) as shown. Ifthe data donot fall ona straight line, the material does not obey a Ramberg-Osgood equation, which is sometimes the case, There is nothing that can be done about this. With n and F, the total stress-strain curve can be calculated with Equation (4.24), IF the data (Figure 4.9) are reasonably on a straight line Equation (4.24) will be a good representation of the measured stress-strain curve as is demon- stvated in Figure 4:9b: the curve through the data points was obtained using and F from Figure 49a, ‘The Ramberg-Osgood equation was intended to describe the true stress-true strain curve. However, up to the point of maximum load, the eyistcin can be used just as well for the engineering stress-strain curve (Figure 49), be it that n and F are different even though the material is the same. The latter is demon- strated in Figure 4.10 which is for the same data as Figure 49. The only restriction is that the equation cannot be used beyond maximum load inthe case of the engineering stress-strain curve. It can be readily demonstrated (9] that n is equal tothe inverse ofthe true plastic train at maximum load. Although this would be an easier way to find nits accuracy is poor. The value of m so obtained from Figure 4.10 would be n = 8.49 which is different from then derived from the slope ofthe line in Figure 4.9. As F still would have to he determined as in 4.10, the plot is necessary anyway. Resides, this simpler method would ot apply to the engineering stress-strain curve,Cy ‘cara rte : 0 ee Ee Piqee 49. Ramberg Osgood equation for engncting seese-strain cre. (1g asic sain) ‘ers oe (tess (b Pgincring sess curve equation “The ‘plastic modulus’ F, tends to have rather unvieldy values, bbe more convenient to replace F by (5; that it may se= (SJ winesar=r 429 0s oo . ‘acsirsicam Masti0 8 eS a eae Fire 4.10. Fit oe sess. curve. (a) Long (ee plat rain) versus log (eres) “Tro ates ~ tre srin equation! W appears from Equation (4.26) that for @ = S,, the strain reaches 1 (100% strain). Sometimes 5, is called the flow stress, a somewhat unfortunate name, because it suggests that 5; has physical significance as a stress. As a strain of 00% is hardly ever possible, the stress will never reach the valve 5; ‘This section could end here if the developers of FPFM geometry factors had used the Ramberg-Osgood equation in the form discussed, Instead they106 employed a more complex form, the implications of which must be understood by the users of the geometry factors 1W should be realized that the Ramberg-Osgood equation has only three parameters, namely E, Fand m o£, S,... No further parameters are needed, ‘oF even permitted. Nevertheless, a fourth parameter was introduced as Follows: oer am ‘There is no objection against replacing Fby (a4) as wae done before in Fquation (4.26). However. introduction of the fourth parazz-1 is permitted omly if and ‘eare dependent, because by definition: es = Av a matter of fact, one may now take any value For op fy, such that & = 4/E. Thi et apas alte eues rete) id The shove equation is perme at Tong as one sity adheres 10 the dependence of parameter: namely Fquation (4.27) was written as: (429) ‘The literature on EPFM often refers to 0, as the Mlow stress. which is a very disturbing and confusing misnomer, because can be given any arbitrary value as long as Equations (4.29) are adhered to. Its generally aken equal tothe yield Hength, and the suggestion is raised that the laiter is significant for the equation. Nothing could be further from the truth when any arbitrary value is appropriate This can be readily demonstrated by the example in Table 4.1. From Figure 49 it appears that E= 30000 ksi, =6.27 and f= 2.1 E13 for the given ‘material. Taking arbitrary values of 50, and 100 ksi for op, and using the mandatory Equations (4.29) the stress-strain equations become as shown in Table 4.1. Any of these equations leads to the same strain for the same sess as shown. Indeed, ap can be chosen at will, I may be taken equal to the yield strength if so desired, as long a ti realized that this i an ahitrary choice Wan be argued that the yield strength depends spon tne stipe of the (E, Fn), and vice versa, the yield strength depends upon F and n, This is true and as such taking , equal to the yield strength is certainly defendable The Auctions ‘The geometry factors H(a/L, m) were developed [4 5 for a mumber of structural geometries and n-vatues, using the pla 107 Table 41. et of dierent deiitions of Rambers-Orgood equations, WOKS, F= ITE KO”, 9 = 627 © = ptm tee Ms Favation (49) “ With $1" = Fone obtains 5, = 133 ksi p °= 500+ (sy ‘As Equation (4.26), (B) With arbitrary ae = SO ksi: ry = $0/30000 = 0.00167 = Oh gk = 500 }0.00167 x 2.1 E13 = 1.28, ami \s Equation © Tonia ‘As Equation (4.28) © With arbitrary a) = 100ksi; r = 100/30000 = 0.0033 Wr MONT LEE = 50 aan Gis) ‘As Equation (4.28). © Results ti Fgation (4) Equation (8) Equation (C)___ Equation (D) @ ‘oontr ooniey oni) O00" ™ dorm ‘002i dois 01554 » oon 0499s ous 204336 (4.28). This requires adherence to Equations (4.29) as discussed in the previous The equation for J, was written as: Je = ansen (2). my equation P is the load, P, is the load at collapse supposing a, were the apse strength. Instead of the crack size, a, the unbroken ligamert, cis used. yh thegeomery iors ge108 ‘Clearly, the load P is related to the stiess, Py t0 ay and the igament 40 the crack size a Page ee aay Feet (iy" (2 which immediately reduces to: Hora _ Hot FH with A= Uh(g/ky"', Equation (4.33) is the basic form of the equation already known as Equation (48) from simple arguments. Equation (4.30) is just a complicated version of the same, Obviously, 2, 45, and ry can be divided out, they are superfluous. Indeed, the solution to Exercise 6 shows that the same results are oblained with Equation (4.32) regardless af the choice of (arbitrary). Equation (430) suggests that J depends upon a collapse load P, but ‘obviously a9 is divided out as well. This should he expected. because a collapse Toad cannot enter into J since the stress-strain equation used has no limit. The clastic energy release, G, could be expressed in the same manner by using P= ga, Pa = hogan oy = GE o = sPawton(2) 9 7 a Bringing in the collapse load or the collapse strength does not make G dependent upon same; it merely amounts to multiplying numerator and denomi- nator by the same number, which does not change the basic equation, Collapse ddoes not enter LEFM equations, nor does it enter EPFM equations in their present form, An artificial introduction does not change this fact. Collapse i ‘competing condition which must be assessed separately, at lest for the time being. ‘Two other objections can be raised against Fquation (4.30), Instead of just ‘one geometry parameter H. it must use four geometry parameters hg. and 1 Every time a calculation is performed double work is necessary: parameters must be derived and are subsequently divided out. Naturally, one could, once and for al, calculate H from H = Uh(g/ky""? and from then on use Equation 109 (4.10), but» computer does not object to unnecessary work and for it the form of Equation (4.20) need not be changed. The second objection is that J is expressed in the load P, while in engineering ‘one works with stress. Therefore Equation (4.33) is more useful; all other Fracture mechanics equations are expressed in siress for this very reason. For ‘complicated structures the conversion from load to stress is dore in the design stage not at the time of fracture analysis, 49. Accuracy wchers have expressed great concern about the large variability of J and nily) J. The reason forthe large variability is obvious. As J depends spon sires to the -th power, xecording to Equation (4.11), a slight dlference of 5% in stress with eg. m = 9, leads to a difference of (1L05)” = 1.63 or a difference of 63% in Jy. (Note that this occurs also in a standard tests: the load-tisplacement diagram becoming almost horizontal, the area under the curve. which determines Jq. changes dramatically with a slight change in 3. This may seem bothersome but itis of little practical importance. The value of Ja is of no interest as long as the predicted fracture stress is reasonably accurate. This the case, because ina fracture analysis the situation is reversed a dilference of 63% in J with n = 9 will lead to only a difference of 5% in the predicted fracture stress: (1.63)! = 1.05. (See Eq. (4.16)) For a difference in J by a factor of 2, and for n = 7, the predicted fracture stresses would be different by 2" = 1.09; hence the error (difference) would be 9% only, Thisis clearly demonstrated in Figure 4.11, showing the results of two calculations with exactly the same input and n = 7. Two Jy-curves were used differing by a factor of approximately two throughout. The predicted fracture stresses differ only by a small amount (9%). In general the stresses in a structure will not be known with better accuracy, so that any of the predictions in Figure 4.11 would be satisfactory from an engineering point of view. Even the predicted amounts of stable crack growth at maximum load do not differ appreciably. as shown in Figure 4.12. Most alloys satisfy the Ramberg-Osgood equation fairly well. However material of great interest to some industries, namely annealed 104-SS, exhibits a slress-strain curve that cannot be fitted with the equation, Yet itmust be fited to such an equation, otherwise an EPFM fracture analysis cannot be performed, Im that case there is « choice as to whether the equation should fi the lower or the upper part of the stress-strain curve. ‘There is no categoric answer to this problem, Any conclusions reached are watcrial specific. Although it miey be argued that must of the exack is subject to relatively small plastic strains, so that aft ofthe lower part ofthe stess-steain curve is the m Res (conseqay ose? ase pos ao ee x ceor zoho - as] 208 8 we a ee 12s es eee oeLtA 6 on tee ono 2 0170 uno 0] ose 2 : * “ 7 g | g 2 = 3g » a a gol g ' = 3 a ae) 20] aa . 109 — re) 7 . a em wo sens ow cn Lover oo : Pre 412. Iaailty reiton according 0 Equations (417) forthe wo eases in igure 411 Fur 4.11, Frc es prediction fo dierent Corso center rahe pal 15) Above . US Above: Case 1: Case? of Figure 41 (Coury EMAST Feouves, Below preiced face tess and fracture popes (Coury EMAS)m2 tip material is subject to strains in accordance with the upper part of the s-sirain curve. Common sense indicates that the decision must be made on. iaterial-by-material asi there is no categorie answer. Any conclusions arcived at certainly should not be generalized, and if they are. they are sill restricted to materials not obeying an exponential stress-strain curve; for other materials there is no choice. When there isa choice, the criterion for the choice is whether the fracture stess is predicted correctly: the value of J is irrelevant. ‘The problem would not exist if the measurement of J, would use Equation (4.15) instead of (4.20), because the reverse operation of Equation (4.16) would then automatically lead to the correct answer (use of same F and i as in Equation (4.15). 4.10, Historical development of J helby [10] defined a number of contour by virtue ofthe energy conservation theor of these integrals can be written as fra rare o where (43s) v = fox jegrals which are path independent in, The two-dimensional form of one being strain energy per unit volume. The integral is taken along a closed contour. 5 followed counter clockwise (Figure 4.138) in a stressed solid, Tis the tension perpendicular to 5. w is the displacement, and ds is an element of S. ‘Although the equation is somewhat elusive, it can he seen that the fist term represents strain energy, while inthe second term T is the ‘force’, and dj a strain, so that (dujdx) dv isa displacement. As force’ times displacement equals, the work, F. done by the force, the equation essentially states that § Which is energy conservation criterion of Equation (1.32), Applying this integral to a cracked body [II] one can construct a closed contour ARCDEF around the crack tip. as shoven in Figure 4.13. The integral ‘of Equation (4.35) along this contour must equal zero: it consists af the sum of four parts febepegee (39 Since T = @and dy = 0.along CD and Ft the contribution of these parts is 13 Piare 413. Conlon intel) Eat hy: h) Body with rack) Path independent conto frfean fon] (an Therefore. the contribution of ABC must be equal (but opposite in sign) to the contribution of DEF. Note that one is clockwise, the other counter-clockwise. ‘This means that the integral i taken in the same direction along F, and F; will have the same value: f,, = J, in (Figure 4.13). As [and I’, were arbitrary paths, the integral over'T is apparently path-independent (one may take any path, beginning and ending at opposing crack faces, and the integral wll always hhave the same value), The value of the integral was called J de [ray 1a ws (39 11 should be noted that there is no proof as yet that this Ji the same as the one used in previous sections; thus, for the time being it should be considered as the“4 definition of a new quantity defined only by the value of the above integral. The path-independence is not of apparent relevance to fractu {the integral is path independent any convenient contour may be taken to des mine its value. The simplest contour is a circle, as in Figure 4.14, with center at the crack tip. For this case y = r sin @, so that 10, Then Equation (4.38) becomes: radws 4, dy = 1 608 040, and d: 10 fff} oo rehean am No matter what the relationship between a and c, tv integral Jo de at any point always evaluates to a,o¢, where a, is dimensionless. Since Tis a stress, it can always be expressed a8 Co, and dujdx being a strain can always be expressed as Cyt. Then T (du/dx)evaluates as C,Cyez » 3.°> Rohm, anda, may depend upon @, but regardless of how complicated th. ~ pendence, the integral of Equation (4.39) will be: J = | (0e0,(0) cos 0 ~ axa} ra (440) No matter what the functions of 0 are, the integral reduces to See erent aan Pure 616 Simple circular path for onto integra us ‘The determined integral between —x and x will evaluate to a dimensionless umber. The solution therefore willbe as shown in Equation (441), while o and. ‘cmay be defined at any convenient point. Taking another point as a reference will merely change a, and a;, and therefore O, but Q is only a number anyway. Wone defines @ and cat the location y = 0, x = r (Figure 4.14) then: J = ard. (4.42) By using the Ramberg-Osgood equation for the stesé-strain curve, namely "TF, Equation (4.42) yields: Js ot UF. (442) ‘Then finaly « gy aay Wn = 1 = B) this reduces to: «= (aay vou Compa this equation with Equation (3.2) as replicated below: K Par Clearly, Equation (4.44 is the same as the very original Equation (37) shows that for m= we have Q = J2x, and JED = K. This meansfor 1 that 1 = K'JE = dUjda. Thus the path-independent contour integral is but the Strain energy release rate, It is now apparent why this integral was denoted by J. symbol already used for the energy release rate. The above being the case form = 1, itwillbe tein general, because nowhere above was a restriction made with regard tothe shape ofthe stes-srain curve ‘Thus, the only significance ofthe ‘path-independent’ contour integra isto show that it represents the strain energy release rate, = 4U/da, so thal the J-integral is indeed the same as J defined previously. Indeed, that is of secondary importance only: the eneray release rate can be defined in a much simpler way as was shown earlier inthis chapter. Nevertheless, the integral has ils use. As itis known now that it equals dU/da, ita be used to caleulate G, Jor K. It ean be applied to the results ofa finite element analysis (most codes have post-processors(o do this). Ifthe analysis is forthe elastic case onecan obiain K from K = /EJ. When the analysis uses the Ramberg-Ovgond it cam provide the geometry factor H trom: *, (445)116 FL ae (446) ue Fe This is essentially the way in which H (and fh; Section 49) were obtained. 4.11, Limitations of EPEM. ‘Although the concepts discussed in this chapler are generally referred to as clastc-plastc, they are in fact elastic. The non-linear stress-strain curve used ‘must apply for loading as well as for unloading. Thisisillustrated in Figure 4.15, ‘Truly elastic-plastic behavior is shown for comparison. Inthe latter case linear unloading occurs, and there will be a remaining plastic strain when the stress is reduced to 7ero. The stress-sirain relation used in EPFM, upon unloading, must produce zero strain at zero stress. The curve may be non-linear, but not plastic. The return to zero strain means that the material is merely non-linear-BLASTIC, ‘This does not put any restrictions on the use of Jas long a there is 0 ‘unloading. Without any unloading it would never be known that the curve for unloading is different from the loading curve. However, if there is unloading anywhere, the assumption of non-linear elasticity will eause errors. Consider for example Figures 3.15 and 4,160, It was shown in Chapter 3 Figure 3.15 that for linear elasticity the energy release extension of a by da i always equal to the change in s whether there is constant load or constant displacement, In one case the does work which is twice the ine ergy. se that the rem ‘of the work by the load (the energy available for Fracture) is exactly equal to the ‘change in strain energy. In the case of constant displacement the load does no work, but the strain energy decreases, so that the released energy is agsin equal tothe energy regardless of id not if the material is elastic-plastic. Dur term Pine 18. Rass of PEM, “7 3 “ o ‘Fie 4.16 Sein cere clese upon acute by da under constant plcement a) Nom ines ‘laste: () Eli platie load the energy available for fracture is equal to the work done by the load ‘minus the increase in (plastic) strain energy. However, during linear elastic unloading the plastic strain energy remainsin the material. Only the smal elastic part of the strain energy is indeed released: the available energy is much less ing constant displacement than during constant load (Figure 4.16b). Hence. strictly speaking. the resulting equations are valid only in load control, so that isis meaningful only up to the point of maximim load: this isnot, ‘severe restriction. because in most engineering analysis one will be interested only 38 a cracked structure can sustai Yet, even the later is somewhat questionable. The material atthe crack tip is highly stressed. When stable fracture is occurring (up to maximum load), it is this highly stressed crack (ip material that is unloading after the fracture has passed through. Hence, the errors due to the unloading assumptions are felt ‘most strongly where it counts most: at the crack tip. The errors will be small ly, but increase with increasing Aa. This sheds doubt on the analysis of stable growth and instability, As a matter of fact the analysis will be only meaningful when the J curve rises very steeply, so that there is very litle stable fracture before maximum load (Figure 4.2) It has been suggested [12] that stable fracture should be limited to just a few percent of the remaining ligament, ‘otherwise the errors become considerable. Using the path-independence of the J-integral it was shown thatthe crack tip sress fied in EPEM is described by Equation (4.44). For an ideally plastic ‘material (n -+ co) the equation leads to a finite crack tip stress, but forall other values ofm the stress is lll infinite at r = 0, It was pointed out already that the Use of Equation (4.14) leads to an infinite fracture stress for a —+ 0, and therefore will be increasingly in error for smaller cracks; approximations for18 imal cracks will be necesear> > a8 in the case of LEFM. Por this and other reasons, a discussed, collapse sill must be evaluated separately s a competing condition. ‘Apparently, BPFM has cured none of the ills of EFM; itis a mere extension of LEFM for n # 1, A host of modifications to J have been proposed [13]: most Of these belong in the category of “patch work’ ay much as do plastic 7one corrections 10 K [6]; they provide few new insights. comolcate the procedure and lead to only marginal improvements. ‘Nevertheless, EPFM is still very useful and has a definite place in conjunction with LEFM. It has extended the use of fracture mechanics to non-linear ‘materials at least up to a point. Judicious use will prowide meaningful engineer- ing answers, provided collapse analysis is done as well, just as in the case of LEFM, and provided appropriate approximations are made for short cracks (Section 3.8). It may be noted again that the accuracy of fracture analysis need ‘only be as good as that of general accuracy “f engin*1ing proced spters 12.14). With theemes;, st geomesry factors 8, SJ. and the por staiy for ple estimates of these (Chapter 8), EPFM has become a useful engineering tool. 4.12. CTOD measurements: ‘An alternative approach to EPFM has been based upon the Crack Opening Displacement (COD). Although referred to as COD. the method actually ‘employs the Crack-Tip Opening Displacement (CYOD), ‘Consider a crack tip in a stressed body as in Figure 4.17, Let forces be applied ‘over a distance da behind the erack tip in such a manner that the crack just closes over a distance da. The crack is now shorter hy da and, therefore, the required closing forces must be equal to the stresses normally present when da is uncrucked. For the time being, assume that these stress are approximately ‘equal to the yield strength ing their closing action the forces travel over the distance v. Therefore, they do work to the amount of dF = 2 x 0.5F,v da (plate of unit thickness). htt Pare 4.17 Closing force ches the crck tip er 19 Since » is related 10 CTOD, the work will be dF = aCTODF, da. Upon release of these forces the same amount of energy is released, and the crack will ‘grow’ again by da. This energy release, dF/da = «CTOD F,,,is what has been called the strain energy release rate G or J. Therefore: @ = aF,CTOD or J = af, CTOD. (447) The firs of these expression would be applicable for LEFM, the second for EPEM, Q Naturally, the stresses over the future da are not uniformly equal to Fy as assumed. but if they are not, only the dimensionless factor a willbe affected. It turns out {6] that ais approximately equal to unity, but various interpretations would put it between n/4 and 4/n in LEFM. However in the case of J the value of e depends upon 1 [4] In any case, the above equations lead to: (LEFM) : (448) crop = 2 (ePrM) F, Fracture occurs at x critical value of G (or K) or a critical value of J. Then, according to Equations (4.4R) fracture takes place ata critical value of CTOD, defined as CTOD,, Consequently, CTOD, should be 2 material property characterizing fracture resistance: as such itis a descriptor of ‘toughness’, and the measurement of CTOD in a test would provide the material's propensity fracture. ‘A test for CTOD measurements was standardized first in Great Britain in British Standard BS-5762. Essentially. the testis performed on a small three- point bend specimen, (Figure 4.18). Asin other toughness tests a record is made ‘of load versus crack mouth opening displacement, The critica crack tip opening. displacement CTOD, usually referred to as COD. can be obtained 2s follows. The ligament is assumed fully plastic so that all specimen deflection can be considered to be due to rotation around a plastic hinge, the specimen limbs rotating by rigid-body motion. Ii is assumed that the center ofthe plastic hinge coincides with the center of the ligament, then CTOD can be obiained from the ‘crack mouth opening in the manner shown in Figure 4.18. tis not necessary 10 make the assumption thatthe center of the hinge coincides with the center of the Tigament. One possibility is to determine its location experimentally. Other ‘options are open {14. 6] Knowledge of the critical CTOD per se, is of no use for damage (olerance analysis. The number will ave: to predict fracture in a structure. For this to be possible CTOD must be ‘expressable in terms ofthe stress acting in a structure, so that the (ractue) stress120 00m ° Prone 8 Cratk opening dipncment et) COD sts: (b) Mesa COD, oni Gro nb " stro ea ee Cae ar Fone. Oota forever | (Lod dglacement eased te) kre wo al reine. crack can be calculated ns that alrese at which the CTOD of the str teaches the critical value. So far only empirical relations between stress (strain) land CTOD have been developed. There is no objection against their use. as Fong. fas they are reasonably general, However, with the aid of Equations (4.47) 121 the measured CTOD can be converted into Jy or G and then the result can be used in accordance with Eqs. (4.16) or (3.25). Certainly, Equation (4.47) is a approximation, but can be oblained as a function of» [}. Besides, an approxi ‘mate general equation is more likely to be useful than a specific empirical one. ‘The development of the simple EPFM procedures discussed in this chapter hhave made the semi-empirical CTOD approach somewhat obsolete for damage tolerance analysis. However, the CTOD testis a useful extension of EPFM, as it can provide Jy or K, through Equation (4.47) from tests on small specimens. 4.13, Exercises 1. For a test bar of a certain steel one measures a load displacement curve as in Figure 4.19. The original diameter of the cylindrical bar is 0.4 inch; the original length is 4 inch. The final thickness in the neck is 0.28 inch, (a) Determine the engineering stress-strain curve, F, and F,, (h) Determine &, Fand 1 (6) Determine a, 2 and cy for a4 = 1OOksi (4) Do the same for a5 = SOksi {€) Check whether (a) and (b) lead to the same strain forat lest two different stress values. e.g, 50 and $5 ksi 2, Determine the true stress-irue strain curve for the problem in Exercise | up to the point of maximum load. Repeat questions (b), (),(d) of Exercise | 4. The bar of Exercise | is unloaded at P = 7.3 kips and testedas a new bar in ‘anew test. Determine the new load-displacement curve. For this cold-worked ‘material, assuming that F, coincides with the load at unloading, calculate F, and Fy. 4. A test om a center cracked panel with 2a = 4 inch, W = 32 inch shows a Fracture stress of SOksi. The collapse strength is SSksi. Given that W = 20 inch (500mm), B = 04 inch and assuming E, F and 7 as in Exercise 1, calculate J, at fracture, Or did failure oceur by collapse? (Neglect J,) H = 87. 5. Using the results of Exercises 1 and 4, calculate the failure sress of a panel 24 inch wide, with a center crack of 2a = Ginch. H = 13:5; sume that the J, can be neglected. Does failure occur by fracture or by collapse? iven that Equation (4.30) for a center cracked panel is: J = aoyey( — (22/W))ah,(P/P.)"? with P= a1VB and Py= o4(W — 2a)8, and that h, for JW = 0.12Sish = 4.13, Calculate the fracture sess with the information obtained in Frercise le (a, = 100ksi) and in Exercise Id (04 = SOksi) and show that the results are the same, independent ofthe arbitrarily selected og Assuine J = 2kips/in and neglect Jym2 7. Ina Jy test on a CT specimen the load rises almost linearly to Skip upon ‘which Tracture begins. and upon which the load-displacement diagram ‘becomes essentially tistzontal. The initial crack size is one inch; the displace- ig begins. The crack (fracture) size reaches and 1.15 inch when 5 = 0.27 inch. Calculate ness is 0.5 inch, W = 2 inch References U1 DP. Roske an DJ: Cai, Compendon of Sr ene ates HOM, Sniney Oceano 0. (CC 3h Ht of es men fan Int Ea ok Meh Lbih Un 119, UW i Tact Ps mah of ck brs Dl Res Cor 97) 1 ¥- Kamara tore strh fo oc Res nt ep cy 19 ara ca desl pl fc ah hi one kt. Rep hana 19 Bet Hoenig ft me on, Ne ON) 17) Aeon Sind ha che mao mf tt gis, AEM Swede UE Ohnand MLD, Carman, ystems fr menace chi of tk i Jetty he HAR stare, OF Teck Ron (1 AS tecman and A. Met, Peo sa i bn Wey 963 110) 20 shy. Cats rs or pc. er mh Shea New 9 py. (1) eA ath pene te nd thao a Byam aad ake? Mh a p78 121 2Wthchnon snd CP Sabha ted ck owt STA STP AMR) prise tu RE Keoinen nd C1. Pope, Adon fcr mechani, od, es (188 1] CC. Veoman nd ute The enn of opuent oon tinned end fein Ep Ptah 4 (17D ppt 1s) eck sya ko, PCF cnr rates Vo FMAS Coney. 8-0. mee CHAPTER 5 Crack growth analysis concepts S.L. Scope {mn this chapter the concepts and procedures for crack growth analysis are used. Fatigue, being technically the most importam crack growth mechanism covers most of the chapter (environmentally assisted growth or ‘combined stress corrosion and fatigue i integrated into this diseussion), Stress corrosion cracking by itsell:" overed in but one section; this is not because it isnot considered important, but because stress corrosion cracking is practically covered by prevention and not by control, while analysis procedures are essent- ally similar (o those either for residual strength (Chapter 3) or fatigue. Fatigue crack growth on the other hand can hardly be prevented in many structures; must be controlled. * discussions cover the concepts of erack growth analysis, retardation and ‘special effets, as well asthe analysis procedure, Examples of analysis are given, However, fatigue crack growth analysis is a complicated subject, and the discussions in Chapters 6 and 7 should be read as well, before attempts to analysize crack growth are made: 5.2. The concept underlying fatigue crack growth Cyclic stresses resulting from constant or variable amplitude loading can be described by two of « number of alternative parameters, as shown in Figure 5.1. Constant amplitude eyelic stresses are defined by three parameters, namely a scan sttes8, 4,8 stress amplitude, o,, and a frequency cv or v. The frequency isnot needed to describe the magnitude of the stresses, Only two parameters are sulicient to describe the stresses in a constant amplitude loading eyele. It is Possible to use other parameters; for example, the miniumum stress, mu, and ‘the maximum sires, ¢ay, describe the suesses completely, and so does the stress ange, A¢ = Own ~ dnin. in combination with any of the others, excepl 0. ‘Almost any combination of two of the above parameters can completely define14 BI ZI V qs cy) Fue 5.1. Parameters for fie crack growth (2) Blting ar sesbarpeing(b A and ARC) ‘Stew rate the eycle. Note that in the above the Greek letter A is used to indicate a RANGE ofthe stress. This i not in accordance with the normal use of Ain mathematics, where A indicates a small change. In this case A stands for the total range of tress in 8 eyCle: B= dour ~ Snun Which need not he small at all. t would be better to use eg o, forthe stress range, but since the denotation & has become common practice, it will be used here ws wel. "Another parameter is often convenient. This is the so-called stress ratio, R. defined a8 R = Oni/Pau-One of the above parameters can be replaced by R to define the cycling, For instance, any ofthe following combinations fully defines the stresses: Aa and R. dan and R, One. and Ra, and Roa, and R. The case of R= O defines a situation in which the sttess always rises from, and returns to 0. When R= — I. the stress eycles around zero as a mean, which is called fully SCE TE rat 1s feversed loading. Note that R = Gxi/@mn = (me ~ B¢)/¢nus an $0 that Ao = (I~ R) gy and COMVETSEYY: One, = ail R). ‘Crack growth life is expressed as the number of cycles to grow a fatigue crack over a certain distance, The number of eycles is denoted by N. The crack growth mechanism, as discussed in Chapter 1, shows that a fatigue crack grows by a minute amount in every load cycle; the mechanism is shown schematically again in Figure 5.1a, Growth is the geometrical consequence of slip and crack tip blunting. Resharpening of the crack tip upon unloading, sets the stage for growth in the next cycle. It can be concluded from this mechanism that the crack growth per cycle. Aa, will be larger if the maximum stress in the cycle i higher (more opening) and if the minimum stress is lower (more re-shat- pening). so that Mag secl FOF agit ANUJOr nail 6H ‘The subscript F indicates the local stresses, atthe crack tip, 1 stands for larger, [for fower. Note that in this case the Greek letter A is used in its normal” sense, meaning that it indicates a small change: growth from a to a + Aa. tn the previous paragraphs the stress range, Ac, was defined 98 dain. The stress range wil be larger when oyu is less, so that the above equation can also be written as Betas et FOF aust andor Bart (52) ‘The local stresses at the crack tip can be described in terms of the stress intensity factor X, discussed in Chapter 3, where Ks fo Yna.if is the nominal applied stress. In a cycle, the applied stress varies from dan (0 dan, over a range Ba. Therefore, the local stresses vary in accordance with Kean = Ba SR Kon = Na Si 6 AK = fio Jaa Again. the Greek letter A stands for range and not for a small increment: the denotation K, would he better and less confusing, but AK is used here in conformance to general practice (sce also Figure 5.1b). With the use of Fqs (5.3) the Equation (5.2) for erack growth becomes: AMdprosel (OF Kant andor AKT (4) the stress ratio is defined as R= that at any given erack size a, the Wain = FRAP Opa SR ne / nn W appears from Fs (5.3) exe ratio ie alto equal 10 Kya Kg HCE (55)126 ‘According to Equation (5.4) there is more crack growth when Kin is higher. It follows from Equation (5.5) that this isthe cace whion AK is larger andjor R is higher, so that Equation (5.4) can be written as: Adpronet for AKT andjor RY. 66) In this equation Aa is the amount of crack growth in one cycle, which would be expressed in inch/cycle or mm/eycle if growth were measured over eg, 1N = 10000 cycles, the average growth per cycle would be Aa/AN, which isthe rate of crack propagation. Inthe limit where N+ 1, this rate can be expressed a the differential da/AN. Equation (5.6) indicates that the rate is a rising function of AK and R, so that the proper mathematical form of Equation (5.6) a 4 = 10K. ®, 67 As shown, Equation (5.7) derives directly from the model of crack growth siscused in Chapter! and shown in Figure 5.1 ‘5.3. Measurement of the rate function ‘According to Equation (5.7) the rate of erack prrwit:ill Larger for higher AK and higher R. The actual functional form of Eumation (5.7) can be derived from the crack growth model in Chapter |. However, this model - although ‘qualitatively in order - is a two-dimensional simplification of a three-dimension- al process that is extremely complicated due to the presence of grains with different orientations, grain boundaries. particles. etc. As a consequence, a rigorous mathematical description of the mode! is not well possible, and @ ‘reliable functional form of Equation (5.7) cannot be obtained from theoretical ‘analysis. This leaves only one possibility to obtain the function: interrogation of the material in a test. Although this might seem objectionable to theoreticians, it should be noted that ALL material data are obtained from tests, such as F, and F,, and even the modulus of elasticity E. Crack growth data are obtained by subjecting a laboratory specimen to cyclic loading, The specimen may be of any kind as long as lis known, so that the stress intensity factors can be evaluated. Most commonly used are center cracked panels and compact tension specimens. The following examples are for Aacenter cracked panel, As long as cracks are small with respect to panel size (e.g a/W < 04) the geometry factor, fis approximately equal to one, so that K = ora {A panel as in Figure 3.2 is provided with as small but sharp central notch, so that cracks at both sides will start almost immediately. The specimen is subjected to a cyclic stress of constant amplitude in a fatigue machine. First nr o 3 . fog 8, t ® D "082g eat pines ing Mend se Ra conse set yee With = 0,40 that R= tought he et. Abo, mm = Ae in that case. how ie . Crack growth is monitored throughout the test by measuring the length of the crack at itera of eg 10000 pels, The euler plated aba oe crack growth curve on Figure 32% Tiss alte information th sont extracted eircom he est must be inetd forte deen of th Tomo guaon (5), Conder a sal cack nrement, ao the crue.) Acorn tothe mesure eve it took AN, cfr ne rch to grow over as, Tet the taf rowth (la) Preamp ie cack rene. ch and this om took AN 10000 yess then the tte hvala 0.1/10000 = 1 x 10 Finjeycle. i * eae) The ete I fo otin te gtonh le denenene mon A, which tcuies determination ofthe aes itensy range The sees ack sea beso, Theses range so that OK = Maoyaas Apres wae AK Ak greed crack fromth a's rat of (lah Te ree128 ram with dajd N= Aal/AN)and AK along the plotted as a data point ina di axes, as show Sie ‘The above procedure is repeated for a number of points along the crack ‘growth curve. Ata larger crack sizea,,an amount of growth At takes only AN, cycles. Because the curve is sleeper, the rate is higher: as a; > aj, also AK; > AK,. Hence, a larger AK indeed produces a higher rate of growth. A. plot of the data points as in Figure 5.2c confirms ths. Because differentiation is a very inaccurate procedure, large ‘scatter’ may ‘occur in da/dW. This problem is discussed in Chapter 7. Wis the reason why in practice da[4N is obtained as a running average of 5-7 points along the crack ‘growth curve. Figure 5.2c provides the growth rate for any given AK. In Chapter 3 it was shown that the crack tip stress distribution is unique and depends only upon the stress intensity facto. If at two different cracks in the same material have equal stress intensity then the (wo erack tip stress fields are identical; there is similitude, Hence, if the stress intensities are cqul the response of the cracks ‘must be the same, This means that the erack growth rate will always be the same, if OK is the same. Thus, Figure $.2c is the material's rate response in all cases IW can be used to analyze crack growth in a structure built ofthis material The validity of this similitude argument can he checked by performing second (est on a similar (or different) panel, but with a different Ag. The erack growth is measured (Figure 5.3), the results analysed in the same manner as before, and the data of both tests plotted in the sume rate dingram (Figure 5.3). ‘The rate data oF the second test will fall om the same line as the data of the first test. This confirms that the same rate was obtained at the same Ain both tests For example, take a point on crack growth curve | ata crack size of a = 0. oe se — peed 7 « ete $1 Date io ews Tesi nt F761 Rom eer racked pe Te mcd ts gb rede data 129 With a stress range of Aa = 17.6ksi the value of the siress intensity range is AK = 11602 xx = I4ksivin. In the second test the siress range was 40 = ILARSi, This would. produce a stress intensity of I4ksiginy if @ = 238 x02 = 048in (assuming = 1) in. the second test, a 114/048 %m = 14.0ksi Jin, still assuming = 1. This means that at a crack size 02in in test 1 and O-4.in in test 2, the stress intensities were the sine, so thatthe rales shouldbe the same, Figure 5.3a shows thatthe slopes (rate) ofthe curves at these two crack sizes ae indeed equal (naturally followed immediately from the fact thatthe two tests led tothe same dad ~ AK diagram), Similitade in behaviour i hereby established, The results can be used fo analyze crack growth ina sructore “The tests diseussed so far were all atthe sme stress ratio, namely R= 0. ‘According to Equation (5.7) the rates alo depend upon R. Thisdependence can be assessed by performing tests at diferent values of. The els are plotted versus AK. to oblain data such asin Figure 54. Indeed, higher R produces higher growih rates as should be anticipated on the basis of Equation (57) The data in Figure $4 show that the effect of Ris smaler than that of AK, but that, is simply the way it comes out Data are alway ploted on logarithmic scales of log(A.K) and log(dajd). Pine $4, Cock prow rates for TL6AL-AV: Mill Anal Lab Air 7OF (210) Efecto Ratio,be because the rates vary over several orders of magnitude. A log scale for AK is ‘not strictly necessary, but it has become standard practice to use a logarithmic scale for AK as wel Crack growth properties of a number of structural alloys are compared in Figure 5.5. Environment, loading frequency, and temperature may have @ significant effect on growth rates. Examples af ron of these effects are shown in Figure 5.6. For a discussion of erack growth «: segative R see Chapter 7 S.A. Rate equations ‘The form of Equation (5.7) follows from the test data; it cannot be obtained from a theoretical model. Naturally, a functional form can be established by fiting a curve through the test data. The resulting equations are sometimes useful as they eliminate the necessity of using a raph From Figures 5.3 through 5.6 it appears that the rate data for one particular Reratio fall more oF less on a straight line in 2 logarithmic plot. The equation ae ie ($8) = moan) + 69 on Scam Faure 88. Type ae properties of dierent alloys t ‘ 1 ‘ ‘ 1 t ‘Taking the anti-og provides: bt ond 86 oes ay : a Par 56. Type elt of evionmen an emprare oak 69) ‘This equation is generally Known a5 the Pais equation. ‘The parameters ('p and * two points 4 and Bin cn be determined ently Fr example, win ‘igure 5.7, yields: pie sing the jer) Roar Pwr 7 eng132 point A:log (1.6 x 10-7) = mplog 6.3 + log Cr (510) point B: log (x 10 *) = mplog 40 + log Taking the logarithms provides 68 = 08m, + log C 4 = Ng + op Ce way 24=0. This provides my = 3; substitution of which in one of the equations leads to log C, = =9.2 oF Cy = 63 x 10°", The rate equation becomes da M0 6310 *aK? (512) a js between 3 and for hi particular mater, Formos! materials the value of a has also largely 5. The value of Cis more strongly material dependent: ‘ifferent values in different unit systems (see Chapter 7). ‘The Paris equation covers only one R-value. The lines for different Rare often parallel, ie. have equal slope as e.g. in Figure 5.4. Thus all these lines would have the same m-value, but diferent C; the latter depending upon R as C(R). Hence, the following equation could cover all R-values: da da 7 5.13) an 7 CURA" (5.13) For many materials, the dependence of Con R can be described in a simple £ ake. (5.14) an ~ @~ ‘Where C. is the value of C when R = 0. Equation (5.14) can be used ass. Often itis further modified by substituting K%, = (AK/(I — R))". so that = CAR Kings where m, Equation (5.15) essentially reverts hack to the original Equation (5.4): of course Fquations (5.14) and (5.15) are equivalent, and both are in use. ‘One may argue that fracture occurs when the maximum stress intensity in ‘cycle equals the toughness, 16. Kao, = K, oF Ky Since Kyuy ~ BAI ~ Ry this would happen when AK = (I~ R)K,. At fracture the growth rate would tend to infinity. A functional value can be made to go to infinity through division by zer0: 133 da aK” an > OTT R aR AL fracture. where AK = (I~ R)K,, the above equation indeed provides an infinite growth rate. This equation is known as the Forman equation, It shows the growth rate to depend upon R and should therefore apply for all R-values: the equation ‘pretends’ to ‘know’ how daldN depends upon R.A strong. objection against the equation is that in many cases fracture isnot governed by the toughness, because of collapse (Chapters 2 and 3). In addition to Equations (5.9) through (5.16) many diferent curve fitting ‘equations can be developed to describe the test data, As a matter of fact there are probably as many equations as there are researchers in the field Several ‘others are in common use. But none ofthese, nor the above equations, have any physical significance; they are merely curve fiting equations. If they do fit the ‘data property, there is no objection against their use. But no equation can fit all data, so that religious adherence to one equation is not advisable. One should use the equation providing the best fit ina particular case. An equation may be used if convenient, but direct graphical use ofthe rate diagram is just as reliable. Most crack growth analysis is done by computer, which can be supplied the rate diagram in tabular form. It makes little difference to a computer whether it interpolates in a table or uses an equation. It should be neted that the parameters for the various equations are different, even if they cover the same data set. For this reason the coefficients C,, C,, and C, and exponents my and m, are used (0 indicate that they are specifically for a certain equ (Paris, Forman, Walker). Use of the parameters of one equation for another ~ ven for the same material ~ may lead to dramatic errors. Conversion of Parameters to other unit systems requires great care (Chapter 7) (5.16) 8.5. Const amplitude crack growth in a structure ‘Most structures experience some form of variable amplitude loading in which cease the crack growth analysis is considerably more complicaled than for constant amplitude, as will appear later in this chapter, However, in the few cases of constant amplitude loading the analysis can be readily performed with ‘oF without the use of a computer. A crack in a structure will grow at the rates indicated by the rate diagram because of the similitude discussed. Analysis of structural crack growth can be carried out if the geometry factor is known (Chapter 8) for the structural configuration at hand. The erack growth (curve) in the structure follows from an integration of the rates: do da GW AK RY on aN = Se 17)da CpaRm gin Generally, the integ done numerically; it can seldom be performed in closed form, because of the complesity of the Funct ns f and fin AK, and of the stress history. The function f might be as simple as the Paris equation Ler oO ene Oe The f fora structural erck is usally a lengthy p\nomial in a’ of known only in tabula form, so that numeral integration and da is constant (independent of o) Imes through the use ofa computer, but in the case of constant smipitde loading + hand computation is very well possible. The principle for a simple numeral tegration inthe case of conlant amplitude foading it shown tel 1 the louding is of constant amplitude, the integration can be done in smal steps with litle err; the step size might be Laken as eg a crack increment of one percent ofthe curren eruck sie. Assume for example (Figure SA ens of constant amplitude loading at eg. Ao = 20ksi further assume that a Pars mn applies with da/dN =6.17E~ 10 AK". Let i for the structural crack nas f = 1.12 + (a/H’Y (a hypothetical case) and let W’ = 4 inches (gue 58) The fist wo steps ofa calculation star ta rack size of 7S inch are shown below: Totally N= Ocyces ‘a = 0.01 x 0.75 = 010075in (one percent increase) BK = (112 + (0.5/4) x20 x yi “SH = IS.5ksi a)AN = 6.17 x 10" x 35.5 = 275 x 10 Pinjeyele; AN = Aafida/N) = 0.0075/(2.75 * 10 *) = 273 cycles: N= N+ AN = 04279 = 273 cycles + Ba = 0.75 + 0.0075 = 0.7575in; a = 0.01 x 0.7575 = 0.007575in (one percent) AK = [112 + 07595/8} x 20% ORR 4af4N = 6.17 «10% 38.77 = 281 x 10 Sinfyce: AN = 0.007575)(281 x 10 *) = 269 cycles; 35.7ksi 1s o ) 3 erro "AK? 7 100 AK (saiVind Franc 88, Crack promth analysis: Hxample fr constant mpi. a Sacre; (b Hypothetical Pure: () Matera rate data N= 270 + 269 = $42 cycles; a= 0.1575 + 0.0076 = 0.76510. ued until the crack size ais reached. It can be programmed JAK, RY a, = $4, (08. 5 = 001):136 : 137 AN, = baftdajd), N= N+ AN: = a # Bags mur sie 961 2008s ° 108 489 206 ose aa 68 amo if a, < ay then return to beginning. 12: R= 0 With the steps on the order of one percent of the current crack size (or fixed step sizes if s0 desired) good accuracy can be obtained. If large sleps are taken an integration rule such as Simpson or Runge-Kutta should be used. These inte- eration rules were divised for numerical integration prior to the computer era, ‘when large steps had to be taken for hand calculations. In (rue integration where the limit da —+ 0 no such rule is required. The computer ean indeed let the Sep size go to zero, so that the integration rules are not necessary. In an emergency constant amplitude analysis as above can be done by hand in rather large steps as shown in Table 5.1. Even without the use of an integra tion rule, reasonable accuracy is obtained when the steps are not made too large. especially a the beginning (Table 5.1). The results of Table 5.1 are compared in Figure 5.9 with a computer analysis of the same problem, showing thatthe hand calculation gives a reasonable approximation. The procedure illustrated i ‘Table 5.1 lends ise very wel for execution with a spread sheet program. or 3 dedicated program such as shown above can be used. Integration isa forgiving procedure in contrast to differentiation. The latter problem is discussed in Chapter 7. Most structures being subjected to variable amplitude loading (Chapter 6), the above procedure for constant amplitude integration is seldom applicable, Crack growth analysis for variable amplitude loading wil be discussed later in this chapter. Such analysis is much complicated by the problem of load inte action and retardation as discussed in the following section : 812 065 am a 104 soa an 5 nn ae ns 3a 10 as «10 390 10 13 «10 67 «0 10 « 10 rns 2 x 10 399 «10 16 928 2670 sar 607 928 ‘56. Load interaction: retardation When one single high stress is interspersed in a constant amplitude history. the crack growth immediately after the ‘overload! is much slower than before the overload. Figure 5.10 shows how three single overloads increase the crack growth life by almost a factor of five (compare A and Bi), After a period of very slow growth immediately following the overload, gradually the original growth rates are resumed. This phenomenon is known as retardation. A negative load : following the overload reduces retardation but does not eliminate it (compare Band Cin Figure 5.10). Crack growth analysis for variable amplitude loading. is not very well possible without an account of retardation effects, Before such -count can be made, retardation must be explained. Consider (Figure 5.11) a crack subjected to constant b= ee aga: 1.02 Lon Los wae ae apa ous 030 oss Las 0 a 5 3 i 2 i i q & : 8 3 g q ? i Be 03 os os ‘Same elation wih ner steps during eaiy growth plitude toad138 ‘Figme 39, Compaison of computer analysis an the hand analyte of Table $1 AANA 55:82, 55733 eanllyn orm Sma8e canal pwn ferres Sa Fique 510. Retardation ae overload (Courtesy Airerall Engineering 0; during the firs cycle the load varies from A to C through B. Before the lnading starts, imagine a litle (dashed) circle (Figure 5.11a) at the erack tip, indicating the material that will undergo plastic deformation in the future plastic zone (Chapter 3). The plate is then loaded to B. Imagine that one could now 139 on" @--- @ Pore 5.11 Resid sees rack tip (a) Load at A: (b) Load a B() Load at Bs plastic ome emoned (Ld at: plastic one ail ou; (e) Load aC: plastic none back remove the plastic 7one and put it aside (Figure 5.11e). Afler unloading to C the situation of Figure 5.11d is reached: all material is elastic - the plastic material hhaving been cut out ~ so that all strains and displacements are zero after unloading. Hence, after unloading the hole at the crack tip in Figure 5.1ld is. qual in size to the dashed circle in Figure 5.11a. The plastic zone has become Permanently deformed: itis larger than before the loading started. Thus it will fot fit in the hole of Figure 5.114. In order to make it fit it must be squeezed ‘hack to its original size for which a stress at least equal to the yield strength is needed (it must be deformed plastically again in compression to be squeezed back to its original size; this requires stresses at least equal to the yield). The plastic zone in fatigue loading is very small. Most ofthe f rowih takes place at low values of AK as can be appreciated from the data in previous figures. At high AX the rates are so high that litle crack growth life isle; most of the life is covered at low AK. Iffor example, AK = 10ksi ' material with a yield strength F, = SOksi, the plastic zone size (Chapter 3) would be rp = 107/(6x50") = 0.0021 inch. As the remainder of the plate is laste and returns to zero strain after unfoading, this small plastic zone indeed will be squeezed back to its original size and made to fit in its serroundings. In order to squeeze the permanently deformed material back to is original size140 reverse yielding is necessary: the compressive stress must be at least equal to the Yield strength. Whether the plastic zone is taken out as hypothesized above, oF whether it remains in place as it normally will the end result will be the follows that after unloading, there isa compressive stress at le the yield strength at the crack tip. This isa residual stress (no external load) which must be internally equilibrated by tension further away. The residual stress distribution is as shown in Figure 5.1¢, During subsequent cycling this residual stress system will be present upon each unloading and it wil have to be ‘added to (act together with) the applied stress. The crack growth response by the material automatically aecounts for this residual stress system; the data in eg. Figures 5.4-5.6 already reflect its effect, because the material ‘knows’ about these residual stresses and shows growth rates in accordance with their presence. If an overload occurs, a much larger plastic zone is formed (Figure 5.12) ‘After the overload a more extensive residual stress system is present than before the overload. This more extensive system, acting against the applied stress, ‘Fier 512, Resa sess before amd afer overload (a) Before (b) Af: (6 Situation afer ‘vetoed ut causes subsequent growth to be slower (retarded). Once the crack has grown, through the overload residual stress field ( overload plastic zone), the original ‘esidual stress field is restored (Figure 5.12) and the ‘normal’ growth is resumed, (In reality the compressive stresses do not extend throughout the plastic zone, because the compressive yield zone is smaller. However. the principle is still maintained). Another way of looking at this problem is illustrative. It was shown in ‘Chapter 1 that crack growth occurs by plastic deformation (slip). As a conse- uence, the crack tip plastic strain range is the best measure of crack growth. The stress and strain being singular at the erack tip, it is somewhat hard to cenvisage how they vary. However, the singularity occurs only in an infin ally small material element, while it will effectively disappear dee to crack tip ‘blunting. Therefore. crack growth phenomena can be discussed on the basis of finite values of stress and strain, their values being bounded by the (cyclic) stress-strain curve First, consider a crack with no previous history (no previous plastic zones) “with a very small plastic zone, cyclically loaded at R = 0 (Figure 5.13). At ‘maximum load, the stress and strain are defined by point A. Assuming an extremely small plastic zone in a large plastic plate, the elastic deformations of the plate will completely dominate the problem. Hence, all deformations wil ‘come back to almost zero upon load release: the elastic plate will squeeze the ‘permanently deformed material in the (extremely small) plastic zone al zero load. This implies thatthe erack tip strain will be reduced to almost zero (point in Figure 5.13). " Prue 518 Ses142 Upon reloading, the stress and strain must conform to point A, because that is the condition dictated by the surrounding. elastic material. Obviously, the crack tip is subjected to a plastic strain canye A., with R= O and R, close to <1. The material experiences a stress strain cveie 1194 The crack tip strain at R = 0 will return to almost zer0, but not exactly so, because there must be °juilibrium of residual stresses as illustrated in Figsze 5.14, The compressive stresses in the plastic zone have to be equilibrated by tension stresses around the plastic zone, which give cause to a small remaining. positive strain. Note however that the extent of residual tension stresses is actually less than shown in Figure 5.12. This ie duc tthe fact that upon return, from point A in Figure 5.13 yielding begins earlier, boeause AE = BD in Figure 5:13 due to the Bauschinger effect. ‘The return of the crack tip strain to almost zero is» result of the action of the large elastic plate, or rather of the remaining ligsrsct, Note that the « of the plastic zone is very small compared to the suiTness of the [Now consider a real fatigue crack with a pr wake along the crack edges i a strip of material representing the acc of all previous crack tp plastic ones through which crack has progressed. This ‘material is no longer loaded, but at one time it underwent plastic deformation ‘Closure stresses arise because the permanent elongation of the crack lips u Pare 5.1, Cenk tipsteses at R= Ooading top at ma faa Rott: atv kl Compare ‘ith Fite S12 13} Courtesy FMAS. 3 CLL AU HI tonding while ncountng for previous pete deformation closure of erck ip at ro lad 13}. Courtsy EMAS close the crack before the load is zero. As such, they ate similar to the com- pressive stresses built up in the plastic zone. As @ matter of fact, both stress systems resull from the same action of the surtounding clastic fed. Therefore, Figure 5.14 can be redrawn by including the crack lips as shown in Figure 5.15. The closure stresses at the crack tip never exceed the yield. This, it requires only an elastic strain to accommodate the remaining strain at the crack tp (in case of complete closure). which means that the remaining crack tip strain is clastic and the return point in Figure 5.13 is still B. Upon reapplication of the load, the crack will remain closed until the closure stresses and the compressive sresses are relaxed. But, the crack tip material is already straining. ts stress- strain condition moves from B to C in Figure 5.13 while the crack is still closed. Figure 5.16 shows the consequences of an overload. Depending upon the ‘minimum stress in the cycte and upon the relative stiffness of previously plastic ‘material with respect to the elastic material, the remaining strain after the overload will be larger or smaller (point F in Figure 5.16a). Therefore, depending upon return to F, or F;, the straining during the subsequent cycles will be as in Figure 5.16b. In any case the eyclic strain range (i. the opening. of the ae loop) is considerably reduced and the crack growth rate will decrease accordingly (retardation According to some retardation models (see Section 5.7) the growth rate fexluction is proportional to the ratio between the overload plastie zone and theewe 516. Consequences of eneshond for eack tip patie sin loop. Top: during evel ottom: nfl eve. [Copyright ASTM. Repent with permssion current plastic zone. Although the retardation models may be somewhat they do contain the relevant parameters. With increasing crack size {and plastic zone size) it becomes increasingly dificult forthe elastic material to ‘estore the zero strain field after unloading (the stiffness ratio between ligament and plastic zone decreases). Hence, at small crack sizes the return point may be F; (Figure 5.16), but at large crack sizes, it shifts to F, so that retardation becomes more pronounced (see also Figure 5.10). This may explain why various investigators find different retardation effects in the is ernck size dependent and panel sir (i The above discussion explains why a ct overload will reduce the retardation, but cannot completely annihilate it essive stress following the 4s (compare C and B in Figure 5.10); during compression the closed crack is no stress raiser and therefore all compressive stresses and strains ae elastic (elastic ‘rains are negligible as compared to plastic strains), 5.7. Retardation models ‘The complexities of the retardation phenomenon so far have precluded the development of an all-encompassing mathematical-physical treatment of the problem in a retardation model. More than half a dozen models have been ‘proposed [1-7], none of which covers all aspects of the problem. ‘A number of models are based on crack closure, As ws previous section, crack tip closure occurs even in constant ampli Some closure models consider only that part of the cycle effect ‘rack is open. However, even during that part of the cycle the crack tip material is staining (B-C in Figure 5.13) and this straining is just as much part of the crack-tip strain loop us is the part associated with the open crack. Hence, considering only that part of the cycle during which the crack is open, is physically incorrect. Certainly, closure changes afer an overload and this affects the general residual stress field at the crack tip, but this residual sires field is changed more due tothe larger plastic zone. In view of this, pure closure models, ddo not cover all aspects of the problem. There is no doubt that overloads affect closure; hence there will be a correlation between closure end retardation. But fetardation and changes in closure are both consequences of the overload. ‘Closure changes are not the cause of retardation. Both are symptoms, and one symptom cannot explain another; both are caused by the ‘disease’. A proper treatment of the model must consider the effect of overloads upon the total residual stress field in the wake of the crack and ahead of the crack, and in particular upon the plastic strain range. Other retardation models attempt to account for the residual stress eld directly by superposing it on the stress field due to the external load. The residual stresses themselves cause a certain stress intensity which can be added to the stress intensity due to the applied loads. Such models have several 2.5K2,/F2, in which 2.5 is a somewhat arbitrary number (Chapter 3). ‘Normal’ stress cycles at low Kau may give plane strain, but an overload may cause plane stress due to its larget Kua. This also causes a more extensive residual stress field and more retardation. Thus, in ‘accounting for retardation the computer code should assess the state of stress imeach cycle. It does not, even the ‘so-called’ sophisticated retardation model ‘may give large errors. For example, the larger retardation at longer eracks in Figure 5.10 may be caused by a change from plane strain to plane siress during, the overionds. But even if the slate of stress is evaluated, the value of a will be arbitrary, the factor of 2.5 is arbitrary and F,, is more or ess arbitrary. Some ‘models use different plastic zone formulations than above, but these still contain arbitrary numbers {mn the following crack growth analysis for variable amplitude loading will be illustrated on the basis of the Wheeler model. This model is used here not hhecause itis believed to be better than any other, but because itis very simple, ‘so that itean be used easily for illustrations. Itis worth mentioning however that if all models are simplifications anyway, the simplest certainly is the most appealing. If all models must be calibrated for general use, even the simplest model can be made to work by calibration Wheeler introduces a retardation parameter #,. 1s based onthe ratio of the current plastic zone size and the size of the plastic enclave formed by an overload (Figure 5.128). An overload occursing ata crack, size wil cause « rack tip plastic zone of size tm = Ki, = Babee mm aah ~ Eee, (520) where oi the overload stress. When the rack has propagated further to a length a, the current plastic zone size wil be Fem, (521) Where oq is the maximum stress in the ith eycle. This plastic zone is still ‘embedded in the plastic enclave of the overload: the latter proceeds over a distance gin front ofthe current crack (Figure 5.17), Wheeler assumes that the retardation factor y will be a power function of rjje. Since @ = 4 + Fre ~ 2, the assumption amounts to: da we (AK, R) | » wit . (52) t= (Stee s)148 ‘Fite 5.17 The mode of Whee. (2) Ststion immediately afer oven (b) Afer some cack ‘wth (6) Stason ater secon meron Ire, = @, the crack has grown through the overload plastic zone, and the retardation factor becomes ¢, = 1 by definition. The exponent in Equation (5.22) has to be determined empirically. This is the adjustable calibration factor. Note that ify = 0 there will be no retardation at all under any circumstances. Hence, the minimum value of ys zero. Typically for variable amplitude loading 0 < 7 < 2, depending upon the material but also upon the spectrum, For the case ofa single overload in a constant amplitude tes the retardation Factor gradually decreases to unity while the crack progresses through the plastic enclave (Figure 15.176). If second high load occurs. producing a plastic zone extending beyond the border of the existing plastic enclave, the houndary of this, new plastic zone will have to be used in the equation (Figure 5.17c), and the instantaneous crack length will then become the new dy Calibration of the above model (and all other models) proceeds as follows. A test is performed under variable amplitude loading. The test result is then “re-predicted several times using the proper da/dN ~ AK data and the proper {but with different valves of the adjustable parameters; (in the caze of Wheeler y-values taken are eg. 0,05, 1, 1.5 etc). The parameter value(s) that producets) the best coverage of the test data, is (are) the values to be used in analysis. An We ‘example of such a calibration [8] is shown in Figure 5.18. Clearly, in this case 7 = I.4is the parameter value to be used. Unfortunately, the parameter calibration is not general. It depends upon the lond-history and spectrum (Chapters 6 and 7). A different spectrum with a different mixture of high and low loads requites a different calibration factor Eg. the non-linear man-induced exceedance diagram (Chapter 6) requires different calibration parameter(s) than a nature-induced log-linear exceedance diagram; the calibration parameters are suitable for one type of spectrum, but they cannot be generalized forall spectra. Failure to perform thisre-calibrat and subsequent general use of calibration factors isthe main cause of claims that one retardation model is better than another. If proper calibration i performed for each spectrum type, any model can be as good as any other. Calibration for a certain spectrum type and material generally gives good results [8] for all variations of the same type of spectrum (Chapter 6), as shown in Figure 5.19. This figure shows results of about 70 predictions for random loading with the spectrum and calibration as used in Figure 5.18. More informa- tion on model calibration can be found in Chapter 7. Crack gronth analysis for variable amplitude loading Most siructures are subjected to variable amplitude loading, i. of the type shown in Figure 1.2 (For a detailed discussion of load histories see Chapter 6). {In such cases the crack growth rate da/AWV varies from cycle to cycle, depending. ‘upon AK and & of the cycle involved, and upon retardation (any cycle can be ice sition londing 18Co Y S Sc“ SSS \ i, \ DBDRMA' \ SSG \ is ‘Five 518. Accuracy of 70 predictions as compared to et [8 Courtesy Engineeing Fracture Meche, an overload eycle with respect to the subsequent cycle). Therefore, the computer rogram must evaluate and add crack growth on a cycle-by-cycle basis. For ‘example if erack growth takes 200000 cycles, the computer must perform the ‘operation’ 200000 times. This may take considerable computer time. Typically, ‘a mainframe computer can perform at about 0000 cycles @ minute, o that the above computation would take about four minutes. A personal computer might take as much as two to three hours for the same job (1987). ‘A logic diagram for the integration is shown in Figure 5.20. Thi based on the Wheeler model because of the latter's simplicity, vs tially different for other retardation models. Naturally, in order to perform the calculation for a structural crack, the computer must be provided with applicable da/4N data for the matcrial at hand (Chapter 7), and last but not least, the stress history for the structure (Chapter 6) ‘A more detailed representation of the algorithm involved is shown in Table Ist Frqre 520. Logi diagram for crackgrowth computation 5.2. A hand-cateulation for a few successive cycles is shown in Table $.3. These {bles show the basic algorithm which is quite simple; a useful computer code with many options, especially for the complicated book keeping for stress histories, is rather involved and will contain approximately 3000 statements (lines). ‘The accuracy of the computation depends somewhat on the retardation ‘model, but useful results can be obtained with any well-calibrated model. Most influential to the accuracy are the input of the stress sequence (Chapter 6) and material data (Chapter 7). A general discussion of accuracy and errors is ‘resented in Chapter 12. The accuracy problem involved in the simple algorithm of Table 5.2 is in the addition of a very small da to a relatively large a. For example, &-bit personal computers evaluate Ia =1 inch and da in a given cycle is 0.000001 inch, the new erack size will become132 Table 5.2. Crack promth analyst in vatiable ampliude Subroutine oF prepracenor:Librty or methods of Chapter Sebrowtne: generation of tes stony (Chale 6, anon iain, counting necessary (Chaper 6) Sebctine sate of aes fo y Subvoatine vations etundation mot Only Whece shown bn thy Subroatne oda Ubeary Newel i, < 4, then return to He 2 a tyors. voyages, gh) — —Subromtne comers up Sobroatines Prats Sebroutinee End ‘error. This means that @ and da must be evaluated in double precision wl ‘an 8-bit personal computer provides 16 digits. Hencea = tandda = 1E ~ 15 Table 4. Example of rearded crack growth by hand cleulation: onthe basis of Table [Note af astumed equal oI throughout bx = 3 (lan strain assur ec tas) few 3 for et line (dod) Retarded 0 so dan J nS na -i 153154 will be evaluated correctly, but ifda = 16 ~ I6and.a = |, the computer will still not recognize growth, There is nothing that can be done about this rounding problem. Usually itis not serious, but it may become a problem in evaluating retardation. Computer programs working in single precision may be one cause of claims regarding the accuracy of retardation models. should be noted, that the above does not chinge when « is evafuated in meters. For example if a = 0.01m and da = 1-E16m, the addition will be performed properly in double precision because leading zeros do not count. Mainframe computers already carry 16 devin: n single precision; they carry 32 in double precision. Even in that case rounding errors may occur, but they are even less significant, Although the algorithm in Table 5.2 is simple, computer codes rather complicated [e.g. 9}, because there must be (2) preprocessors for for a fi library, (©) options for various rate equations, da/dW table and/or library of dat (6) options for various retardation models; (4) accounting for state of st {€) options for random loadi (1) accounting procedures for stress history; (@) options for cycle counting If all of the above are included, the main coe of 3000 stateinents as ‘mentioned above can easily triple or quadruple in size, OF the above (e) (f) and (2) may be the most important and most involved: they are discussed separately in Chapter 6. Further discussions of the subject are found in Chapter 7(data and calibration) and Chapter 12 (errors and accuracy). is well-known that fatigue predictions, in general, have a low accuracy. In the ease of crack propagation, w linear integration (without interaction effects ‘or retardation) in Figure 5.10 negative loads reduce the retardation caused by positive toads, but the net effect is usually a deceleration of crack growth, so that retardation ‘models must be used Figure 5.21 shows results of crack growth in rail steel (10] under simulated {rain-by-train (Chapter 6) loading. Retardation hardly plays a roe in rail steels, therefore predictions were made by means of linear integration. The gure shows that they are within a factor 2 of the experimental data Better accuracy can be obtained in general, provided the retardation model is adjusted. Predicted crack growth for a titanium alloy subjected to aircraft service loading [8] is shown in Figure 5.22 together with experimental dala, ‘Generally, part ofthe discrepancy between computation ind test may be caused by scatter in crack growth properties. Most retar! odclscan be empirie- ally adjusted. In this respect, the Wheeler model is stractive, because it con only one adjustable constant. ‘eenerally 155 ‘Fare $21. Predictions and test data for sevice slaion luding inal et {10} nelgwee — i, 4 ie ‘ren 522. Pred crack gromih and test data fr sterat specirom [Coates Engineering Fracture Mech. Apparently, a crack growth prediction can be substantially more accurate than a fatigue life prediction. Admittedly, afew experiments, 1 spectrum of certain shape to empirically adjust the retardation model Parameters. From then on, predictions can be made for the same general spectrum shape and variations thereof, for structural parts subjected to lower156 and higher stresses and for cracks of different types (i properties of most materials show considerable scatter, The the subject of Figure 5.21 showed variations of almost a factor of ten in constant amplitude erack growth (see Chapter 14). Therefore, discrepancies between predicted and experimental crack growth are not a shortcoming of the predictive method per se, but are due to anomalies in material behaviour. Fortunately, most materials ate well-behaved in comparison, by sho. scatter in crack growth, Nevertheless, there is enough scatter that predictions ‘ill always have some uncertainty. This would still be the case if better retarda- tion models were developed. ‘However, the prediction procedure in general contains many more uncertai ties, which may be just as detrimental tothe final results as ae the shortcomings of the retardation model. These are: {(@) Uncertainty in the local stress level. (b) Uncertainty in the stress intensity cate (€) Insufficient knowledge of the load spectrum. (4) Possible environmental effect. Consider first the uncertainty in stress level and stress intensity. Inthe case of ‘a complex structure consisting of many elements. an error of five percent in the ‘stress analysis would be quite normal. The subsequent determination of the stress intensity can easily add another five percent, especially in the case of ‘corner cracks or surface flaws. Thus, the final inaccuracy of the stress intensity may be in the order of 10%. If the crack growth rate is roughly proportional to the fourth power of AK, the error in the crack growth prediction will be on the order of (1.1)! = 1.45 (85%); see also Chapter 12. Despite extensive load measurements, the prediction of the foad spectrum is sill an uncertain projection inthe future. Slight misjudgements of the spectrum can have a large effect on crack growth, Even if possible environmental effects are disregarded, the errors in crack {growth prediction due to uncertainties in stress analysis and loads analysis can bbe just as large or larger than the errors due to the crack growth integration. Development of better crack growth integration techniques will not improve this situation. Therefore, the shortcomings ofthe retardation models an hardly be used as an argument against crack growth predictions, “Faking into account all errors that can enter throughout the analysis, it is obvious that a safety factor should be used. This safety factor should not be taken on loads of stresses or da/dN data, Doing this would make some predictions more conservative than others. The complexity of rack growth behaviour dacs not permit an easy assessment of the degree of conservatism atiained through the application of such safety factors. A safety factor should rather be applied to the final result, ie. to the erack growth curve, by dividing 157 the number of cycles to any given crack size by a constant factor. The problem of accuracy and sources of error is discussed further in Chapter 12. ‘5.9. Parameters affecting fatigue crack growth rates ‘When predictions of erack propagation have to be made, data should be service. Such data may be hard affecied by an endless number of parameters, and the circumstances during the {est will seldom be the same as in service. The influence of the environment is the most conspicuous. ‘Theeffect of environment on crack growth rates has been the subject of many investigations on a variety of materials; the rate of fatigue crack propagation in wet air may be an order of magnitude higher than in vacuum, the effect being attributed to water vapour. The influence of salt water (seawater is of particular interest to marine structures. An example of is effect will be shown in Chapter 7. 1s generally accounted for in the crack growth analysis by submitting the ‘computer program with the actual data in tabular form. Its then assumed that luring each cycle in a variable amplitude sequence the rate will immediately adjust to the one found in the constant amplitude test data atthe same AK. This, is assuming that chemical/load equilibrium will be immediately attained. More claborate accounting can be implemented however. No single model can explain the influence of the environment on the rate of propagation of fatigue cracks. Different explanations apply to diferent materials, The effect iscertainly a result ‘ofcorrosiveaction and assuch its time-dependent. Therefore, heenvironmental fect is dependent upon the cyeting frequency. ‘Among the many factors that affect crack propagation, the following should be taken into consideration for erack growth predictions: (a) thickness; (b) type of product; {c) heat treatment: (a) cold deformation; {e) temperature; (1) manufacturer; (g) batch-to-batch variation; (hy environment and frequency. FFor the factors lower in this list itis less likely that they can be properly ‘accounted for. No attempt will be made to illustrate the effects ofall these actors with data, because some have greatly different effects on different materials. ‘Many of these ellects cannot be accounted for property in the ar structural cracks, primarily because the data are simply not av pragmatic approach to solve the problem is discussed in Chapter 7. At this point138 it is sufficient to note that the necessary use of estimated data may be of ‘considerable influence on the accuracy of the analysis. With this in mind, the ‘acclaimed inaccuracies of eg. retardation models may well become secondary (see also Chapter 12), In sheets there is a systematic, effect of thickness on crack propaga especially before the fracture mode transition. Fatigue cracks in sheets perpendicular to the sheet surface. When the :tack grows the size of the pla zone increases and plane stress develops. this c1nses the fatigue erack to change to single or double shear, as depicted in Figure 5.23, Plane stress develops when the size ofthe plastic zone is in the order of the sheet thickness (Chapter 3). In thicker sheets the transition will require a large plastic zone and occur at a greater length of crack. The data suggest that crack growth is slower in plane stress than in plane strain at the same stress inomity Although the effect of thickness on crack grow has been recognized for over 20 years, little effort has been expended in developing # useful model for everyday damage tolerance analyses, Figure 5.24 enphasizes the necessity t0 include this effect in the crack growth anelysis, A tative semi-empirical model hhas been proposed (11), bul the best way to account tor thickness effects is probably (o submit the proper data to the computer code. A factor of two error {due to thickness) in da/dV data may overshadow any effects of ‘inaccurate’ retardation models. iy investigators hold that there isa threshold for fatigue cracking: below 8 certain AX the rate da/d is supposed to be estentally zer0, At least one conference resulting in a two-volume book [12] was devoted to this subject The threshold would be reflected in a vertical dajdN ~ AK curve at low AR, as shown in Figure 5.25. The threshold is usually determined by gr decreasing the stress in a test until crack growth comes to a halt In possible retardation this procedure is subject to some doubt. Besides, the threshold is definitly crack size dependent; it is not unique. However, if one accepts the presence of a threshold, the practical question is “what is the effect (on predicted crack growth”. Figure 5.26 shows what it may amount (0: the two curves are indistinguishable. Generally speaking. the effect is hardly worthwhile ‘considering, but of course, in each case one would have to se judgement. If for ‘example the initial AX is below threshold 10 wegesat all Ta this eespect Figure 5.26 is somewhat deceiving; one cosid where the eflect 1s larger. However in random loading many cycles will be above threshold and the effect on life (in years or hours) is small, especially when there is retardation, Most computer codes provide an option to use a threshold, Referring back to Section 5.8 it should be noted that the computer ues a threchald automatically because it rounds any da smaller than a certain value, depending upon the current crack size. ‘Thereis often much concern regarding the anomalous behaviour of very small 139 Figure $28. The tamsion ‘heat (lop) a singe thst (sera ttm) in Alloy specimens: (8) Single shea (A) and double sent (8). cracks [13]. As shown in Figure 5.27 small cracks tend to show growth rates ‘much higher than would be expected on the basis of the acting AK, Various explanations have been put forward but.the ‘cure’ prupused is mainly artielal use of an apparent crack size a + 2 where 2 is a fixed quantity determined empirically Most ofthe short crack data stem from strain control fatigue tests at R30; t crack size (mm) 20 Sa: 6St9/mn? 30 a0 NCOP eyetes) Figure 5.26, Els f shel thickness on crack growth, Fiare $28, Teshold in aes (tematic fon small notched coupons (usually with central holes). Consider a short crack at such a hole asin Figure 5.2Ra and compare it with a Yong crack at the same ‘AK (Figure 5.286). By the nature ofthe test, the plastic one at the hole is much larger than the crack tip plastic zone. Since completely reversed plastic strain is enforced in most of these fatigue tests, the crack tip will also be subjected to completely reversed strain (R, = —1). A larger crack at R = — will close ia 7 sar o ® Brom According omeB ofD to Diswibuied New Remains ‘ecur Our ‘Remainder fr ‘Orsi in 36 ypc Onur inc Tee exons dagnm Fig te Tope a ype ig 620 189 & 3 1s +368 51 ‘Same for negative ve Wappitabie, Periods JA > 1B + 360 + AD = 100 tu190 toleae the remaining cles in column 1. Period Ci consreted n the sme tmanne. There can be 36 periods Cand then the re of leet are eave ‘One could go amin his manne, but sae sh: now ae only 4 pesos ete tue the remaining ees smn 15'by 49 onder to UE ape ain ectpe ii enon nin ating Arie fre some cycles unaccounted fo ad avs co tan ep were est Shown in colin 1 Then are cee of tore mnie combating Wile to crack growth, and sine the degram i oly aisien average and he procedure an approsimate one © this ile Sacrepaney could be el at However ifone would want abe pei they col fe scountd for by a ie Change i the conten of period Cah sa 120 By mean ofthe manpuation a column 16 te excrdance diagram of ype Dis determined es shown in column 21. Were the exceedanc diagram for D determined in accordance with the above principe would be shown in Firure 6 0 and the excetances would be as tn column 22. The flere Cetanly somewhat ferent rom tenn colunn 2 ut since they corre only the lower sets the elect on crack gromtr aap wl Be ata thatthe procedures bed upon stitial average) se dileent) types of paid canbe goneatd. However thee no need ogo tatemes along asa senisandom hor isoblied recognizing tat prods ov ileent severety do occur and that the higher load are clunered in the feviods No mate how reed the predaree acua! ad sequence i frac wl be diferent anyway. No mater how many lee re wed and ow tran) diferent types of pes are generated. there comes point where he temalting cytes most be lumped ine the remaining periods Thus the milder Pett wl never be entirely vepresenave but sn ns concer ony he Iowest ses no get harm done ‘Cal the otal numberof ples othe ses Wisory a5 gneated above les than the total number of exeredancer el the sey level. Thi i totaly acepabe, because the ver) sale eye cen wl ave noe Sneath growth, the procedure beng for erack growth wih tardaton The ttre the numberof eels wes the more mal] eee here wil be Bt shown in Figute 6.16 the esl ofthe analy wl be these (ce als seton Son aneton). in ccordance withthe nature of the loading there are only thee severe gerinds A and 13 prods Bin the tao 100'n Table 62. The marty consts I mil periods D9) and C(3), Regardless oft number feels hen and the number of periods concerned the ave procedure wil ret ths ay vara lhe procedures a Ucn, tte ahve incertae tal Oe ad any 1 pee: Bese rn Teil cnk growth analy the vere ered sold eat in random oernd the yes wn each peri ssl he applied randomly. Th te eee ee ee Frere 6.21. Cipping second occurrence of e.g. period C would apply the stresses in a different Sequence than the frst occurence, but the total eycle content of period C would be the same. Ifthe "basket" with 100 periods is empty, itis ‘refilled, and the Process starts anew; yet because of the randomization the periods appear in different order. Pra .22. Elke of clipping om ie: text dat 8192 ‘The above procedure can be implemented by hand calculations, and by submitting a resulting table of stresses and occurrences for use in a computer alysis, Aternatively, it can be programmed and included in the computer code. If this is done. the input is very simple: only a few data describing the exceedance diagram need be submitted; the computer does the res. I there are deterministic loads these must be interspersed in the above sequence. For example, in case of aircraft there will be taxing cycles and ‘ground-air-ground cycles after each period (fight). They should not be applied randomly among the other cycles. 6.6. Clipping “The spectrum is a statistical average of previous load experiences. A level j may appear to occur e.g. 10000 times. It would cause no surprise if it would actually ‘occur 9900 oF 10 100 times in a future case. There will be a certain level in the sires history that occurs only once. Since this is an extreme statistical number, i is very well possible that the level is never reached in service. For example, it would be quite natural iflevel 6in Figure 6.20 and Table 6,2 were never reached. Two high loads might still occur, but only go as high as level 5. In that case there would be twenty-one cycles (0 level 5 instead of eighteen to level $ plus three to level 6. The total number of cycles would not change, but the level 6 cycles would be ‘clipped’ to level 5. Also the exceedance diagram would be clipped as shown in Figure 6.21 Since the highést stresses cause most of the retardation, clipping can cause ‘dramatic effects on crack growth (f}. In the older literature, clipping is often called “‘euncation’, but presently the word truncation is reserved for the procedure discussed in the following section. Note that clipping merely reduces the magnitude of the highest loads to the clipping level: no loads are omitted, as shown 6.216, insert The first tests on the effect of clipping were performed by Schijve [8] using aircraft fight simulation loading. Figure 6.22 presents a summary of his test results, Various clipping levels were used by reducing the size of the largest ‘cycles-which are small in number-to the sie of the next highest level (no eycles omitted). In tests with lower clipping levels all stresses above the clipping level were reduced in magnitude to the clipping levels. The exceedance diagram in Figure 6.22 shows how many cycles would be aflected (up to about 800 at the very lowest clipping level, out of a total of over 300000). According to 6.22 clipping of the (wo highest levels (affecting only 80 of the 200.000 cycles), already reduced the erack propagation lives by almost a factor of about 2. It appears that crack propagation may be faster than expected if the structure encounters less severe service loading than was foreseen, The effect of clipping. ‘on erack growth can be demonstrated also in crack growth analysis as shown 193 Figure 6.2. Ee of clipping. Courtesy Eng. Fact Meck ie 6.23. These are computational results [9] for four diferent spectra ipping has an effect only if thee is retardation; in an analysis without retardation the differences would not be noticeable. ‘As discussed above. it should be expected that clipping occursin service. The ‘exceednce diagram is a statistical average, and loads that are occurring only a few times might reach toa slightly lower level only. Should this occur then crack growth would be faster; ifthe analysis did not account for cliping the erack growth in service might be much faster than predicted. It is sometimes argued that clipping is unrealistic and that all those load levels should be included that ‘may be anticipated to occur in service. The latter part ofthe argument is crucial; i they indeed occut. they should be included. However, whether they will occur is questionable. The spectrum is only a conjecture or, at best, an interpretation ‘of measured data. Slight variations ofthe spectrum may be unimportant in the lower part, but they are very significant in the upper pa194 ‘Piere 6.24. Trueation. (a) Complete pet: (h) improper truncation omision of wer level (6) Proper truncation a evel 8. Clipping of the spectrum should be a factor for serious consideration in crack His easy to produce ‘optimistic crack growth curves if high enough -re-reave included (Figures 622 ‘and 6.23), but the objective of the analysis i¢ 4 obtain realistic information. Engineering judgements the only guideline forthe section of an appropriate ing level. A reasonable level might be the one that is exceeded ten times. ‘AtBough tis felis often seks ios ary. Depending upon the clipping level, and upon the retardation propectics of the material and the spectrum shape, the life may vary by a factor of two or three as shown in Figures, 6.22 and 6.23) so that a categorical selection of ten exceedances would sil be siving disputable results for some spectra and some materials 9s ‘The best solution to the problem is to perform multiple computer rns. A. well-designed computer code will have options for automatic clipping. Once the preliminary work is done, multiple runs do not require any further labor other than by the computer. Multiple runs using different clipping levels wil establish the sensitivity to clipping in any particular ease. Ifthe effect is small, no special problems arse, bu if it is large the upper and lower bound of the crack growth curve can be established, as well as an average. 6.7. Truncation takes as much computer time (and testing time) to deal with one small eycle as with one large cycle. Thus the small cycles a the lower end of the exceedance diagram consume most of the time (cost) while their effect may be very small ‘Therefore. it would be advantageous if their number could be reduced. This is called truncation, In essence ine spectrum approximation by discrete levels already causes ication of lower stressess. Figure 6.24a shows truncation from 100000 exceedances to 80000 due to the selection of the lower level. Further truncation could be achieved by raising the lower level to level 5. This is sometimes understood to mean elimination ofall levels 6 as shown in Fighre 6.24b, In that case 70.000 cycles would be simply thrown out without any account of their effect. This is an improper procedure. ‘True truncation involves reconstruction of the lower step as shown in 6.24. This is in accordance with the stepwise approximation of the idagram which is known to be legitimate. In this way the 70000 cycles of level 6 are replaced by an additional 10000 cycles of level 5, and as such they are accounted for in a manner consistent with the entite procedure. Yet there is a savings of 60000 cycles. The total cycles ae reduced from 80000 to 20000 and Truncation requires judgement and it is recommended that the effect of truncation on analysis results be evaluated by making differen! computer runs todetermine whether truncation i permissible (giving the same esults asthe Full istory). The larger the number of stress levels the less the effet of truncation ‘crack growth. Hence, truncation is better justifiable if the user selects eg 12-16 levels instead of the smallest possible number of levels. Figure 6.25 and ‘Table 6.3 show that proper truncation is indeed permissible (conservative), improper truncation leading to unconservative results. (Note labels in Fig, 6.25 are reversed) IK truncation is understood to mean the elimination of small cycles. Then concerns about the effect of truncation are legitimate, and a sensitivity analysis s med. On the other hand, if done properly, tre truncation is196 Toble6 3. Sites levels incase of truncation (compare Figure 624 Reals poled in Figure 625). Unit of tess Ke Cae no trneation 1 eel Freeda ‘Ox Mina Ma a eee oer T v ~ 5000 3500 4 3 23500 20 w 6 2000 nom 2s 5 0500 20500 e " 1060 pen 158 95 25m ism a8 20 400 ‘com om” a 5500 is asi 1512 7.000 bon oo sm 30 ts00 ase 2; proper uaneaion: 9 vee accede ‘Osco Min Tan ae gee eels 1 1 5.000, 500 4 a 350 230 © 6 = 2000) 7200 x 5 osm 29300 3 a io 19.000 158 95 2500 17500 ™ wo tom 16000 om rh ss i650 a et rom how Case 3: imprope uuneation 9 eves ‘Occ 5 * 95 0 a 1512 ‘ioe below 17h med (compare eT 197 ‘aw .28. fet prope and improper ianction encontervatve) om predied crack growth, Waring abel or pop an improper unc ae ives in ose Fare perfectly legitimate as can be seen from Figure 6.25. A well designed com; ‘code includes provisions for proper truncation, the truncation beng performed automatically upon user specification of the truncation level. The issue is then of little importance. If the computer code understands truncation to be the climination of the lower levels multiple runs will always be necessary to prove that this "improper be Figure 6.26. Maniplation of Spectrum.Table 64. Stes bistony for Min sees Seedy see a 1667 10% 4 ies 600 wee Las <6 000 63096 467 000 2sii876 1000 1000 tv000000 Toni of aves 1004 isor for 100 years 1 uni of sess = 10k (an Min sir Occur ears——Venrs Yours Ven —“Veure oo 167 wo uaa 0 nz mo “100 eI aa sao 7 a no 167 oy 6.8. Manipulation of stress history ‘The stress history generated in the manner discussed in Section 6.5 may become a litle awkward if the number of exceedances is very small oF the number of Flights, time, voyages etc., covered by the diagram is very large. The best results 'eobiained when the total exceedances are on the order of 2000 to 100000, and the number of periods (voyages, months, ete.) an the order 50-1000. Therefore it may be advantageous to reduce exceedance diagrams for smaller of larger numbers to the above ranges. For the following discussion all exceedance diagrams are presumed to start at ‘one excetdance. If this requires extrapolation 1 0 tesheatte higher stresses, the spectrum should be clipped atthe desired level. Examples of how the exceedance diagram can be manipulated to obtain a realistic stress history follow below. Let the exceedance diagram A in Figure 6.26 be for 100 years of usage (1200 ‘months or 3200 weeks). The diagram starts nt 100 exceedances and ends a 100 000.000 exceedances. Extrapolation to one excerdance brings the highest siress level to 50, which cannot be justified. The problem can be rectified by 199 Table 63 Sess Nstory 1200 months Eneeedance table for 1200 months iia Min are ‘Sie ares Saw 1667 000 20 ar 1000 Mo 600 Wo 260 “08 1000 ik00, 0.00 1000 10.00 Toni of ses 100) istry for 1200 monthe {unit of sess = 100k Max. sti Min st Occur Monta Months Moniba Months Months oe i wo ° opeaeaa; T ‘oo 9 ° oo ° ‘oo % ° ° 2 s oo 16 ° 3 5 Py 369 ms ° 4 4 16 mt sé 3602 2 0-9 a so 3m as 37 ae Mo mass 65 aie ne a my Th sotto MR ano 33 tema 161567599 er Se TMssI0 634 i 73 om ba 1a seatsend 29587 331660 nao clipping the spectrum at 40, Using a stress history with different periods of years {e. years with different cycle content) leads to the results in Table 6.4. The period “year’ may be too long, and it may be more realistic to construct 1200 different months, or even 5200 different weeks. This would result in stress histories as shown in Tables 6.5 and 6.6, Obviously, the latter two stress histories Aare not realistic, because (o0 many cycles are concentrated in one period (month ‘or week), A much better result can be obtained by using 2 spectrum for 10 years (ie. 120 months), which is spectrum B in Figure 6.26, Note that B and A are sil the same spectrum: a factor often in exceedances is made up for by a ten times smaller time interval. As can be seen from Table 6.7 a more realistic stress history is obtained. If weeks are used as a period, a spectrum for one year (52 weeks) can be used, which is spectrum C in Figure 6.26. The resulting stress history is shown in Table 6.8. Note that the highest stress is now 40 without clipping Wis advisable to reduce the spectrum (or stress table) to one that covers 50-1000 periods in the above manner, in order to arrive at a realistic stress history. Allo the above may seem a litte artificial, bt it shouldbe realized that200 201 Table 64. Sue bishry for 5200 weeks Table 67. Sues history for 120 months ceedance able for 5200 weeks Exceedace lable for 120 months Man ae in ae Steady sess na Min are Steady ares Erceedancs Soon Too" T a nM To.00 t 20 1000 “ Ho 867 10.00 By 3400) 10.00 uses M00 00 1000 on 2600 1000 106 uw 066 00 1309 1809, 000, asters pw 3H to.00 398106 10.00 000, 00000000 om 1000 1000 0000000 Tonio wees 100808 itor for $200 wks 1 Max. str. instr, Ovcurr Weeks Weeks 419 oD Toni of wee LOO istry for 120 month unit of tes = LOD Max xr Min st Oscurr Months Month Monthy Monit Monihz se 4 ' t we 167 oo 7 5 oS ES aoa 3 ° r oo nua o 0 6 4 9 a ea a A 4 4 a oo m ° 0 6 6 % oa a : ° 5 3 ‘ 00 “100 ° > oo : 6 al fs 0 ; 3 : x67 w18 ms 8 od 5 ™ ‘ 2 u mt sé 2 4 7 6 on a os6 3602 er) ay a 5 ok ow x00 33 16905 4 s 4 38 ms i yo 8 kw 267 au Tass wm om ue a was a aa un 364138 % MSM nr 2 Ms oats 10d mo 31 v690264 52 mes iy a 209 ate amas 16? se 7Hassio Wir 97 mse ine er ios komen mo TR Moaisom 627 ast) Re AY arisane_wers 38591 amo nd ste the purpose of crack growth analysis is to make a projection in the future, so a to exercise fracture control (Chapter 11). As the future is not known in detail, any reasonable projection that accounts for the salient features of the load history is as good or better than any other. Furthermore, such seemingly trivial things as clipping may have a greater influence on the predictions than all other painstaking efforts for accuracy and realism, Clipping levels are often selected fon the basis of a mete assumption, yet flight stress histories. All ofthe latter's ‘sophisti “simple” assumption with regard to clipping. More often than not, the ‘sop Aicated” stress history is than applied randomly with (otal disregard foe the Fact that high loads are clustered as discussed extensively in this chapter. It should be clear from the foregoing discussions that there are only a few iswues that count in the generation of # stress history. ‘These are: (a) Different petiods of severity (flights, voyages. etc.) must be used and applied semi-randomly if in practice the loading is semi-random. Random application of stresses derived by complicated means will negate al complicated efforts to determine stress histories, (b) Deterministic loads must be applied at the point where they occur, For example g..g. cycles must occur between flights; random application may defy all other sophisticated procedures (c) A reasonable number of strexs levels (10-12 posi be selected. More levels will complicate the procedure usefulness of the results, (4) Largest positive and negative excursions must be combined, and s0 on. Random combinations will require subsequent counting, the result of which can be foreseen, while the stess history was based on an already counted history in the first place (€) The total number of periods and cycles must be in accordance with the exceedance diagram, ‘The above criteria account for what may be called the signature of the loading. Small changes in these including clipping, will usually have more effect and negative) must ithowt improving theTable 68, Sues history for 52 weoks sceedace lable for $7 weeks Max sees Win, wee Sith eres Faeredance ow “000 v Moo 600 6 0 200 aI nw 200 3981 1600 oo 209s 000 1090 1.00000 Tait of sre 00 History foe 52 mek ae Max st. Misi. Oseur Wes Wes Weeks Werks a x 2 r oo Toor 7 ° ° ° t ms “a 5 ° ® 1 2 60 cor 0 ® ' ' 2 as 50 a o ) 3 4 300 =33 m 2 3 3 ° ns “0 as 7 * . ” 20 0 126 2 n 3 a ns W 3es n » % 1a 200 My nip ons eo a ns 50 Mast ne ms om na 50 67 assests 4075 ns yas mosh ksh fon crack growth than any complex means of establishing stress levels Hypothesize for a moment that it would be necessary to use a complicated procedure to establish e.g. the loads or stresses in every segment of a flight {ascend, cruise, descend etc.) or of a vos:apr ofa ship (river, port, locks, cruise, ‘elc,). Any such procedure would be full of assumptions, and it needs no explanation that no airplane or ship would ever encounter the assumed circum. stances. If it were necessary to go through this “sophisticated’ procedure, it ‘would only be because the predicted crack growth would be different if other assumptions were made, But in that case the prediced crack growth would have ro relevance to the service behavior in the first vluce. Actual service loading, ‘would be different from the assumptions made, and if the results were that tical (0 the assumptions the predictions would be useless, regardless ofall sophisticated procedures used. The simple fact that one would consider the result uselull nevertheless, implies that one makes the assumption that 1 does fot really matter whether the loads are as derived from this sophisticated procedure. If so, why would the “sophisticated” procedure be necessary. Sur- 203 prizingly. if swch complicated procedures are used, the stresses often are applied in randox nce, while they should be semi-random, and clipping is ignored, Wis very cy to perform many computer runs with different stress levels, simple and sophisticated stress histories, et. Ifthe results do come out about the same (which they do), the above point is proven. One might ask whether tests would prove the same. Not enough test data are available to conclude that they do. But, afterall, crack growth prediction are made by analysis, and hence, ‘only the analysis results count. Ifthe computer cannot distinguish the difference, the results apply. Next hypothesize that there would be a difference, depending upon the sophistication of the stress history generation. Then of course, sso the behavior in service would be different, In other words: no matter how sophisticated the sess history generation, it would have no bearing upon the service behavior, and the effort a waste ‘The above ‘cussion shows that procedures 10 derive stress histories in a compin std way are defying themselves, If they were necessary, they would be tuseless by implication. If they are not necessary, they are useless from the start. must be concluded that only those effects count that could make a signifi ‘cant difference in the calculated crack growth. Those are the effects listed above concerning the signature ofthe loading and clipping, They have to be accounted for as was demonstrated by test and analysis results, All others are secondary: they complicate procedures without adding (o their usefulness. ‘A final problem in the generation of the stress history is inthe definition of sires © use nominal stress, local stress, or hot spot (welded structures) stirs Should secondary stresses, residual stresses and dead stresses bbe accounted for, and if yes, how? To a certain degree, this can be dealt with through the baseline data used in the analysis. However, there it more here than ‘meets the eye. the main problem being that local plasticity during the highest load cycles tends to make even the most sophisticated load-interaction models ‘g0 awry. This problem can be solved only by pragmatic engineering judgement. 6.39. Environmental effects there is load- or load-environment interaction itis especially important that clustering of high loads in storms is recognized by comparing analyses for Fully random and clustered loading (mild weather or storms). Computer runs with similar clustering, but with a totally different sequence, should provide essenti- ally the same result. For marine structures especially; the effect of environment should receive ample consideration. Quite obviously, the main environments (o consider are ‘sal air’, sea water, and the ‘splash zone’. The question arises: ifgoing from one cyche tu the nest (eg. high AK ~ low AK oF vice versa) in successive cycles, does208 the new rate immediately fallin place with the (constant amplitude) baseline data. (In other words: is there no environment interaction). Models for environment interaction have been proposed. As the environ- ‘mental effect is time dependent, these models bring in the element of time, oF the frequency ofthe loading, They are certainly of scientific inters, but the practical problem is one of load (retardation) and environment interaction. As the latter problem has not been addressed al all, the complication of accounting for cevironment equilibrium is hardly worthwhile for practical crack growth analysis. Thus, the models mentioned above remain in the realm of research and “theory” (hypothesisy and are not too relevant to practical applications. as discussed in this book at this time. A pragmatic approach calls for the ‘assumption that chemical equilibrium is indeed established immediately. In that ‘ase, the problem is solved by using crack growth rate data (da/dN-AK) for the relevant environment, Retardation effects are accounted for as discussed. Re- {ardation parameters must be determined empirically, and therefore, these will automatically account for any chemical interaction effets, included in the test data used for calibra 6.10. Standa spectra So-called ‘standard’ spectra have been developed in Europe for general use for 1 varity of structures. The word standard is to be interpreted to mean a general norm based upon a very great number of measurements: it is not meant to be ‘a design standard or specification. These standard spectra were intended primarily for use in tests, so that results of various experimental investigations ‘might be better comparible. There are standard spectra {12-19} for airplanes (Falsta and Twist, for helicopters (Helix and Felix), for offshore structures (Wash), ete. ‘Stress histories for these spectra were obtained by using algorithms very similar to those discussed in the previous sections using semi random loading. with periods of different severity. In essence, not the exceedance diagrams but the stress histories derived in this manner are considered to be the standard. A limitation of the standard spectra is that they essentially always perform in the ‘same way; this is useful for testing and data comparisons, but may be too severe a limitation for practical ws. Using the standard stress histories ina crack growth analysis, would necess- itate input ofall stress cycles inthe sequence generated, and thus require special pu facilites. Fortunately, the algorithm as discussed here (which is somewhat impler than the one used for the standard histories), can be rather easily incorporated in a computer code for crack growth analysis. In that case only @ ew data points describing the exceedance diagram would be needed as input, ‘upon which the analysis code would generate a load history. perform clipping. 205 1d truncation as prescribed, and subsequently the crack growthcalculation. In this manner, the standard spectra would be more useful and more easy (o use. ‘Their exceedance diagrams could be applied for crack growth analysis if a specific exceedance diagram were lacking. Most stresses are proportional 10 loads, so that the stress levels indifferent parts of the structure can be obtained through a multiplication factor. The stress levels are given as relative numbers; sion factor is sufficient to determine all stress levels. Herce, judiciously used. the standard exceedance diagrams may provide spectrum information for image tolerance analysis where none is available. Suitable crack growth analysis software only needs input of the exceedance diagram. 6.11, Exercises 1. Approximate the exceedance diagram of Figure 6.27 by 8 equally spaced positive and 8 negative stress levels for equal exceedances as the positive levels. Determine the number of occurrences ofeach level. Assuming that the 00% level represents stress of 15 ksi, determine the stress ranges for each level by combining positive and negative excursions of equal Frequency of Fine 627, Ener206 2. Repeat exercise I by constructing 6 levels starting with the exceedances: 2,10, 100, 1000, 10000, 100.000. IF the stresses of exercise 1 and 2 were used in a crack growth analysis, would there be a difference in results? Using the results of Exercise 1, generate a stress history with five ‘periods’ of | different severity. 4. Truncate the spectrum of Exercise 1 properly at the lowest level but one, Generate a new stress histgry with five different ‘periods’, Next use the improper truncation proced @Pliscussed in the text and compare the results (of the two procedures (Note that for the latter grt the resulls of Exercise | ‘can be used «hnectly; only for the frst part a new history must be generated.) How would clipping 1. ch highest level but one change the stress histories of exercise 47 6. Repeat Exercises 1, 3,4, 5 using the spectrum of Figure 6.27 and by selecting Tievels. Change the stress history developed in Exercise 3 for the case that the maximum level represents a stress of 21.5 ksi 8, Why is clipping not important if there i no retardation? References UT ME, Mayela et Febeany 1980, [2 Ineratona group on Crack Growth in Nucla, Sra HCCGRY, (9F. Schive, The amas of random loadtime histories eth elation ‘aleadavons. Fatigue of Aral Siractuex p15, Peron (1963), 14) 1B. de Jonge, The meniterng of frig lnk, ICAS compres, Rome (1970), Paper 0-3 15] G.M. van Dit Satis! fad deta roesing, IAF sympsien Miami (1971) 16) D. Brock ea, Fugue senth of ereav-akes ships, Battle Ret to Arm Bu. Shipping 97), 1D Private communication (8) 4 Schijve, Cummulative damage problems in nical structures and material, The arom J 24 (197 pp. 517-522 [P| D. Brock and S11 Smith, Fatigue crack growth prediction wider aircraft speci sing, Bg Pract Mech, IV (1979) pp 122-12 Ho] 1. Rroek and RC. Rie, Prediction a Fie crack g0= neo al, SAME a Shoo 9 (1977) pp 392-408 U1] NU Bolles, The chance of aro ld Royal Alcea Fat. TR 63039 (1965), 12] GM, van Di and Bde Sone, traction fights eu dng sndard fr oie ‘tolaton, FALSTAFF, NLR-Report MP TSO17 (197) 113) 1.8. de Jonge, D. Schuetz, Nowack, and Schyve, A stndard lad sequen fr fight sunudvion vesting, NLR= Report TR T29, or LBF-Report FB 10604 RAF Report TR naies 190). 114) 14. Gerhr id eg inerey evaluation, USNRC Report NUREGICR.1MS, Fatigue tee Wife Stondaraced emironmeniel fie sequence for the eration of compose iu icra (ENSTATH), Vat ar Te ox baenigho, Frnunlatey Tt fe 207 ran rb a (NSAP ab else rnb [Betriehfestigkeit FB-179 (1987), “ 151 GE tito Ba apache dele of TURDISTAN, ag and, a ir rh ene ik, ASTM conference Cincinati OH (1987), = fe 1M ks We Sch ser de of Cs ecm fr ge pin aie rita tar batcenigechah ABO Spe A 111 Marte Sema Be ene NOMRDSCED hit 101 PR Eden) Dn, Sn ddagger epee (EL and FELIN. Royal Aca Eaaishnem RAE TR BOLL Pat | tnd cpa oe ne hn oes AST ene el' CHAPTER 7 Data interpretation and use TA. Scope Material data are an essential input to all fracture and crack growth analysis: applicable data analysis is not possible. Misinterpre ‘of data are major contributors to the acclaimed shortcomings of fracture mechanics, because the data interpretation problem is not a trivial one, especially where it concerns fatigue crack growth, The phrase: “data are dat and cannot be argued with’, s commonly misapplied. The statement may be {we for the raw data, ie. a load-COD curve, or a measured crack growth curve, but cannot be applied to the derived data, K,, Ky. Jq. and dajdN ~ AK. The latter are obtained from the raw data through an interpretation process full of ‘assumptions such as the data reduction procedures stipulated in the relevant ‘ASTM specifications (1-5), Although these reduction procedures are probably the best available, they are not indisputable ‘This chapter is not intended to argue the shortcomings of data reduction Drocedutes. Rather, it is concerned with how these data are subsequently interpreted and used in the analysis. Questions arise regarding constraint. scatter, equation fitting, data errors and inaccuracies, retardation parameters, ‘mixed or changing environments, etc. These are the (ype of problems addressed here, because they may affect the accuracy of the dumage tolerance analysis, ‘more than the shortcomings of fracture mechanics Problems in the use of toughness data, both i 1 Kand J, will be briefly addressed first. As the toughness affects the caleulated permissible crack siz, a, inaccuracies change a, only. From the point of view of fracture contro this 1s ‘ofien not very important as will be discussed in Chapter 11. A small change in ay does not affect the life much, unless a, is already small ~ especially when im rare cases it ic below the detection limit, or where it affects arrest ant Teak before break (Chapter 9). Errors in rate data om the other hand, affect the life directly. A factor of two difference in rates (which is not uncommon) changes the life by a factor of two, which is significant, For this reason, the main 208 209 ‘emphasis in this chapter ison the interpretation and use (estimation sometimes) of fatigue rate data, In many cases the analysis must make use of data provided in handbooks [6] (oF the general literature [7], oF of data in magnetic files in software libraries, ‘These are almost always reduced data, as opposed to raw data: they have been pulated. Ofien some interpretation has been done as well, such as averaging or curve fitting. The user of such data should not overlook this ‘wiviality’, and if possible, check whether reduction and interpretation was appropriate for the application envisaged. In practically all eases a decision ‘must be made on how to deal with scatter, in particular for rate data, Data for F alloys or circumstances, ‘These kinds of problems are addressed in this chapter as well. Recipes cannot be given: only guidelines can be provided. In the end the user must exercise ‘engineering judgement, possibly honed by the follo ‘No attempt will be made to provide any material data here as this is not a ‘materials handbook. Considering that the DT-handbook [6] consists of ‘thousands of pages, any attempt to present data here would be inadequate and selective. Real data will be used for illustrative purposes; in some cases hypo- thetical data will be used to better demonstrate a specific point. 17.2. Plane strain fracture toughness ‘The plane strain fracture toughness, Kis commonly obtained from a standard {est [1] om compact tension (CT) specimen, Standardization can be defended ‘on many grounds (also the tensile test and the hardness test ate slandardized). Does this mean that data obtained from non-standard tests are inadmissable? this question had to be answered affirmatively, the result ofthe test would be ‘of no use in the first place. Its assumed that the test result, K,, can be used for the prediction of fracture ina structure, on the basis ofthe similitude argument that fracture takes place at the same value of K asin the laboratory test. Hence, ifa specimen other than the standard could not be used, the implied assumption would be that the other configuration does not fracture at K, ergo, it would be impossible to predict fracture in a structure on the basis of the Ky from a standard test, which would render the standard test useless as well The concept of fracture mechanics i that fracture in plane strain occurs when K = K, regardless ofthe configuration. Therefore, K, is obtainable from a test ‘on any configuration, provided one knows the expression for K. ar rather for £8. Similarly, fracture can be predicted for any configuration for which the expression for fis known. The main justification for the standard test is that f! for the CT specimen has been calculated to great accuracy, for siraight-front210 through-the-thickness cracks. The standard declares a test invalid ifthe fatigue crack front is not straight. This is because the P-expression provided isnot valid for such a crack, but fracture still took place at K = Kj, and the test resull could stil be used if f were known for the actual crack shape ‘The test also is declared invalid when the thickness is less than 2.5 (K, Fy ‘As was shown already in Chapter 3, the factor of 2.5 was deterii-t by ‘committee agreement on the basis of test data, but it appears from Figure 3.8 that a factor of 2 or 4 would be just as defendable. Yet, appears in the standard, which does not make tle number certainly not sanctimonious. The ‘candidate tovehoess’ obtained for a test ‘here the factor is 2.4is not by defination worse than the one obtained fora test where the factor is 2.7. As constraint depers upon the yield strength and K, the factor in realty will depend upon F,, and Ky, and upon «fguration (Le. itis, ‘material dependent and configuration dependent) It should be pointed out also, thatthe toughness obtained when the factor is eg. 2 may not be declared a valid plane strain trmghness by the standard, but itis the toughness for the thickness at fun. I wr the given thickness the number is at least as reliable as K;, obtained in a valid test. Thus, if one is not imterested in K;, per se, but in the toughness for the «hoven thickness the result is useful. If one insists on knowing the plane stcain toustoess the test must be ‘valid, but one may want to exert caution with regard ta the factor 2.5. AAs pointed out the required factor for plane strain is likely to be material dependent. Almost certainly, itis also geometry depenstent. This can be appreci ated from the data (6 for large center cracked pancls shown in Figure 7.1. The ae Pie 7.1. Toughness daa {6 fr 7078 Aluminum alloy (rg center cracked panels) au same reference [6] quotes Ki, = 30ksi/fn forthe same material, obtained from a standard test As the alloys yield strength s F, ~ 80 ksi the standard would Predict plane strain at a thickness of 2.5 (30/80)' = 0.35inch. Yet, the center cracked panels of 0.35inch thickness have a toughness much higher than 30ksi in, Similar discrepancies ate found for other materials. Constraint in CT. specimens is higher than in some other configurations. This is convenient because it permits the use of small CT specimens and tends to lad to conser- ‘ative toughness values. But clearly, ifthe factor 2.5 is applied to the center cracked panel, a toughness of only 30ksi/in would be counted on for a plate ‘0F 0.3Sinch thickness, while the actual toughness fas high as 50-60 ksiin, For 4 given crack size this would be underestimating the residual strength by almost 50%, certainly not a negligible error. Clearly, plane strain in center cracked panels has not been reached yet at a thickness equal to 2.5 (K,/F,,)’. At this thickness the toughness is higher Airansitional) Plane strain ours ata thickness of 4-5 times (Ky, Besides, there is another reason why the center-cracked panel behaves differently, namely the rate of change of K(a) as explained below. Every material except a truly beitl one, exhibits a rising R-curve (Chapter 7.2. Ths is usually not problem ina standard test on n, because Ga) and K(a) tise very sharply with crack size inthis specimen (G = K’/E; Chapter 3), so that the beginning of fracture usually coincides withthe instability. The standard prohibits excessive non-linearity of the load displacement dingram in order to exclude cases in which instability i not immediate {8}. In a center eracked panel Gla) or K(a) increase only ‘moderately with a. Hence, fracture is first stable, and instability ccursat higher values of K, especially for large eracks (Figure 7.2). Usually the question of ierest is how much load a structure can carry and at which sess it breaks. Thisis determined by the instability, and thus the K a instability isthe relevant number, called K, oF Kg (Se alo chapter 3). ‘Which cases warrant the use of a valid 7 Il is obvious from the above arguments that there is no categorie answer to this question. Engineering ” o ‘Figs 7.2 Difeest behavior of compact tension specimen and center Specimen (6) Centereacha2 judgement must be exercised unless conservatism is desirable, Cases where the use of K;, is always indicated concern surface Maws and corner cracks. It is ‘emphasized that in these cases the thickness plays no role (Chapter 3). For the analysis of part-through cracks refer to Chapter 9. Although possibly superfluous, itis pointed out again that even in plane strain, and regardless how low the toughness, the predicted residual strength for mall cracks is too high, the predicted permissible (or crtieal) crack size for high stresses too large. Fora —+ 0, the predicted stress tends to infinity. Using EPFM inthis regime (For an otherwise LEFM material) would not solve this problem cither, An approximate solution must be used. This is amply discussed in Chapters 3, 4 and 10. Finally, in using toughness data from handbooks, the relevant conditions (emperature and direction) should apply. Like most other material properties, toughness depends upon temperature. Engineers are used to the material Droperties being temperature dependent, but the (often) strong, dependence ‘upon direction of fracture and crack growth properties is sometimes not recognized. The going nomenclature for erack direction and material direction is shown in Figure 7.3. The first letter, L, T or $ indicates the direction of the Joding, the second letter (again L, T ot S) the direction of the crack. Toughness values for the ST-direction may be 30-60% lower than For the LT direction. Not accounting for such effects may lead to errors. As the problem is most often associated with the growth of partthrough cracks, the reader is referred to the ‘more detailed discussion of the matter in Section 7.6. 1s, toughness data for the temperature and direction of interest are ‘not available. Section 7.5 provides some guidelines for estimating the toughness in such cases, 7.3. Plane stress and transitional toughness, R-curve Data on plane-stress and transitional toughness should always be assessed with extreme caution. There are many ways in which raw data can be erroneously interpreted and derived data misquoted. There are equally many ways to misuse them. Many K, data were derived from panels of insullicient size, which renders them useless. As was shown in Chapter 3, when the toughness is high ~ whi it usually isin plane stress ~ small panels will fil by net section collapse, ie. at K < K,, Thus the K, derived from such a test is too low. It is sometimes quoted as an apparent toughness, Kir, but this practice is suggesting more than it should, Fracture due to collapse should be treated differently as discussed in Chapters 3, 4 and 10, and not on the basis of K, nor Ky. A test on a panel of insuficient size yields no useful result other than F,,, and the knowledge that higher than the ‘apparent Ky,'. A good data handbook {6} quotes the pane! 213 Fare 73. Nomenclature for lading and Crack Ditecton, size as well s the raw data so that the user can readily check whether collapse ‘or net section collapse (yield) occurred, and then make a judgerrent.Ifno panel sireis quoted the numbers are suspet. For example ifthe toughness provided is 80ksiJin fora G-inch wide center cracked panel of a material wth F, = 6Oksi, the number is welss. A 2-inch crack (a = 1 inch) in such a panel would provide oy = 80/1.07Yta = 42ksi. bout failure by met section yield is at ay = (6 ~ 2/6) x 60 = 40ks. Obviously, the tex pane failed at net section yield. The valve of his tothe user is only that it is certain that K, > 8Oksiv/in and that all center cracked panels smaller than 6-inch wil fil by collapse ‘Another problem arises because of the definition of toughsess. The value quoted may be the actual K, or the effective toughness, Kg, whichis sometimes called K,. As explained in Section 3.12 nether ofthese is actually a material Property: both depend upon crack size and configuration, because they are derived for fracture instability. tn principle, the actual K, is not avery useful number: if the amount of stable fracture is unknown, the ‘rack size itsll is inknovn; the residual strength cannot be calculated. Conversely the calculated “eritical crack size is always too large (Chapter 3) If some information on the amount of stable fracture is available Kg can be estimated from K,. Without such information, a sae assumption i not without probiems ether but it isthe most the Recurve. The criteria for pane size apply. Using the Rcurve and an instability analysis may seem to sole the dlficlties ‘ith Key but i practice it does not. First ofall, Recurve data ae searce. More important, accurate measurement of R-curves are dificult. As a consequence, the inaccuracy of the data usually negates the advantages of the more refined procedure. Furthermore, collapse considerations apply to R-curves aswell Tike the plane strain toughness. K, and Kg depend upon temperature Direc tional dependence is usually small (TL. versus LT), because plane stress and transitional toughness always apply to through-the-thickness cracks and these24 do seldom, if ever, occur in ST direction. The problem of estimating the discussed in Section 7.5 7.4. Toughness in terms of J and J Toughness data for use with EPFM are scarce. Available data pertain mostly {o materials used in nuclear power structures such us 304SS, A-533B elc., many ‘of which can be found in a series of reports iewued by the US. Nuclear Regulatory Commission (NRC - NURFG:} trata for other materials are scattered throughout the literature. Ne sy-temutic data compilation is as yet available. Practically all data have been obtained from CT specimens and in almost all cases through the J-estimation procedure preseribes:.n the test standard [2]. The fatter is based on the assumption of full net section yield, which by itself is subject to some doubt (Chapter 4). Although many estimation schemes for J have been devised, the only viable general procedure for the application of EPFM in engineering fracture analysis is through the wse of J = Ha" 'a/F, oF equivalent, as discussed in Chapter 4 (excluding finite elements for routine fracture assessments). Since H is known for CT sprcimiens, it would be better to ‘obtain J, and Jy data through the use of the abwye equ>ion rather than from doubtful approximation, especially i the data are to be used in this manner. Fortunately EPFM equations are so forgiving that the discrepancies seldom are ‘of great relevance in the application (see Chapter 4, Figure 4.11), This has the ‘dditional advantage that the user need not seriously worry ifthe data appear to show large scatter: the scatter is logical and rather immaterial in the fracture analysis (Chapter 4). ‘Apart from scarcity there remain two significant problems with the toughness data, The first and most important is tc constraint. All data are obtained from through the thickness cracks in relatively thin specimens, Often constraint is not ‘made an issue, but in most fess there will uot be plane strain, This does not ‘matter if the data are to be used for through the-thickness cracks in the same thickness, But most cracks are surface flaws or corner cracks where plane strain Drevals. In such cases the toughness J, and J willbe less than for the test specimens. One can only speculate on the magnitude of the difference, but it ‘would be reasonable to expect similar differences as in LEFM, Constiaint criteria for plane strain have been proposed {9}. bu (i comparison with the LLEFM plane strain conditions) these seem to he 100 optimistic (favoring plane strain in small thickness, as discussed in Chapter *, Another problem is that the J_ dats ave witaird from small specimens and thence. usually are only for small values of Aa (generally for Aa up to0.2inches at best). In large structures the amount of stable fracture will be considerably ‘more than 0.2inches (Figure 7.2 applies just as well to the Jy-curve). Analysis 2s then can be performed only if the Jy curve is extrapolated. In view of the fact ‘that the analysis is very forgiving, such extrapolations can be made rather easily without sacrificing much accuracy. Nevertheless, this is an unfortunate situation which should be corrected through a thorough revision of the standard test procedure with regard to specimen size requirements, 7. Estimates of toughness ‘As shown in the foregoing sections, estimates of toughness are often necessary; even if data are available they may not be for exacily the circumstances Prevailing in the structure. Considerations upon which such estimates might be based will be given, bu the actual estimate remains the responsibilty of the user. Wino data at all are available, the problem of estimating a reasonable toughness value, Jy or R-curve becomes more difficult; the success ofthe estimate depends ‘upon the availablity of comparable data. The procedures discussed below lead to ROUGH estimates: they are not recommended for general use. When there is a lack of data, tests are always preferable. Consider a structure built of alloy Y, and suppose that a handbook provides data for alloy X for LT and TL direction, and for alloy ¥ for LT, while the value for the TL direction is sought. IT it can be ascertained that alloy ¥is very similar to ¥ and that yield strength, tensile strength, grain size and production process are similar. the estimate might be based on: Toughness (Yu) = Toughness (X;.) > Toughness (Yr) /Toughness (Xr) following common rules for estimates based on reference values Ino data at all are available for alloy ¥, the estimate is more precatious. Usually, if the alloys are similar Toughness (¥) = Fy, x Toughness (1) / Fy should provide a safe (conservative) estimate, but only if Fyy > Fyy. This should not be used if Fy < Fy. In that ease it would be safer to use the same toughness. as for X. For dissimilar alloys such estimates should not be attempted at all Estimating the effect of temperature is more dangerous because temperature affects the yield strength. One may try: Toughness (Xn), Fy YT Tounkness (Fa) * FLV, * Tovehoess (Fn) but it is advisable to base the esti available di Data handbooks (e.g (6}) provide an abundance of toughness data for high t 1 ' 1 ' ' 1 Toughness (Y,) = le on more extensive comparisons of216 strength alloys. But even for those materials estimates are often necessary: because of the enormous number of parameters not al cases are covered. The situation is worse for the most widely used materials: common structural steels (On the other hand for those materials one often knows some Charpy data. The lower shelf Charpy-value is essentially a fracture energy and can thus be expected to correlate with Ky. The Charpy test measures the total fracture energy of the specimen, which is essentially the integral of Ra) over the ligament. If the R-curve would be horizontal the value ofthis integral divided by the igament would indeed be Rand K = JER. For the low toughness at the lower shelf the R-curve will be neatly horizontal, so that a correlation between and Charpy energy is indeed likely. However, there are some essential differences between a Charpy test and a toughness test. The most important of these are the difference in loading rate (affecting the yield strength) and the difference in notch acuity (affecting the state of stress atthe notch root, and as such the stress at yield. rom empirical comparisons [10, 11,12]. it appears that, Ky = 12. JG for K inksi Jim and Cv in filbs; Ky = AJC for Kin MPa Ji and Ce in Joules A conservative lower bound [11] is claimed to be: K = 225(C* for ksi git and fs; K = 216(C9°" for MPa Jan and Joutes. ‘Toughness values so obtained would be for the same high loading rates (impact) as prevalent in the Charpy test. ‘The toughness for slower loading rates may be obtained from the same equations, while accounting for a transition temperature shift given by: AT = 215 ~ 15 F, for °F and ksi and 36 < F, < 140 ks: AT = Oi F, > 140 ksi AT = 119 ~ 0.12 F, for °Cand MPa and 250 < F, < 965 MPa; AT = Oi F, > 965 MPa. Note thatthe slower loading will cause the transition temperature to be lower (lower F,),s0 that the estimated toughness values will be useful - if the loading rates low ~ even at these lower temperatures T ~ AT: Other empirical correla tions have been derived (12) jor example the Charpy value is 20 lbs, the yield strength F,, = 6OKs! at ‘a temperature of 65F, the toughness for high loading rates would be estimated as Ky, = 12/20 = Saksiv/in. This would be a safe value to use, because the 207 toughness would be higher for slower loading. One could probably use this toughness value for temperature as 65 — (215 — 1.5 x 60) = —6OF (see shove equation for AT), Empirical correlations to estimate the toughness at the upper shelf from Charpy data have been developed as wel [10 The geertty ofthese es certain, because the upper shelf Charpy energy is not directly related to toughness. If the R-curve rises steeply - which it does atthe upper shell ~ the integral of R(a) over the ligament is not universally relatable to. Besides, much of the Charpy energy on the upper shelf is energy used for general specimen deformation rather than for fracture: in the extreme case Cherpy specimens do nol facture a al. but are simply folded double, so that only deformation ‘energy is measured. Nevertheless, in some cases the empirical correlations may be the oly way to aie at an extn, nnn nn stcomriains may “Toughness estimates can also be based upon the results of COD tests. Cor- ‘lations Between CTOD and toughnes were discussed in Chace a tating K,,oF the effect of thickness is usually somewhateasier. Assuming that trae plane stress occurs when the thickness is equal to the plastic zone, plane stress will develop during loading when B = (K,/F,)'/2n. Note that K. is used in this equation, because at very low stress (low K) there willbe plane strain, but pane stress must have developed when K = K,, otherwise K cannot. be {increased (othe full plane stress toughness K,. Further assume that plane strain isreachedat a thickness of B = 2.5(K;,/F,) These two pointsean be identified inthe diagram of toughness versus thickness as shown in Figute 74, A straight approximation between the full lane stress and fll plas sean value is ually a conservative one (compare Figure 7.1). The straight line can them wed 8 the bio the ext provided te pane ain euphc se rere 74, Conservative estimate of tsntional toughness on the basis «f two dat points (compare with Figure 1).28 {toughness for a last one other thickness are known. If only K, is known, one may estimate that Ku for full plane stres is between 2 and 2.5 times Ky (K. is another 10-20% higher), and then follow the shove procedure. (For eeamples see solutions to Exercises ) Actual and reliable data are preferable abowr esiimates. In the lack of handbook data a test on a specimen of sufficen: size is always preferable However, most engineers, unlike researchers, do not have easy access to a laboratory; even if they do, economic conditions and/o pressures for immediate answers often preclude obtaining data for each and every alloy, condition, heat {eatment, temperature, material direction and thickness. Thus, estimates are ‘ofen necessary. Used judiciously, the above procedures should provide conser, vative answers 7.6. General remarks on fatigue rate data Usually da/di¥ data obtained from handbooks or the literature need further interpretation and so, naturally, do test data. A typical data set as might appear in @ handbook (6] is shown in Figure 7.5. The set shown is rather complete as ‘covers several R-ratios. Not always are such complete sets available; in general R = 0 is covered (or R = 0.05 oF 0.1), but often this is the only R-value for ‘which data are available. ‘There may be a question about the effect of negative R. Some schools hokt {that negative tresses hardly affect crack growth, while «thers maintain that data, Clearly demonstrate a substantial effec. Both schools are essentially right. the dichotomy being a matter of interpretation, Consider Figure 7.6. A small negative stress may still have some effect until the crack is fully closed. However, after complete closure there is no longer a stress concentration: the compressive stress can be carried through and does not hhave to by-pass the crack. In tension the crack forms a load-path interruption Which must be bypassed and which causes the high erack tip stresses, crack growth and fracture in the fist place. In compression the crack faces catry through the load, a8 a loose pile of bricks can carry compressive loads. Thus indeed the compressive part of the cycle has no effect (low elastic erack tip stresses, as opposed to a very high stress concentration and yielding in tension), and e.g. crack growth curves obtained for R = 0 and R= — 1 with the loading a in Figure 7.72, would be almost identical a3 shown in Figure 7.76, But the do/AN data can be represented in two ways, By accounting for the above argument the negative part ofthe cycle coutd be nored and the stress ratio be defined >.” 0, with Ag = Aa, and AK, = ido, Jna. The dajAN data will svn in Figure 7.7c (schoo! ny + will no apparent effect of rcpscive &. On the other hand, In a ta interpretation, R = — 1, the stress range Aa = o,,and AK, = ido, “fo 219 Fate Ponti Mae, et Re 3 4 Se nny act Rak a Figure 73. Example of rat dia in MCIC Handbook [6k Courtey MCIC, faa would be employed: Since the rates are obviously the same ones asin the devo cos ae Ss ars ve 130) oe cso ea twice as large a AK, namely AK, = 2AK;, a8 in Figure 7c (school 2). Now there isa considerable effect of negative R; nevertheless the cack growth curves are identical; both rate diagrams in Figure 7.7c are based on the same data of Figure 7.7. oth interpretations are enable a ong as cach entre inthe ae manner in a crack growth analysis. Ina loading case such as in Figure 7.74 school must take R = 0,86) and AK, = Ao, na, and usedata set |t find a/4N, but school 2 mort takp the larger Aor and AK, ~ hes ra with negative R 10 find dal from data set 2. (Note that both wll find the same {lofAN as they should), Consistent interpretation and usage will prevent errors.20 ” o Fire 76, Crack Tip Steses in Tension and Compression. (4) Tenon (6) Compresion. rc) “ “ ‘Pure 7.7. fect of Negative R (2) Loading 8) Tet data (hypothetical (2) Rate pot fortwo eva) Use of dts als a One should be aware however, of the ‘schoo!’ built into computer software. Some programs use the school | interpretation, others use schoo! 2, some provide the option for either one. Fatigue crack growth properties depend upon direction T-L, L-Tete. (Figure 7.3). In particular in the case of surface flaws one must be cautious in the selection of the appropriate data set, especially in forgings where due to the different grain flow the identification of L, T and S direction is not always trivial Figure 7.8 shows a part produced by (wo different methods. Machining to [provide forthe seat of the bolt head will expose grain endings indifferent ways. ‘The erack will elect the weakest path along the long exposed grain boundaries, tnd tend to grow in the ST (or SL) direction, so that ST-data should be used for the analysis. The analyst must be cognizant of the production procedure and use the appropriate data. A crack assumed in the wrong direction (ie. reversing, the directions in the forged and machined-out-of-plate parts), results erroneous predictions, because ST toughness values and rates are usually much lower than those for LT or TL (cross grain growth), 1 is erucial that the proper crack direction and data set are selected for fracture and crack growth predictions. The analyst must be aware of how the art is made in order to identify crack location and direction of growth Arbitrarily choosing the crack in the same direction in all cases (and using the ‘wrong data) will lend to predictions with no bearing upon practice. "The . is no excuse; neither the compuler code nor fracture mechanics can be blamed if wrong assumptions are made. ‘When the environment to be considered is different from air, the rates may Faw 7.8, Elst of Production Procedure on Crack dieton.(t) Rolled ‘Oven Forbin, (0) Forsng (0)m2 be substantially different and also the effect of R-ratio. A typical data set for a Pipeline steel in seawater {13] is shown in Figure 7.9. 1! needs no emphasis that the data used, must be for the proper environment. The damage tolerance requirements may prescribe a data set for environmental effects (ASME). But {in general, an estimate must be made for the ‘average’ environmental effect as will be discussed later. 17.7, Fitting the da/AN data Data sets such as shown in Figure 7.5 still must be be used in analysis. Clearly, the scattered” data pow erpreted before they can -annox be used directly kita 24 600» 0 wm i 14 [| i i md wt io vo * 10-4} wo aoa ket Figure 7.9. ft of Sen Water on Crack Growth in Pipe Stel [13] (Courtesy ASME). 23 The problem of how to deal with scatter per se wll be discussed in Section 7.8 ‘This section concerns the first step of the interpretation, In some cases a growth rate equation may be desirable; in particular if the ‘data appear to fall on straight lines, the Paris or Walker equations may be ‘convenient. Most computer programs have options for the use of a number of ‘equations (Walker, Forman and some threshold equation are the most ‘common), but accept tabular data as wel. The later eliminates the need for force-fited (sometimes poorly-fitted) equations. Whatever the shape of the rate ‘curves, they can he represented in tabular form. This is especially convenient if the data can be read from a permanent magnetic data fle that can be called by the program. However, the program does not interpret the data, so that one 3t submit the data points as they appear in the da/dN — AK data plot. rst lines must be drawn through the data for different R and points of these Tines must appear in the table, be it that this fine may have any form without being fixed by an equation. data are used, ‘All equations derive from ‘curve’ fitting and have no physical basis. None of ‘them is fundamentally better than any other; none is more universally useful than any other. The most appropriate equation is the one giving the best fit for the case at hand, which in turn depends upon the material, environment etc ‘Also for the use of tabular data, the best fitting fine must be drawn through the dai ‘Commonly, equations are obtained by using cg. a least squares fit of the da/ AN ~ AK data, This all that can be done if the original raw test data (a versus 'N) are not available; the original crack growth data usually are not reported in the handbooks of the literature, The best fit through the da/AN ~ AK data must then be used, although this ‘best’ fit may not give the best predictions as shown below. ‘A measured crack growth curve is shown in Figure 7.10a, the da/AN data in Figure 7.10b, In this particular case a straight fine is appropriate (Paris equation). A least-squares fit of these data provides C, = 6496E-11, and ip = 3.43, When these parameters are used in a crack growth analysis routine to-re-predict the original crack growth curve, the result is as shown in Figure 7.10a, which i certainty not the best fit tothe actual crack growth curve. In this case the curvature of the predicted'curve seems appropriate (which means that the value of mis correct), but there is a more-or-less proportional error, which can be corrected by adjusting C,. The predicted life is too long as compared to the test life by a factor of 1937483/1 871 080 = 1.035. Mulplication of C, by this factor to obtain a new Cy = 1.035 x 6496E-I1 = 6.72E-11, will result in 4 better prediction as shown in Figute 7.10a, ‘Apparently the regression fit of the da/4iV data is not necessarily the best fit for analysis. The reason for ths is that in all curve iting procedures every data24 Sed 8 ao 5 Chee ats eves 4 _ 3 z 4 ‘fe dette Fo ast a1 wy ure 7.10, Finding he bet representation orate dats, Materia A-$33 "Test Dan and Predicted curves ae text (6) Regression Fit (1590 FR = 01.09) point gets equal weight. But the points for high da/AN affect only a small portion of the life, while those for low da/AN are of much more influence onthe life they are the felevant data during most of the life); the latter should weigh more in the curve fitting, 2s In the above examples, the curvature ofthe predicted curves was appropriate, tnd only C, needed adjustment. If the curvature is found to be incorrect, mp ‘must be adjusted first by trial and error, as shown in Figure 7.11. This usually causes wild gyrations in the predicted curve, but these can be ignored. The objective is to arrive at a line with the proper curvature, which will have the proper my. Once this my-value is found (TRY 2), the curve can be adjusted by adjusting C; in the manner discussed above (Figure 7.11). Clearly iting an equation i not trivial; it requires judgement, and re-predic- tion of the original data using a predictive computer code. Ifthe original crack growth curve isnot available, there is no other option than tofit the da/dN data 1s well as possible. This cam be done by regression, but this isnot necessarily the best; regression analysis may be a refinement and not necessarily an improve- iment over a hand-drawn best fit line. With the later the parameters for some ‘equations can be derived very easily by hand. ‘The Paris equation for example sa straight line on a log scale, with the equation Y= mip +b, where y = log(da/N), x = log(MK), and b = logy. By {aking two points (as far apart as possible) on the hand-drawn fit, the values of dda/AN and AK can be substituted in the equation. This provides 2 equations with two unknowns (C; and m,), which can be solved to obtain the values of the parameters. An example was given in Chapter 5. If data are available for more than one R-ratio (Figure 7.12), the Walker ‘equation can be derived in an equally simple manner. All datasets ae fited with, Faure 7H. Aéjytment of data for best preitons. Try | provides tes m adjusiment of Ca in Figure 7.10 oe tent) Materia: A-S3} Bat 350 F; R = 0.226 parallel straight lines (same m). For each line m and Cy are determined as for ‘the Paris equation, Cy being the value of C for each particular R-ratio and Cy. is the value of Cy for R = 0, The Walker equation reads: Se ge OR = Coan am oF With Kye = AKU = R): (a 4 —o _ we OR Co Rog AK oa sot Mm Ca ke 03) an = Co AK Ke By plotting the calculated Cy values versus the corresponding values of (1 — R) fon a double-logarithmic scale and drawing a straight line, one obtains C ‘ny in the same manner as above. Table 7 | 1-1 Vipure 7.13 show the det Fire 7.12. Rae ata wed fr examples of equation fae text) Tralloy (ce Tables 71 and 72. a Tale 71. Desivation of Walter equton fr data of Figue 2.42 Tai cqaton for = 0, sng AK = Pha and 2046 ae on Se = miog Bk + toe C Lop a7 9 = moe? + tog € 822 = 0301 m4 top.c Lop S66 = 1 hog 20+ tog ¢ 2525 2 Totm + bose 301 =m Loe = 82-0 x 307 = 90 C= STEWIR = OF Same for other Rat's or Cy cam be read directly fom Figure at AK = log AK = 0 esate ose a1 16EI 09 02 9 RIO on “007 ray 0) Es o7 oss LE os 13e9 06 =a 1260 05 tees as =o20 208 ED ‘Average $$ et log (1 ~ versus tog Cy pled in Figure 7.13. From Figure 7.13 and Equation (7.1: WRC, = bon Cem log = Ri M8 = top, +03n, 20 = og Ch +0 03 = 03m, m7 08203 = 17% ogc, - 90, C= 63 E10 pe ecemne are esc ete eo PAA Fork = 05, = SEN. son eo For other predicted values east cola in table m= mem = M0117 = 1396 do 63 ENO eae a a28 sean alt 8) Fire 7.13, Fit of Walker equation for data of Figure 7.12 (se Table 7.1. then be used in the form Equation (7.1), or in the alternative form of Equation (2.3) with my. = me ~ te Some computer codes use Equation (7.1), others use Equation (7°), s0 that the required input may be different (either Cu ts Mr (of Cy. ta) {Alto the parameters of the Forman equation can be obtained in a simple inanner. The Forman equation is da Kk" da Lc, ak a4) an ~ OUTRO who which can be written as: {a - RK - an} $e = GAR, sy Hence, the Forman equation ASSUMES that in a plot of og (1 — R) Ke — AK)da/AN J versus log (BK) all data for all R ratios con- solidate to one line and that ths line is straight, Figure 7.14 and Table 7.2 emonstrate this. From this line C, and mt can be determined in the same manner as for the Paris equation, Should the data (for more than one R-ratio) ‘hot fall on a single line as eg, in Figure 7.15 then the Forman equation is not 1 good fit, at least not with the value of K, that was used. Note that A, isthe Only adjustable parameter in the equation, the effect of R-ratio is implied. Since fracture of the crack growth specimens may have occurred by collapse, Ki, oF XK, would not be appropriate to ft the data and the value used becomes an ‘djeiment parameter indeed, although the generality of the equation becomes oubiful as it would apply t0 the specimens only. Yet, it can then be tried to consolidate the data using a diferent K, value Figure 7.15 ne ‘rune 7.14, Desivation of Forman equation for dat et of Figure 7.1 (ee Table 7.2 then the Forman equation is nota good fi tal Similar al, Similarly, if og(Cy) versus fin dat ease a more complicated curve i couk be used at curve filing it appealing ony ithe eqution sa sinple one not. it is muchmore convenient to use tabular data inputs instead of an equation. A computer works jus a easly witha data table (AK, da/AN for various R as with ax equation This section would not be complete without a warning with regard to conversion of equation parameters lo other equaions oto olker unit systems Mayen ha Cy Cry, te NOT taal or a Cae A le exercise deriving all parameters fr all equations from the same data se tga) wieasy denote irene Fr expe, Tabs ‘and 7.2 show the Walker and Forman equations for the same data set The ee eli ote eee Unit conversions can be made, but they are treacherous. F , ut they are treacherous, Fortunately the nd m values do not change for diferent unit systems, becase exponents ae Simeone by ature, But Cd hag tcaeCf dmenonen ot example in a Paris equation wilh AKinksiVin and dfd in ine, th oie for Crow hom (nls tht ope e Sansone one infeycle = Unit (C,) 6 fay" or Unit (CG) = — MI Uni) = aaa co)230 rman equation frm data set of Figure 7.12 nay be evar Yo obtain beter = Woks Je lam ane Mea tea [aOR = oe fn = Thin OK pa skie |e ekSee : Table 7.2. Destin of ‘empon: Ke = 100 bs Jn terval dan fom Figure 712 (ine ‘ ; ae ace ae ep gee ae fe ‘Dara of comes 4 and Soe plied in Figure 717 On o log lpia, thes cannot e propery parsed and mus be pled nxt to each ther instead ofa proper MK uation of seg lin in Figure 7.14 mffe-ns-a 2 pints: lets £7) = med a foe (07 £4) — my log 20 + te C cyan nation 144 9 = 108 8 aR herfoe hat no aaa cones aba yh ave Wad bese Gy theor has o Trlr tence Conversion ofthe unit opin and ine would eq Monveron factor of in 2 = (te) ot on Vai gin ~ (1000 psi Vin (oa a Ja ° aversion depends upon mand C, shouldbe multinied by (1000) Haaren cgeation with pl yit ACC = TED apd my = 32 forks (J ts wluebecores 13/1000)? = 25-19 fr pst Zonversion to AK in MPa mi dajAN in mjc would require a conver to. © 8 @ © @ 4 2 for 4.39) B tet for vavious Ratios pote of Formen eqeation. Date do not consolinte Lo one line and do nt ft equation Fire 7.15. Da omsim ans C6 MPa OORT (097% (MPa Jim 8) Inthe above example the conversion to MPa mi and m/eycle, C, would become 1E-9 x 0.0254(1.09)"? = 1936-11 Conversions of equations other than the Paris equation are even more tricky. They can be made, but usualy itis easier to convert the da/dN and AX data first and then derive the new equation parameters from the converted data se. I is advisable to make it a common practice to specify the unit system used when Quoting equation parameters, a practice not often followed in the literature. Equation parameters without specification of the unit system are useless; the ‘effects ofthe unit system are such that indiscriminate use of the parameters will ee tes Jin i the use of computer software some caution is needed when working with MPa Jiifor AK. In that case the erack sire, a, must be given in meters (instead ‘of the more customary mm). Since | MPa = IN/mm?, it is more rational to use the unit MPa mm, a8 some codes do, This may requ unit eonversion forthe coefficients inthe rate equations or even for the AK values in tabular input Although these sem trivial matters, itl slips can lead to suck lage eos i [redctons thatthe problem is worth mentioning may be obvious rom this section that indeed the iting of ate data snot2m Urivial. As was shown in Table 7.1_and Figure 7.13, the Walker equation is a ‘good fit (o the data set in Figure 7.12. Yet, the resulting Cy-values are off by a factor of 1.11 on the average (Table 71). Thus, al rates will be 11% higher, and therefore predicted lives will be 11% too low (conservative). Naturally this i largely due to the difficulty of reading data from a log-log-plot. However, itis not really the reading of data that counts; afterall rate data vary as strongly as they do and direct use of data would nol improve the situation. The reader is challenged to derive more accurate equation parameters (Exercise 11. It should be noted that these inaccuracies are not due to shortcomings of fracture mechanics; they are caused by material behavior. The latter isthe ‘real world’: is, theory of not. No better ‘theory’ would improve the situation (see also Chapter 12 on accuracy). ‘As may be lear from the examples, the Forman equation often will bea poor (Figure 7.15). This is because the equation iraplicity accounts for the effect of R, which is usually more complicated as can be seen from Table 7.1 ‘Any equation fitting leads to errors, even iit may scem (om » log-scale) th the fits very good. On the other hand. the use of tabular ‘because the reading of data from a fog-plot is Full of error also, It shoukd be realized however, that thi is not due to the log-plot. but due tothe materia! ‘behaving in the manner shown, No theory or equation can accurately account for material behavior if it depends as strongly 2s shown upon AK, Any small in AK will cause large errors in da/AN. 7.8. Dealing with scatter in rate data Despite regular practice a regression fit is seldom the appropriate fit of rate data, as was explained in the previous section, Similarly, use of statistical procedures to account for the scatter are seldom in accord with physical reality. Most statistical procedures ignore the physics and) mechanics of the problem, ‘Commonly e.g, the 90% confidence curves ate determined by using the individual da/4V data points as the statistical population sample. However, apnlying correct mathematics does not lend credence to the physical result First consider the nature ofthe scatter. The three main sources of scatter are (a) Consistent difference between heats 4 and of the same alloy. (b) Local differences due to inhomogeneities and ‘weak spots’ (©) Errors in measurement. Essentially, only the first source of scatter is relevant; the other two have little bearing on the problem. The three sources of scatter will be considered indepen- dently as ifthe others were absent, Consistent differences between various heats fof batches of materials, and {0 a certain degree diflerences from location to Tocation in one batch or plate, will be reflected in a more-or-less consistent difference in crack growth curves. As a consequence, the rate also will be 233 ‘consistently different as shown in Figure 7.16. This is true material scatter and it must be accounted for in an analysis because itis not known what the exact Droperties are of the batch used in the structure Next consider the scatter due to inhomogeneities and ‘weak spots’. These may ‘occur anywhere in the material, but only locally. At some locations the crack will accelerate, but soon aflerward it resumes normal behavior. Another erack (in @ different specimen) will encounter such ‘weak’ spots at other locations and will speed up locally (at different stages in life than the first one), and then resume ‘normal growth; similarly, it slows down sometimes locally. On the whole, the two crack growth curves will be essentially identical as shown in Figure 7.172, Also the rate data will be identical, except that each test provides a few outlying data points due to local higher or lower rates as indicated in Figure 7.176. The ‘outlying data points occur at different locations in the two tess. If instead of ‘wo, many tests are performed on specimens from the same plate, more and ‘more outlying data points appear and the scatter seems to become well estab- lished (Figure 7.17e). ‘Taking the upper (or 90%) and lower bound of this scatter (Figure 7.17), and reconstructing crack growth curves on this basis, results in Figure 717d Clearly. the upper and tower bound data lead to unrealistic results, as all measured crack growth curves are essentially identical, This is caused by the implicit assumption that itis possible that in some cases all crack growth could be through a continuous string of weak spots (note that this scatter is caused by local weak spots). This is an untenable assumption. Weak spots are local; in cach case only a few will be encountered; the entre bulk will not be one great ‘weak spot. Hence, the average curve isthe relevant one; the scatter’ is apparent only and not real. ‘The third type of scatter is due to measurement errors, shown in Figure 7.18 Fie 7.16. Type seater due to hatch-o-batch and hea to-het sriations.24 {In general the crack size will not be consistently over-measured or consistently tunder-measured; both over- and under-measurements occur. Even if the error would always be to one side, it would result only in a shift of the erack growth curve; the rates would not be much wrong: B+ 5-48) Aa Sat - sat = &. sy ‘Actual crack growth measurements, sometimes evra -hnw the crack to hecome §smaller. This indicates that the measurement inter! soo small itis estoncous to believe that more measurements are always beiicr. Ifthe crack appears to become smaller the data point is useless (negative d/AN). Ben ifthe crack ‘appears’ not to grow the data point is uceless: rer cannot be plctted on a logarithmic scale’ The above shows that wild gyrations can occur if measure. iments are made too close and scatter can actually get worse because too many data are taken, each of them having an etror. The main reason for the problem is the differentiation (obtaining da), an inherently inaccurate procedure. It tends to exaegcrate measurement nae, curacies. For example consider a case with a measurement accuracy of ©.00Sinch, which is about as good an accuracy as can be attained by any Fine 7.17. Sate de tla ihomogeities (2) Mensur dat for wo test(h) Rate dat for {wo tet (€) Mali ess: (d) Predictions. 25 o ‘Figne 718. Seater due to measurement esos (2) Measured dats for one te (0) Rate data for ‘one tet) Mulple tet; (@) Predictions, ‘measurement of crack size. Ifa = 0.5inch nominally, its value is between 0.495 ‘and 0.508 inch: the possible error is only one percent. Let the next measurement bbe 0.52inch, indicating a crack size between 0.515 and 0.525, again with an accuracy of one percent. The value of Aa is then between 0.525-0.495 = 0.030 and 0.515-0.505 = 0.010, with an expected value of 0,52-0.50 = 0.020. Hence, the error in rate will be as large as 50% (0.020 + 0.010). This problem is not ‘unique for crack growth; it always occurs where (numerical) differentiation is performed. To counteract this problem, the ASTM procedure recommends (o take @ ‘moving average for the rate determination. Yet the differentiation will stil ‘exaggarate measurement inaccuracies, which will appear as ‘scatter’ in the rate diagram. If a sufficient number of such data is accumulated the scatterband becomes impressive, even ifthe other sources of scatter were completely absent, ‘The resulting problem is the same as was discussed on the basis of Figure 7.17. Upper and lower bound would give crack growth curves bearing no relation fo the tests (Figure 7.184), since all tests essentially showed the same crack ‘growth curves. Using the errors of all measurements and assuming they all worked in one direction, ignores the mechanics and physics of what causes236 this ‘apparent’ scatter. Determining a 90% band with a statistical treatment that does not account for this reality is inappropriate ‘The question now arises what to do about the scatter in practical damage tolerance analysis. Most of the scatter observed in data plots can be ignored, ‘because itis apparent scatter only. The line representing the average ofthe data is the ofly realistic one (Figures 7.17 and 7.18). However. it is prudent to account for the batch-to-batch, heat-to-heat, and manufacturer-to-manufie. turer variation (Figure 7.16), because itis not known a priori which batch of ‘material will be used in the structure. As a rule of thumb, these effects can cause a difference of about a factor of 2in crack growth rates between worst and best. ‘Thus the worst would be about a factor of 1.41 higher than the average. the best ‘factor of I.4I lower than the average (1.41 x 1.41 = 2). On the logarithmic scale of the rate plot this will only bea slight shit. I a Walker, Paris or Forman uation is used, the correction can be effected by multiplication of C, b factor of 1.4, after the coeficient has heen determined (see previous section) from the average of the data. Ifa threshold equation is wsed, a assumption. rust be made on how the threshold is affected. The threshold being a disputable Property, no general rule can be given, The damage tolerance analyst will have to use judgement. Inthe case of tabular data, a shift of about 1.4 coull be used 8 well, and if threshold is used the same remarks apply as above With regard to the threshold in tabular data itis worthwhile noting that most computer codes cannot handle da/dN = 0, because tog (0) is undefined. ‘Therefore the rate at the threshold should be given as a small but non-zero umber, e.g. as 10 ®, otherwise the computer may declare an input error and the run may be terminated prematurely, {In conclusion, itis worthwhile mentioning. that integration is as forgiving. as dileremtiation is inaccurate. Integration, being the inverse of differentiation, tends to ‘cover-up (ie. compensate for) the errors of differentiation. Examples ofthis were given in Chapter 5, comparing hand analysis wih computer analysis. 7.9. Accounting for the environmental effect Should the environmental effect result in data of a form such as [13] in Figure 7.9, then none of the common equations is applicable. It might well be possible to-use a more complex equation if the computer code is equipped to deal wi these, but most general computer codes can deal only with the common equations of with tabular data. Thus a tabular representation of such data would be indicated, Inthe case of variable amplitude lowding AK varies froin one cycle to the next In one cycle the AK may call for rates atthe ‘plateau" level of Figure 7.9. inthe next cycle AK may be low and call for rates cose to the threshold, The baseline 27 {data were obtained at continually increasing AK. As the environmental effet is time dependent and measured at ‘chemical equilibrium’ at the crack tip, itis ‘questionable whether in variable amplitude loading the effect isthe same; the ‘equilibrium condition for the given AK cannot be reached immediately in one cycle. Although procedures have been proposed to deal with the problem of time-dependence. they do not appear in general computer codes nor do most of them consider the problem of variable amplitude loading. Thus, the user may take a risk when indiscriminately using the data for variable amplitude without the benefit of some variable amplitude test data. The latter can be used as a ‘check, If the analysis can reasonably reproduce the test resull, one may have soine confidence that it can deal with the structural problems to be analysed. ny cates the environmental effect is less complicated and common be used if so desired. The questions about equi variable amplitude loading and of time dependence, as mestioned above, remain however. Should the environment change during crack growth, a new and major problem arises. This occurs in many structures exposed to weather. In winter the environment is cold, dry ar, in summer itis warm, wet ait. It should be noted hhere that there may be 100% relative humidity in winter, but this cold air (even if saturated) does contain much less moisture than the summer's warm air; for example air with 100% relative humidity at room temperature contains approximately 30000 ppm of water, but air with 100% relative humidity at ~ 55°C contains only about 200 ppm of water (relative humidity isthe fraction of the possible moisture content). ‘The problem occurs for transport vehicles, bridges, airplanes, and many other structures. Thus, while the following example is for an airplane, it applies in the same manner to other structures. A wing fll of fuel may warm yp considerably when the airplane is serviced, standing on the tarmac. During ascend when many of the gust and maneuver loadings are encountered, the material is warm (heat content of fuel) the air is warm and the moisture content high; at this {emperature saturated air contains 30000 ppm of water. When fying in the stratosphere the structure and fuel cools to — $5°C. Cyclic loading still occurs, ‘but at this temperature even saturated air contains only 200 ppm of water. IFthe flight is over the ocean, the air will contain a certain amount of salt. Upon descend, the structure is cold (due to cold fuel) but the air warm with much more water. On top of this, the air may contain sulphuric acid (polluion). One side of the crack (iF it isa through crack inthe skin) is exposed to air, the other side to jet fuel. Clearly, such an environment and the changes thereof, defies any theory and any modeling. Not even tests can conceivably provice a solution to this problem. Only pragmatic engineering is of use Pragmatism can be exercised in many ways, depending upon goals, outlook, ‘and desirable conservatism. Assumptions must be made and they can have far238 reaching consequences. No categoric recipes can be given, only an example, Consider a case where the data set can be represented by a Paris or Walker equation, all with the same exponent(s), Consequently all environmental effects would be only in the coefficient C. Estimating sie relative times spent in each environment, one could take a weighted averase 0+ asshown in Table 7.3 The final equation then will be as shown (oote: the numbers in Table 7.3 are fictitious). "Naturally, these assumptions are disputable, bt so are all alternative assump- tions. Clearly, the time spent in the differen: nments can be estimated only, so that mv. “refined” assumptions are as good as the above estimates, Will different assumptions and refinements do any more than complicate the Procedure while not giving better results? The reader be the judge. ‘The above procedure is an example only. but it shows roughly how the problem is handled in some cases. Every analyst encountering similar situations ‘must make such approximations and judgements, but the problem cannot be ‘ignored, Obviously, a problem such as this isnot likely to be solved by theoreti- cal considerations. Fracture mechanics cannot be blamed. since this is an engineering problem. With the potential inaccuracies following from this ines- ‘capable engineering approach. itis hardly realistic to require extreme accuracy in stress and geometry factors. 7.10, Obtaining retardation parameters, All retardation models contain ‘unknown’ parameters; they do not, too many. assumptirus were made, e.g. about the accuracy of the plastic zone size and constraint equations, the yield strength (arbitrarily defined), the strain Toble 73, Pragmatic estimate of da/dN for changing covironment fr the at of eal forall i = Sain) TEA Gal ai) Fane Evo Wagied G ohne) io Bi Tes is teas 108s » nent ie Sit hes 2 ses @ Chae set0 tat trio = os Avie 6, = 226%, = 2 239 hardening etc. Models without any unknown parameters can hardly be expected {o have general applicability other than by happenstance. Any model that does Contain such parameters can be made to work because the parameters can be obtained (adjusted) empirically. (Chapter 5) From an engineering point of view there is no objection against empirical adjustment, provided the result is reasonably general. Objections against cempericism are somewhat exaggerated. : ALL material properties used in engineering are obtained from tests. Crack growth data (da/d¥) must be obtained from tests as well. Then it is hardly objectionable to use empirical retardation parameters. The only objection can be that retardation models are primitive and may not reflect the actual physics of retardation, so thatthe empirical parameter may not be applicable for general sage. For the latier reason a model with one and only one (adjustable) parameter is the most attractive. Without any such parameters, the model can hardly be expected to be general: with too many parameters, it will always be possible to fia specific case, but the generality becomes more dubious. Regardless of the number of parameters, an experimental determination of their values is necessary. This is called calibration. For any stch calibration at least one testis needed for the type of variable amplitude loading relevant to the structure. It must contain the signature of that loadin 6. The test can be done on a simple specimen (fis not, analysis that is at issue, but the retardation model). CT specimens should not bbe employed for this purpose unless the slress history contains no compressive stresses at all, In the case of compressive loading, the stress distribution (load path) in a CT specimen bears no resemblance to that around structural eracks in compression. (The specimen can be used in tension because ofthe uniqueness of the near crack tip stress field.) Although one test is necessary, results of several tests are preferable for a reliable calibration, Performing the calibration is a simple matter, especially if there is only one parameter in the model, Predictive crack growth analysis is performed, Proper loading spectrum with the appropriate dajdN data, and parameters forthe retardation model. The analysis is repeated several times with different parameter value(s), and the results compared with the test data. The Parameter value giving the best reproduction of the test data isthe sought value ‘An example for an aircraft spectrum loading case using the Wheeler retar- dation model was shown in Figure 5.18, Note that the case with a Wheeler exponent, 7 = 0. represents the case without retardation, called the linear case. ‘The resulls in Figure 5.18 show that y = 1.4 gives the best representation ofthe ‘The generality of this calibration must now be brought to tral, In practice it ‘mus be assumed to be rather general, but in a research project generality can be240 investigated by performing analysis for many different stress histories and by comparing the results with data obtained from tests using similar stress histories. Examples of such a generality check were presented in Figures 5.19 and 5.22. Data from tests with considerably different spectra (causing largely different crack growth) were all covered very well by the analysis, 1 should be ‘noted that an analytical result within 20% of the test life is very good, in comparison with errors caused by the data - interpretation (see previous sections) and those arising from inaccuracies in stress history prediction and lipping (Chapter 6) Unfortunately, true generality cannot be claimed. First of all, retardation depends upon the yield strength, and therefore the parameters are material specific. They must be determined for each alloy separately (as must the da/dN data). Second, and more important they are spectrum dependent. They can be used for variations of the same type of spectrum as shown in Figures 5.22, but they are not applicable to a spectrum of altogether different shape (Chapter 6) retardation parameters determined for an off-shore spectrum cannot he used for 8 pipe-tine even if the material were the same). Retardation depends upon the mixture of high and low loads. A different spectrum shape (Chapter 6) will give 2 different mixture of high and low loads. Iis retardation parameters must be determined separately, because afterall, retardation models are simplifications. Hence. retardation parameters are spectrum specific. Nature induced log-linear spectra need different parameters than man-indhiced spectra (Chapter 6), nd so ddo spectra for eg. rotating machinery. Carrying over retardation parameters from one type Of spectrum to another is not permitted: it is one of the ‘main reasons why some retardation models ate acclaimed to be better than others. Each model should be calibrated for the spectrum relevant to the application. 1f itis, any model will perform satisfactorily, if the provisions stipulated in Chapter 6 are implemented. Naturally, one can ignore retardation altogether, which is almost always conservative. Doing this i also making an assumption, namely that there is no retardation. It certainly will not provide accurate analysis resulls (though conservative). Ignoring retardation on the basis of the argument that it cannot bbe properly accounted for, is naively believing that the implied crude assumption of no retardation is better than an engincering approximation Calibration of the retardation model has several additional advantages. For example, the problem of data interpretation is partially resolved, {9 the ea bration analysis one already uses the interpreted daa, so that any misinterpre- {ations will be automatically compensated for in the calibration parameter(). Also the problem of rainflow counting (Chapter 6) is largely resolved. If the stress history is not ‘counted” in the calibration analysis, potential errors due to ‘counting are automatically compensated for in the parameters derived from the calibration analysis. Hence, in subsequent predictions the problem can be mat ignored as well, hould one elect to use counting inthe calibration analysis the parameters may be found to have different valuesif there happens to be an effect of counting at all (Chapter 6). Then of course, one must use counting in subsequent analysis as well. ‘A problem often not recognized is that the calibration is alo specifi to the computer code used, Some computer codes account for changes in slate of stress, others do not. Some use different equations for the plasic zone size than others. Consequently, the calibration parameters obtained with different computer codes will be different a5 well; they depend upon the equations used. Every retardation model must be calibrated with the same computer code as used for analysis. Transfer of calibration parameters from one code to another is not permitted, unless the codes use exactly the same equations for plastic zane, slate of stress, etc, and use the same definitions of F. This may scem cumbersome and unscientific, but the reason is that the above equations and definitions are arbitrary. [fone specific computer code is used however, for calibration as well as for predictions, this is only a one-time problem. The transfer of calibration factors may be another reason for acclaimed inaccuracy ‘of some models, Despite the arguments in this section, objections may still be raised against retardation handling and calibration, Unfortunately then there is no other alternative than using linear analysis, which is not very accurate either. In the extreme one might forego all analysis; this would provide no information at al ‘Waiting forthe perfect retardation model is no way out. Even the perfect model ‘must work with interpreted estimates of future stress histories. In that respect, they are hardly dilferent from weather predictions; the reader be the judge of the later's accuracy. Again, the resulting inaccuracies are not due to acture mechanics, but due to assumptions made for input. TAL, Exercises Estimate K, fora material with F,, = 80ksiand B = 0.3inch, ithe quoted plane strain toughness is K,, = 40ksi J, and the plane stress toughness SOs in, ‘A center cracked panel test shows @ K, = 75ksi Jin. The panel was 16inch Wide, and the fatigue crack size was eight inches. Estimate Ke if there was a stable fracture causing Aa = 0.3inch, and F, = 80Ksi Supposing ‘Sksi for the material in Exercise 3, estimate K, 4. Ifthe Charpy energy is 16 /Ihs at room temperature und the yield strength F, = Wksi, estimate the toughness. For which temperature can this toughness sil he used if the loading rate is low?m2 5. The LT toughness for material A is SOksi Jim. Estimate the toughness of ‘material B, assuming both materials have the same composition and if material A has a yield strength of 6Oksi and B a yield strength of 75ksi Estimate the toughness for B if the yield strength values are reversed 6. The transitional toughness for a thickness of 0.Sinch of » material with F, = T0ksi has been determined ax K, ~ 6Oksi in, the plane strai toughness as K, = 40ksi Jin. Estimate the toughness for a through-crack ina plate of 0.3inch thickness. 7. Determine the parameters for the Walker equation for the data in Figure 7.12 in your own way by reading the data yourself, Then calculate the data lines with your equation and draw your own conclusions. 8. Determine fie parameters for the Forman equation for the data in Figure’ 7.12 in your own way by reading the data ani! ice your own conclusions for recalculating the data with your own e:::00 9. For the cases in Exercises 7 convert the parameters to units of psi Jin. 10, Rederive the parameters in Exercises 9 hy first converting the scales. ‘Compare with the results of exercise 9 1, Using the data of Table 7.3 obtain the best coustant for an equation ‘governing five percent usage in salt air, 25% in room temperature air, remainder in cold air. References ard tet method or plane sta ci ouphncs of metal materials ASTM standard E99, 121A standard method for the determination of J, + meatoe of facture toughnes. ASTM andard ED sh Stands stitution BS 572 ‘ASTM standard C6 15} Tenative test metho for constnt-tnd-ampitae fatigue crack growth rates above 10 * imeyele ASTM standard F647 161 Anon. Damage vlron design handbook, Mat & Ceramics Ifo Center (Columbus) MCIC HBL: yearly updates. 1 1E. Compl et le), Application of fate mechan for seeton of metalic sacral mae, Arm. Soe. Metals (198). {8] D Broek Elena engineering facare mechani, thf, Nil (1986) 1B] MF. Kanpenia snd CH Popelr, adres! Faia alo Un, Press (198) {00} $7 Role and JM. Barson,Prctre and fr sd m sce, Pretce Wall 197) IMR Roberts and C. Newton, Interpretive repo await ae est conrlation with Ky da Metamg Ree Coun. baan 2 (1981). [12] B Marandet and. Sunr valuation of he toughness of thick medom strength ste by wing LLFPM and covrtions, ASTM STP 6M (1970 pp TAK 113] 0. Vosovaky,Fatirsr sack growth io 6 fine pipe seta fw eequecice in aquous og a eg 8 1 1 CHAPTER 8 Geometry factors Scope For the solution to any fracture or rack growth problem the analyst must know the geometry factors for ether K, J, or both, for the structural crack of intrest. Geomeity factors for many generic configurations already have been obtained land compiled in handbooks (1, 2, 3}. This can be done a priori for generic foading and geometries. but actual structural details are often unique so that ready made handbook solutions cannot be expected to be available. in such a ease a formal solution can be obtained in principle, but due to the complexity of structural details formal analysis may be too costly oF prohibitive for any number of other reasons. Structural cracks usually are of the part- through type 30 that 3-D analysis would be indicated. Ifthe structure is expen- sive, the cost of fracture high and only one type of crack is of interest a finite clement analysis may be a solution. If on the other hand one must consider Titerally hundreds of potential crack locations, it is not pessible to obtain ‘geometry factors in this manner other than for a few of the most critical cass. inite element analysis of models with cracks for hundreds of erack locations prohibitive. “There is a great need for simple (be it approximative) methods to obtain ‘geometry factors. Fortunately, there are many practical and easy procedures ‘which can provide geometry factors with good accuracy; they are almost always adequate for engineering applications in view ofthe general accuracy of damage tolerance analysis (Chapter 12). Methods to obtain geometry factors are the following: {a) Direct use of handbook solutions. {h) Inditect use of handbook solutions through superposition and compound ing {(c) Methods based on in and (9. ht and engineering judgement, combined with (b) m324 (2) Use of Green’s functions or weight functions, if necessary in combination with finite element stress analysis of the uncracked structure {) Detailed finite element analysis of models with cracks. Although all these wil be discussed inthis chapter, emphasis will be on simple ‘methods. Where possible, the success and accuracy of these willbe illustrated ‘through application to cases for which solutions are known. ‘Methods (a), (b) (©), and (d) are already included in some of the generat purpose fracture mechanics software [4], but they can be done easily by hand in ‘a short time, Use ofall methods requires access to one of the handbooks [1 2, 3. In many cases one must define a reference stress for the siress intensity factor, and this reference stress must be used consistently throughout the analysis, Large errors may occur when this is not done correctly, For this reason the problem of the'reference stress will be discussed first in Section 8.2. 8.2. The reference stress The stress intensity factor is defined as K = flox/Ra, in which o is the nominal stress away from the crack. The geometry factor ff accounts for the fact that average stresses in the cracked section are higher, as well as for all free boundaries affecting the erack tip stress as expressed by K. Thus f= [ha], aD, afS, ...), where W, D. and S are relevant structural dimensions (Chapter general fi = PlajL), L being a generalized length, Determining ‘geometry factors means derivation of the Function f(a/L) forthe specific loading and geometry details relevant to the crack to be analyzed. Simple methods to derive these functions are the subject of this chapter. The way in which the Function is to be used in crack growth and fracture analysis depends upon the definition of stress as well, as shown below. In the case of uniform applied stress there is no problem in the definition of ‘in the stress intensity factor. But if the stress distribution is non-uniform it may not be immediately obvious which siress should be used in the expression for K. ‘An example of this was given already in Chapter 3 where the stress intensity for the compact tension specimen was discussed. Commonly the latter is expressed in load instead of stress, because ‘the stress’ in a compact tension specimen ‘cannot be defined easily. Use of the load presents no problems ‘of a toughness test result, but in structural analysis stresses are used, not loads. Especially in crack growth analysis for comple structures, the use of loads ‘would be awkward; more specifically, computer crack growth analysis codes are ‘based upon stress. It was shown in Chapter 3 that the problem can be mended for the compact tension specimen, by simply defining a reference stress as @ = PIWB. Although this reference stress has no physical significance, the sess intensity is evaluated correctly, provided ff is changed in accordance with 2s cayaion (2.23) All analysis ca then be based upon this ference tres. "orang to hee af rofl en dso, my te eth te conser another example, Fora cetel crack ofsize aa pate wrt geder eiform tension has ben shown that) = ysetee/ WV, and Fea ange er eeT jaya. Should one inst on sing reference rest ea dre a i tiseapeon, one can legitimately 80, For example one aoe avene. deni ihe cracked selon, Ou, whichis given y Cp emi 2) Then K would become: ur = (ies ERT aa Bom = ETE oa BS a with JETT. 110.0) a 1 2a) ‘Obviously, the valve of K obtained would be identical to those based upon onistent use of Equation (8,1) in a residual strength analysis would provide Cone cat nn ems ofa, rom wich 9 cul be btn). ‘Shmiary, cyclic stress input in a crack growth analysis should then be in terms Stara the input for f should be in accordance with Equation (8.1). In this case's reference stress other than ¢ offers no advantage, but the example iouates the principe. Any relerenc ses can be used proved Bis aust accordingly, x0 that the product lo remains unaffece , Fr ee oa nn-nierm sredibtion i «bending moment (Figure Bt), The fi for this case can be found in handbock, but dierent hanitooks may provide different f's. This dilference is due to the use of @ Uierent reference stress. as shown in the figure. In one case the reference isthe trai bending ses in the outer fiber, yy, wien the other case iti the fering srese a x = given a5 0, = (I~ 22) Yq: Defining fs fa 3% Fa respectively for the two cases, a conversion can be made as follows! K = Patan SRG = Mata Vt De 2 Clearly fi. g0es to infinity Fora = W/2, because a,, goes to zero, while the stress intensity i finite, Numerical evaluation of a few cases will readily demonstrate that the two lead to the same value of K. In this case, the maximum bending sires is obviously the casiest to use. In employing handhook solutions one Should ascertain which reference stress is used, and make conversions if desirable. It is good practice to define the reference stress when quoting fi.6 Fire 81. Gcometry factors or dierent reference sess in bending. ‘The need for a clearly defined reference stress i 8 generally non-uniform stress distribution, such as io stress is usually the best choice. t obvious in the case of =f local stress can be used as the reference, by ly factor. With the nomenclature of Figure ici Keagilee noi Hen AQ Anns, FT Frere 82, Refwence sues in ewe of nom sion see dita 27 82: f= @3) Itis emphasized once more that the reference stress must be wed consistently. “The B-input tothe crack growth analysis must befor the proper reference stest, and the input of stress ranges, exceedance diagrams, stress histories and/or stress tecurtence tables must be in terms of the reference aress. The output of a ‘residual strength analysis will bein terms ofthe reference sre nd may require further interpretation. For exaropl, let in Equation (8.3) fora crack size of eg. two inches be 1.1, and alo, = 2, so that fy = 22. Supposing Kz = Dksidin, the residal srength would be evaluated for the two cases 3s a= (22 YR.) = O9ks:, and cy ~ 6O/(L.L RZ) =21.8 ks. The Former “tution predicts fracture vo occur when the local sires , has the value 10.9 “This isthe case when the maximum stress in. the distribution is 2x 109 = 21.8ksi, Although this may seem trivial this example, misintr- Pretations are easily made when input and results of a lengihy analysis are Feconsiered at some ater time or reviewed by others. Providing the definition Of the reference stress in input and output tables is good practice. 8.3. Compounding, In the general expression for the siress intensity factor K = Box/ka, the geometry factor i accounts for the effect of all bounda ‘alD ...), where W, D, etc. are relevant dimensions of the str ‘cases the individual effects of these boundaries can be found in handbooks; Their composite effect is obtained by compounding, which is multiplication of all individual effects. Possibly. the most prominent example of compounding is demonstrated by the classical solution (other solutions have since been obisined (5-8) for the clliptical surface flaw (Figure 83), The various boundary effectsare due to: back fee surface (BFS), front free surface (FFS), width (W), and crack front curvature (CFC), ie. K = Paes Dees Be Beec @ Ra = fo Jaa 4 B= Ps rs Be Base IEW is large, he = 1, and Pes = 1.12; then the classical solution provides: Ko = 112 Bere =p Hin + ale 008 9 ge 6s [f= = 212) sit 0) de8 Fine 43. Sess ites of Surface Flaw under wri ae) Bre Ye terion. (a Surface a: (4? versus re ‘The effect of the back free surface is simple, and almost always taken as Bass = 112 regardless of crack type (Section 8.5). The effect of the front free surface depends upon crack shape (Figure 8.3c). Finally, the effect of curvature isa complicated function of the parametric angle, @, and the flaw i ajc. The elliptical integral in the denominator can be and has been ale us is shown in Figure 8.3b. Its value is @, but the common representation is as in Figure 8.3b, providing Q = instead of . For a crack with a certain aspect ratio afe (or a/2e), the value of Q-can be read from this diagram and then = JQ is obtained. Alternatively, the curve can be fitted with an equation so that ¢ can be calculated in a simple ‘The function of g in the numerator must be evaluated for a certain point at the crack front, Note that Equation (8.5) implies that the stresa intensity varies along the crack front. For the deepest point A (yy = 1/2: Figure 8.2), the value of fp) equals 1 because cos(r/2) = 0. 50 that 9 fone n= ae vol re Kt = fo Sma twit set ft trom pe Fee. Tor point C athe surface (@ = O,sing = 0;coxg = 1) the numerator in [Equation (8.5) becomes equal to Jae nd therefore A= LID Iire gg vale K = fo Jra Note that f, = fro/ae.s0 that with a < c, the stress intensity at the deepest point ofthe Maw is higher than atthe surface (fy > i) ‘There is no objection in expressing A’ in terms of ¢ instead ofa; in that case: @n ke 112 than 7g Wate ale SE = hada (a) a! vale B= LDP gale = Equations (8.7) and (8.8) are equivalent, provided ft is properly adjusted as in Equation (8.8). In crack growth analysis itis often more convenient to express both K* and K’ in terms of a by using Equations (8.6) and (8.1). The different stress intensities at a and ¢ have significant consequences for the behavior of surface flaws, as discussed in Chapter 9. in obtaining f for a structural erack in a complex geometry, the effect of the individual boundaries can often be found in handbooks or determined otherwise (see following sections), By compounding these effets the ‘total’ is joblained. It should he noted that rigorous compounding adheres to slightly Uifferent rules [9]. but the procedure shown here is generally used and accepted. For example ifa free edge increases the crack tip stress by fy. then it will still do so if there isa hole, s0 that iy may be multiplied by By to get the total. 8.4. Superposition While compounding is multiplication of geometry factors, superposition is addition of stress intensity factors due to various mode / loadings. For example (Figure 8.4) in a combination of bending and tension the total erack tip stress from Equation (3.1) i:250 Fire 84. consbination of tension aml heling Ku Ku 6 FAO + FH) (89) Because the solution of the crack tip stress field is universal, the Functions f,(0) {in both terms of Equation (8.9) are identical for each i—j evmbination, so that the equation can be weitten as: Kee + = a = AERA = FEO. in ‘The total stress intensit Kan = Kart Kew ey Apparently the total K follows from the addition of stress intensity factors, ‘owing to the fact that both terms in Equation (8.9) for any of the individual stresses contain the same function of 0. It should be emphasized that super: position of stress intensities of different modes of Inavting is NOT paasible. In ‘such a case different @-funetions apply tothe differont movies and the sep from Equations (8.9) to (8.10) cannot be made Far 4 «l-russion of combined mode loading see Chapter 9 ‘As long as all loading is mode | superposition is permissible For example, i there is tension, in and out of plane bend {otal stress intensity becomes: pilus pressure inside the crack, the 251 Kr = Keaoan + Rosin + Kan + Ky 2) After the superposition is completed, f must be obtained by selecting 2 suitable fefetence stress. Generally, all damage tolerance analysis K = foxfna, and Equation (8.12) is not very suitable for aralysis, ‘Assume a combined bending and tension case as in Figure 4; the geometry factors forthe two cases are as shown, The stress intensity i Ko = Bren Oren SRG + Ben Gime A (8.13) In ordar to obtain f for the combination a reference stress must be selected. This ‘can be any of dy, Cras Set = Suen + Suns OF any other Suitable stress (Section 8.2), Selection of a, leads to: K = Cran + Baa) om VRB = B Gan VB amy B= Pr A + Bae while the use of a. provides: Ke (Fan St + fee) = Pom Va as lent and both can be used as long as one terms of the reference stress. For ‘one wants to find the residual strength for a crack size of a= 24inch (in an 8-inch wide panel with K, = 7Oksi/n) the solution is as shown in Table 81. The same result is oblained in either case, provided the numbers are interpreted correctly. The solution in terms of ois leas likely 10 erpetation ervors, but careful application of the rules will provide cotrect answers in all eases Equation (8.13) can be used diretly, provided one of the stresses is defined in value. The fracture condition K = K, (or Ky) e2ds t0:Ky + Kim = Ke. which represents a straight ce, a8 depicted in Figure 2.5. It shows the combi solutions as in Figure 8.5b. Ifthe tension stress in €g. 12.62ksi, an additional bending of 8.45ksi would cause fracture (compare with results in Table 8.1) Bourg are Bg laste a soy by mg Of Eg ons (ge282 233 ele “| 1 g[33833 jn & Slanase i — a I : aos ee ee eee eee eee ee en ! or 815 is usually more convenient, Most software is based on this soltion, is the wser deciding the reference stress, In the case of (aigu) crack grow Ko analysis Equation (8.13) cannot be used and some form of Equations (8.14) or ; B15) must be applied. Allstresses whether constant ampltde or variable - plitude most be expressed in terms ofthe reference srs, mull taken 38 tra, In most leading cass, ihe Bending and tension il bein phase (e8 ef ctenrc load) 0 tha the ratio Sexe aby the same (20 30 oye je OF en/ ue). FeBATless of the actual stress values. Hence fi can be evaluated a r Prior! with cither Equations (8.14) or (8.15) As long as the table of stresses oF bg theenceedance diagram andthe values submited tothe crackrowth analysis computer program are compatible for the proper reference stress, correct Table 8.1, Residual sength analysis wit ferent reference sess, (Cate of Figue 84). Compare resus of columns 7 and 9 a 3 wale Tthereis more than one boundary to be considered, compounding should be ac 5 performed fst and then the superposition executed, Eg ors crack als hole a] gd ina plate of width W: af] ba tele Fou = Pres Bator ames | Bea = Pre Base deat] 3 oun ux 169 a aves) oy B= Pracren Bora 2 + Phan Poe e acl ae a a oe ese], K = fom Sra i dss] 3 The analysis then proceeds a8 above i aie ok Tos esol of saperpoalion hasan important and very interesting com- " Taal sequence as is demonstrated on the basis of Figure 8,6, Consider a plate under La gue =naen ‘uniform tension with no crack (case A). A similar plate with a crack of 2a (case g 3 ae ereaeled 8) can be ‘fooled’ into believing that there is no crack by applying stresses to the i j anes faces ofthe crack. Originally the material atthe erack location was carrying a284 Pere 48. Superposition fr unerackd pate uniform stress (case A). Hence, applying this uniform 6 to the crack faces asin case B will result in the identical situation as case A. Since case B is the superposition of 2 mode I loading cases, the stress intensity of B (and of A) is equal to the sum of the stress intensities of caves C and D: en a1) ‘The stress intensity of case A is zero (K* = 0), the case of no crack, so that: Ka, 18.18) Case D cannot exist by itself asthe crack faces would interfere, but if Cis applied first then D can be superposed. I we reverse the applied stresses in case D, the sign of the cracks tip stresses (and hence of K) also will change (case E), and hence: Kem K?, 8.19) ion of Equation (8.18) and (8.19) yields: Kean KE (8.20) Equation (8.20) provides a new stress intensity factor for a crack loaded by internal pressure. This is useful for surface laws along the internal wall of here the pressurized medium enters the crack (Figure 8.7). (8.20) has more important implications. Note Picture of Figure 8.6 can be redrawn fie any jours nd snes di as long as the proper stresses are appived vo the crack ices (Figure 8:8). The following rule emerges: ‘The stress intensity for any loading case is equal to the stress intensity ‘obtained by applying to the faces of the crack the stresses that used to be there ‘when there was no crack (K" = K4) ‘This rule can be used to obtain stress intensities and geometry factors in many Combi 255 o o ‘Pare 8.7, Cracks in pressried contin (a) neal: pressure nse cack; (External I [a ) i TB/s I hel bal 7 Fire AA. “Uncraced sre dai tion’ re. cases by simple means as will be demonstrated at various places inthe following sections. In the following the rule will be referred to as the uncracked stress distribution rufe; it means that A‘ can be obtained from the solution to K'. 8.5. A simple method for asymmetric loading cases Pictorial superposition as used in Figures 8.6 and 8.8 is often very helpful in obtaining geometry factors, especially in the case of asymmetric loading, The ‘most prominent asymmetric loading case is the lug (Figure 89). This case can be built up [10] by superposition of two symmetric loading cases (B and C), ‘Then of course the downward load P and the upward stress system @256 3} {8|.fo]. [3]. wl. oo Pio 89. Seperpostion to Derive for Anymctte cs. (a Base comlguation: (h) Iver ‘operon superfluous and must be subtracted (case D). Superposition yields: Kee EK RO a2) is obvious that K® = KY (inversing the picture does not change the stresses) ‘Then it follows that the stress intensity of the lug crack is: KN K 2 ‘The geometry factors for both K* and K‘ can be found in handbooks, so that indeed K can be derived. Should the hole be negligably small, and W large, then Equation (8.22) leads to: w= (feos am where P's the load per UNIT thickness, Equilibrium requires that P = 1.50 that 2K = REARS or KY (82) 257 (Sr esi rs (624) meter D and small #¥, compounding of ff must be Khe (Foe be Fe + Poteet Vta)|2 = pola (625 B= 3 Por Bar + Sn Ban [As the various compounded geometry factors may be based on different feference stresses, the equation should take proper account of, and be modified for, the selected reference stress in the way shown in previous sections. ‘Geometry factors for almost all asymmetric loading cases can be obtained in this manner. The procedure isto build-up the case from symmetric cases and to subtract superfluous loadings until the original case is obtained with opposite sign. (This isthe same procedure as sometimes used in partial integration, where eventually the opposite of the original integral is obtained and the solution follows from 2 7 = sum of partials) ‘A solution for the general case of load bypass is shown in Figure 8.10. In the case of rows of fasteners (rivets, bolts, spotwelds) each fastener transfers part of the total load. The stress intensity is (see also Figure 8.9) Kem ea bre ee) (8.26) ‘Figne 410, Genera cone of seyret ong (combine with Figure RY258 which can be manipulated in the manner shown in Equations (8.22) through (6.25) to obtain f. Proper accounting for the reference siress is essential 8.6. Some easy guesses Geometry factors for complicated configurations often can be obtained in a simple manner, provided one develops somic insights. Admittedly, the accuracy may be somewhat limited, but errors are t="; uss than & few percent Since the accuracy of damage tolerance ansiysi: is «-cided primarily by loads, Stresses, material data, and assumptions (Chapters 6, 12), a 5-10% accuracy in Bis usually adequate. Several examples of these simple procedures, are Provided inthe following sections, but frst some trivial eases are discussed here, "upon which the reader can explore other possibi Consider a uniformly loaded plate asin Figure 8.114, No stresses ate acting along the center line. Cutting the plate in two (Figure 8.11b) makes no difference: two half plates can carty the load just as well as one full plate, a fail-safe or multiple-oad-path fexture often used. Next consider the cracked plate of Figure 8.11¢, Provided o/W is small, the geometry factor is appro. ‘mately f= 1, Cutting this cracked plate in two is not permissible because there Aare stresses acting across the center line. If the cut is made these stresses are released (Figure 8.114, e) It is easy to see that this will open the erack a little Pees sg 259 ‘more: K is higher. Since K = fax/ra and o is the same for the two halves (W is large), this implies that f > 1. The cut stresses existed very close to the erack ‘only (Chapter 2), hence the difference from ) = 1 cannot be lage. Itis certainly Jess than of 2.On the other hand one would expect the difference to be ‘more than one percent (1.01 < f < 2). Intuitively, one would guess the difference to be about 10 percent: = 1.1. In actuality the geometry factor for ‘he edge crack is Bf = 1.12 for small of (Figure 3.3. Hence, fone would guess f= 1.1, the estimate would have an accuracy of two percent. Belter accuracy certainly not required. Note that a guess of 1.05 would have been within seven Percent and a guess of 1.15 within three percent For large a/W the differences might be larger. However, let us explore the premise holds for larger a/W. Then fi for the single edge erack would be evaluated as 1.12 écexa/21P, while the proper solution is the polynomial shoven in Figure 3.3 (note that W = 1¥/2).Itis easy to evaluate these two alternatives for various a/W: results are shown in Figure 8.12. The agreement, for the case thin a few percent up till a/W ~ 0.5. But cracks larger than half the structural size are of no technical interest. Hence, the above "guesses" se perfectly acceptable from an engineering point of view. (See also Chapter 12 ‘on accuracy.) Section 8.3 showed the classical solution for a surface flaw. The geometry factor forthe back free surface appeared to be i = 1.12. Clealy, this geometry Fiure 812. Gcomeity facies for edge enc (esimated for L/W’ ~ 2260 factor has the same origin as the one for the edge crack discussed above. The classical solution is for an embedded elliptical crack (Figure 8.13). Cutting the plate in two (0 oblain a surface flaw, cuts some stresses acting across the center plane. Again, if one estimated these to be 10%, on would only introduce 2 two percent error, if any. Iti not difficult to figure that the actual factor of 1.12 (Equation 8.5) was simply taken from the edge crack solution above, and as such was an estimate as well. This estimate is not necessarily better than the 10% estimate. Continuing this procedure, the same arguments can be used for ‘a corner erack (Figure 8.13), Again some interface stresses ate cut, so that a geometry factor for the side Free surface must be introduced. From the above ‘one would estimate AsysPars = 1-1? = IL2I- Indeed, most handbooks provide the fee surface geometry factor for corner cracks as 1.21. The above ‘guess “error free’ as compared with the classical solution, 7. Simple solutions for holes and stress concentrations Consider the symmetric case of two very small cracks at a hole in a wide plate (no effect of width), The stress concentration at a hole is k, = 3 (Chapter 2) 50 ‘that the local stress is 3o (Figure &.14a). Hence, the case is equivalent to the one ie 8.L4b (note that the radius of the hole is very large with respect it a + 0), The stress intensity is K = or/na, where o is the nominal remote stress. Using the uncracked stress distribution rule discussed in ‘Section 8.4, it follows immediately that | = 3. One may apply the free surface correction (Section 8.6) and obtain ff = 3 x 1.12 = 3.36, [Next consider [10] the same situation with long cracks (a/¥ sill small), as depicted in Figure 8,15. 1t will not make any difference whether there isa hole for not. Ifthe hole were filed with material (Figure 8.15b) the latter would not ceay any load because the crack face isa free surface; the filler material would ‘simply ‘go along forthe ride’ undergoing rigid body motion only. Hence the case ‘of Figure 8.15b is identical to that of Figure 8.15; the crack behaves as if the bow, oss Fue 812. Solution for embeded erck applied Surface cack: ()C ———-4 261 o ® o Figure 8.15. Large crac at hoe (0) Phyl crack; (b Efe crack far asthe plate is concerned there is a crack of 2a + D.C Wis large, the stress intensity is: am effective length 224 K = fe Jaq inwhich = 1 for large W. ean Expressing K on the bass of the true erack sie , yields (with B. = 1) Kn orig = 0 ela DR TF Dita a fra = flo Sra (8.28) 8 = JTF DRao Figure 8.16.8 or symmetric cracks This P can be calculated for various o/D values and plotted as in Figure 8.163 (curve) Also the point f = 3.36 for a = Ois plotted. The curve is aceuate for large a, but not for small a. The point 3.36 is accurate for small a, The curve between is certainly not like curves A or Bin Figure 8.16a. Hence, a line faired between (0, 3.36) and the curve (accurate for large a) must be very close to the truth, ie. within a few percent. The result, shown in Figure 8.16b. can be ‘compared with a solution by Rowie (11) Clearly, the two are very close. The Bowie solution is considered the best available, but it was obtained by numerical ‘methods, and hence, it will have some error. Then itis hard to say whether the Bowie solution or the above simple solution isthe better one, But even if one were to accept the Bowie solution as the absolute indard, the simple solution reat effort and cost, while ick of an envelope’, as shown ibove. the simple solution is hardly worthwhile if a SEMI-rig- ‘orous solution is available anyway. However, the above example demonstrates that accurate results ean be obtained by simple means. More important, the simple procedure cen now be used with confidence o generate solutions to more ‘complicated problems for which no sem’ vigorous answers have as yet been ‘as shown below. ile crack at a hole (Figure 17) the above procedure leads 10 {1 = 3.36 for small a, and for large a (large 1) One might argue 263 ow Lb 3 d ‘Fire 8.17. One crack a hole. (a Single long crac; (6) fo singe cack. K = oVRig = 0 fala + DE = SEF Diao Jra a) pean) (8.29) B= JRF Ra Which results in Figure 8.17b. For large a this provides <1, which can be understood if i is recognized that the physical crack has only one tip and is defined as a (not 20), "Next, consider binxial loading, as shown in Figure 8.18, Taking one step back, in the uniaxial case, (Figure 8.14) there will be a compressive stress (— being the nominal applied stress) at the poles of the hole, This ean easily be demonstrated by pulling a sheet of paper with a cicular hole: the paper buckles above and below the hole as a consequence of these compressive stresses. Thus, in case of biaxial stresses (Figure 8 18a) with o, 2, there isa tension 1¢ due 1o 4 at the equator, and a compression — o due to ¢, = total st of Xe ~ a = 2s. This leads to ff = 1.12 x 2 =264 je o Stee a ole de 7 deviation a short rack (ses component perpendicular rack) Na fod path ‘deviation a ip flog cach. Pau (Load p cracks (a ~ 0). For long cracks Figure 8.15 and Equation (8.28) apply, because the transverse stress is not deviated at the crack tip: one can cut the plate in two horizontally if only a, is present (only when the crack is small will he load be deviated around the hole and have an effect on K causing f! = 2.24 a8 shown in Figure 8.18c, d). Hence for long cracks Equation (8.28) does apply, and the geometry factor is as shown in Figure 8.19. ‘This procedure can be extended to any kind of biaxial loading. Ifo, = 0.50), as in a pressure vessel with end caps, it follows that the equator stress is Je — 050 = 2.50. Hence, for smal cracks f= 1.12 x 25 = 280. For long cracks again Equation (8.28) applies, and the resulting fis as shown in Figure 8.19, Similarly, geometry factors for other biaxiality ratios can be obtained as shown in Figure 8.19. The same procedure can be followed for single cracks using Equation (8.29). Should the width 1 be small, compounding will be necessary, a may be ‘demonstrated by the following example fora symmetric crack (both sides). For small a the geometry fact 3.36 (a + 0; 6, = 0). For large cracks K = Jee Rag]W 0 Sra 8.30) Substitution of age = D/2 + a yields: 065 Figue 819. Goomety fcr for rcks a hole ender basa lxding K = fofna 30 B= \eemDE + NW JTF Dpa ‘curve forthe above f can be drawn for various o/D, and the results for small cracks ate obtained by fairing between l = 3.36 for a’ = 0 and he curve in the same manner as described above. “The above method as applied 1 uniaxial loading with large W was shown to bein excellent agreement with the Bowie solution. Thus it may be expected that the extension of the procedure 1o more complicated cass will provide geometry factors of good accuracy. The method can be used for cracks at any sires concentration. Inthe case of Figure 8.20 the stress concentration is Aw Affe chaper 2. for example 7 05" then 1 + 2/2/0.5 = 5. Using the uncracked stress rule (Section Bayes bore onc bia fo Storamalleachs Forsarp notes teens questionable whether one should apply the factor 1.12 as in the case of holes However, it should be noted that this would introduce at maximum a 12% certo. Using judgement and taking the factor as 1.06 would reduce the error to 6 percent. Clealy, the sharper the notch, the smaller the correction should be. A rule of thumb could be to take the facior a8 fis = 1+ 0.12] (&, ~ 2)fork, > Band f= 112 fork, < 3. The equation provides P= 1.12 for k, = 1 (circular holed. but some engineering judgement i tequied here Tor large cracks, the notch can be considered part of the effective crack (aq = 1 + a; Figure 8.20) as in the case of the holes. Hence:26 fe Bow (oa) JTF Ta K = Bo Jig = By (oalW) 0 Salat = fo Ra | a MF Tia ace oP for a central notch f= rr Tafusa 02s!" + wos6(-E2) 1+ a)" Lay —n(49) + 2082 ain (Et) + me (4 for an edge notch. By plotting = fish, for a = 0 and the curve of Equation (8.32), and by fairing for small cracks asin the case of holes a very good approximation of will be obtained, ; A case in point is a crack developing from a hole and causing a cracked ligament as illus deed very common case figuration is evaluated as discussed ubove,Once the ligament is broken, will emanate from the other side of the hole. In estence this will be equiva- leven nck waren iwc = d+ D2 and = 9/2 regu 2.2), Hence, k, = 1+ 2J@ + DAVADID. shih provides for small a, whi Equations (8.32) provide f for larger a. -»y she procedures as described will 267 ” co) Fre 8.21. Cracking at hole with edge distance d from fice surface. () Fist crac (6) Broken ment pos crack, {ead to geometry factors with good accuracy (compounding and superposition for various toasing cases are to be performed as described), 8.8. Simple solutions for irregular stress distributions Even if the stress distribution is non-uniform, several simple procedures to ‘obtain geometry factors can be used. The first, and easiest approach would be to approximate the stress distribution by a superposition of uniform tension and 4 (any) number of bending moments. An example is shown in Figure 822. Once is is accomplished, all methods of compounding and superposition as discussed in previous sections can be applied to arrive at the appropriate Beometry factors, provided the selected reference stress is adhered to and ‘accounted for conscientiously. Alternatively, use can be made of the uncracked stress distribution rule, discussed in Section 8.4. The rule states thatthe stress intensity factor (and thus o » ” Fire 822. Approximation of srt distbuton: superposition o b,c, shold provide god ‘resus up 0 4-0, which it wsualy «large enough erack sn for pratcal problems268 ‘B) can be obtained by applying to the faces of the crack stresses that used (0 be there when there was no crack. Hence, if the stress distribution is known (as it was before cracking), these stresses can be applied to the crack faces to obtain Kand f., beit that some work is involved. Naturally. only that part ofthe stress distribution is used that covers the crack ‘Suppose the stress distribution in the unc 8.22a. By applying this stress distribution to the crack faces a stress intensit actor can be derived using so-called Green's functions. The latter are shown for central and edge cracks in Figure 8.23. They provide the stress intensity for a point load pet unit thickness (stress) at one particular point on the crack face. ‘The total stress intensity can be oblained by (numerica Green's functions, using the stresses in the uncracked section as a series of point forces. A-certain caution is necessary, because the stresses close to the erack tip are the most influential. “This ean be demonstrated by using an example for uniform siress distr for which fis known. An approximation could be made, using the Green's functions of Figure 8 23 and the resulting rough analysis (performed by hand) would be as in Table 8.2. A more refined approximation would lead to the “analysis of Table 8.3. Clearly, the cesults depend strongly on the approximations close to the crack tip. Solutions of sufficient accuracy can be obtained only ifthe approximations at the crack tip are adequate, where the adjective ‘adequate’ is not well defined; engineering judgement is necessary. Computer software (4) ‘using Green’s Functions will account for this problem. should be noted that the above procedure provides the stress intensity factor K. What is needed is the geometry factor f. AS K = flox/na, the geometry fesneat et Figure 8.23. Most wet Gren Factions Table $2. Hand analy ueng Greens functions woth cade sept Tale (Figure 4a Rint ade Figure Sar mee ke 102 3 oon 0277 ome 5 : 8 ons 02 16 ons os 00650261 ous Lm aus 02 00s 0316 040 ous 02 0209 ao aus asi om 00% ons 02 439 ows an 02 02 im 156 = I oss tnowe = t ero 12%, butte Table 83 Wet ) With above sumptions integration canbe done i lod form, but no srs non storm this an example for a ease with Known so that Fess ca bechecked [Note (Bete reat i obaind in Table 8.3 with ier eps. (2) Assumptions central crack 22 = Zine uniform tes, ¢ = 1 vit aerage ates over iin ths ate o = I overall in ee of non-uniform se There are two sides ofthe crack (— x and), both must be accounted fr. ‘ayeat Sri see =| Ser AE rignrace ip doe fo rene on let sige of rack (Figure 823) 269eRe R oan oan 0261 ou om ous oon oo oom owe oon ms ona one 0006 oom Ke oon 5 on osm asi om ome 5 3 018 ons ane a 2086 26 0.008 00051 oon Tet side Fare 63) 0.0001 02 a o aos oon on 0.008 ons aise 02 a1 02 ous ous? ou oon 0108 oon 00% 106 g tm asa 066 000 9950 08 Same notes ply asin Table 82. ous oom 0081 00006 63.38 0056 0.056 00% oon oon | 02 02 02 on 005 oor oo 00% oo 0976: known 02 02 a2 a1 1 a1 205 002 oor oon 0.0001 03 os os ons ons 096 ome ses sms a 4 = a Fs 2 os om 4.090 09 able 5.3 More rec ind integration of cae of Table & 095 - am 85 fl = Kia, which means that ¢ has to be defined. As ussed any reference stress can be used as long asthe damage tolerance nalysis is based consistently om this same reference stres. The problem of reference stress has been addressed at several places in the preceding sections and specifically in Section 8.2. Inthe case ofa complicated stress distribution, the best option is to select the highest stress in the cracked section as the feference stress. Nol: this is not meant to be the highest strewat the crac face, because the latter may vary with crack size, The geometry factor always musi be derived fora number of crack sizes. Hence, the highest sess in the section, and not the highest stress on the crack face isthe best reference, Well software will account for this problem as well ‘An alternative procedure to obtain geomelty factors in the case of com- Dlicated stress distributions is to make use of so-called weight functions. Although this isa sound method, it requires a great deal more knowledge ofthe effect of cracks on strain and displacements, because it is based upon a displace. iment reference fora known ease(12, 13,14} tis not particularly suited for hand f), as shown in Figure 8.25d. Inthe plate with stringer, the effect ofthe latter will be negligible when the crack is small; fl = 1 for ab ~ 0. However, if the crack extends from stringer-to-stringer, the situation is quite different. In ‘an unstffened plate all load in the section of the crack musi bypass the crack inside the plate, which gives rise (0 = 1 in the firs place. Ifthe stringers are Present a second load path is available: part of the load can now by bypassed ‘outside thy >t. via the fasteners into the stringers and then back into the plate, ‘again via the fasteners (Figure 8.25c). If part of the load bypasses outside the plate, then the stressesat the crack tip (inside the plate) willbe lower. This means that K is lower. which is reflected in a lower 8 then in the unsiflened plate. Since for the latter i =, the f for a/b = 1 in the stiffened plate will be less than 1, as shown in Figure &.25d. For intermediate erack sizes f will gradually decrease from 1 at a = Oto the lower value at ajb = 1 Thus the stress intensity is lower than in the absence of stingers, which can et =| q 7 = o ——— © pass inate: cy pas hcoughscngr; () Geomelty Tato, ()216 have a dramatic effet on crack growth and residual strength, as will be discussed in Chapter 9; fracture arrest hecomes a distinct possbiliy, The above example is for mechanically fastened stringers, but similar results are oblained if the second element (doubler or arrester) is welded or an integral part ofthe structure, For example a transverse web crack in an I-beam will show a similar decrease in ff when the crack approaches the flanges. The decrease can be significant: = 03-0.5, depending upon the slfiness of Manges o steinges, the stringer spacing, and the fastener spacing ‘The decreases in K (orf) is beneficial fr the crack tip stresses (plate), but the side effect is thatthe stresses in the stringer willbe higher and thal the fasteners willbe subjected to very high shear loads. IF the crack is small, the stresses in plate and stringer willbe oughly equal; since stringer and plate are attached they must undergo equal strains and equal strains require equal siesses if the tnoduli of plate and stringer are equal. If the local stress in the stringer at the location of the erack is denoted aso, then 6) = @ as long as cracks are small, bout o) > @ when afb -+ 1 ‘stress concentration factor Z wll be defined ss 1. = Ja. F1om the above arguments it follows that L = 1 for aff = 0, and that L'> | for ab» Ny a8 shown in Figure 825e. Depending upon singer stiffness and spacing and fastener spacing, L can attain values of 2to 3 (ie. the local stress inthe stringer will be 2 to 3 times higher than the applied stress). In the sme vein the shear com the fasteners (or welds) will increase from essentially zero at a/h = O10 @ significant shear load (siess) when a/b reaches |. Because the load bypass occurs ‘ery close to the erack (Chapter 2) only the fasteners close to the eracked section carry the shear. The so called sifening ratio, n, which reflects the ratio of cross-sectional area of stringers and skin, and the Fstener spacing determine the magnitude of f and L. For any configuration. fl Land fastencr shear cam be really calculated by smeans of numerical analysis of elosed form solutions [16, 17, I. oF by faite clement analysis. The former are preferable because they permit parametic analysis. Due tothe high stringer stress and fastener loads plastic deformation may occur, which may affect load transfer and therefore alter and 1. The closed form analysis procedures can properly account for thes effect, but then ‘only specific solutions ate possible. For cases without plastic effets generic and ‘parametric solutions have been obtained and the results are readily available in handbooks {1, 19. s0 that they can be used for general damage tolerance analysis (Chapters 9 and 14). Unfortunately, the handbook solutions (1, 19} nly provide A, and neither L nor fastener shear, while all thre are necesary for residual strength analysis, However, ood estimatet of L can be obtained (ror fas follows. Define the seometry factor for the unstifened plate asf, and forthe siffened pane! as t 4 | 2 ‘The reason that f, < f isthe load transfer to the stringer. If there no stringer, the load carried by the plate in front of the crack tip is: P= fo, Bar ea For 0 the sress field solution provides (Chapter 3): Kk | fa Jia o-g (640) ion over a distance equal to r = a from the erack tp is the bypassed load. Hence Equation (8.3) becomes: ee (ah = JE fo ab. Clearly, the additional toad carried by the stringer is P,(aifened) minus P, (aunstiffened) and the additional stringer stress is (P, — P,)/A, isthe stringer ross sectional area. The total stress in the stringer is 0) = @ + (P, ~ PVA, so that the stress concentration L = a,fo becomes, with Equation (8.41) Lo = 1+ aB BU, ~ Bid, (8.42) The handbook provides 8, and f,; the value of L cam then be obtained from. Fquation (8.42). By taking fl, and fi, rom the handbook for a number of crack sizes between ajh = 0 and a/b = the stringer stress concentration can be cakulated as a function of crack size. An example is shown in the solution to Exercise 15 Obtaining fastener shear loads is somewhat more precarious, but a sensible estimate can be made as follows. All load is transferred by the fasteners closest to the crack plane: assume that three fasteners above the crack transfer the load into the stringer. three fasteners below the crack transfer the load back into.the plate. Together the thrce fasteners transfer the total load which is (a, — 0)4, Typically the fastener closest to the crack transfer most of the load (e.g. 60%), the other two transfer e.g. 30 and 10% respectively. Hence, the highest fastener shear load would be: Prune = 0.6 (L ~ 1) 04, (8.43) where L follows from Equation (8.42). Clearly, the fasteners must be made Strong enough to carty this shear otherwise the whole scheme will not work (Chapters 9) It turns out that the actual fastener load on the average is only in the order of 60-70% of the value in Equation (8.43) due (0 fastener hole ovalization (plasticity).28 8.11. Geometry factors for elastic-plastic fracture mechanics ‘The definition of J is (Chapter 4: DH Int Ip = PRO OE + Natya. (aay “The geometry factor fis obtained by means of any ofthe procedures discussed in the foregoing sections Ifa stress-strain eysation is available, the plastic part ‘of Jcan be expressed in o only. The commonly used equation (Chapter 4) is the Ramberg-Osgood equation which provides for the plastic strain: « = oF. ‘Then the plastic part of J becomes: te = Wg) aay ‘By performing non-linear finite element analysis on the cracked structure (direct method in Section 8.9), using the proper n and F values, J,,can be calculated from the integral formulation. With J. thus known, the stress applied to the ‘model known, and a known, the geomeirs factor cam be extracted as Fi, Sm = Eero (8.46) MG = Gig ‘The procedure must be repeated for various values of a in order to obtain ‘H(a]L) fora certain material with given » and F111 has to be determined for ‘various materials with different n, the procedure must be repeated forall these different n-values, a costly proposition Geometry factors, H, have been determined in this manner for a number of ‘configurations and n-values. These have been compiled in handbooks (20, 21) which provide hy instead of H, but H can be derived from hy if 30 desired as discussed in Chapter 4. ; ‘Simple methods to obtain H have not yet been devised, but the following. procedure might be used. If = 1 then F = Eand Equation (8.45) reduces to: J = Hot alE. ean) “This is obviously the linear elastic ease For which (Chapters 3. 4) J = xP Oral. (8.48) Clearly, in this case H = x f.Since + 1 = 2, the square in a? comes from n+ 1. Assuming that the squate of f also comes from + 1, it follows that either H = xf"! or H = (afiy!. At present there se wont whether either able agreement is ‘may not be is correct, AS a matter of )s oor sally 30 that se =e =aie (of the two express ‘obtained only for small a very useful. I appears th 29 a(sn) = afe(2)f" ao may be used if a is small and mis low, but the solution is up to the user. As forgiving as EPFM analysis is, the results are usually within acceptable engi- neering accuracy. Hence, H can be obtained from f through Equation (8.49), where fis derived by any of the procedures discussed in the previous sections, The fact that Equation (8.49) does not exactly cover the computed H values does rot make the equation invalid, because there is equal reason to suspect the finite clement analysis, which will contain increasing errors for larger a/L and larger 8.12. Bxercles 1. Using the maximum bending stres as the reference stress determine f for a crack with «/W = 05 from Figure 81. Calevlat the residual strength ven K, = TOksivin, W = inch, and F, = 6Okst 2. Calculate and plot the f.curve for a plate of with an edge crack subjected tocombined tension and in-plane bending, using the maximum total tress 8a reference; note take six values of a/W at increments of 0. starting at 0.1. A remote load P is applied 2inches from the center of the plate to the side of the crack: IW = inch 1. Repeat Exercise 2 using the uniform tension stress as a reference, 4. Given that IV = inches, a = 2inches, F, = 100ksi and K, = Soksiin, calculate the residual strength forall cases in Exercises 2 and 3. (Ignore collapse). 5. Calculate and plot the f-curve for a single crack at a hole of diameter D ina plate of width W subjected to uniform tension, Take 10 values of a/D increasing with increments of 0.2. 6. Use the result of Exercise $ to calculate the rate of fatigue crack growth Of a crack of O.3inch if D = tinch, W = 6inches, Ao = 10ksi at R= 02, da/dN = 28-9 AR”? Ki, 7. Modify Exercise $ for biaxial loading with oyioy = 3 aind for og/ a1 = ~ 0.5 (oy negative). Then repeat exercise 6 for both cases, but with a crack size of @ = 0.1 inch,280 " Determine the f-curve for symmetric cracks ata fastener hole where the fastener takes out 20% of the load for five values of a/D at increments of (04 (Note from Exercise 5 that for a/D > 0.4 the hole may already be considered part of the crack. Assume that = 8 inch, = 1 inch, A= 0.5 inch and that the stres distribution atthe ends is uniform. Calculate the residual strength diagram for the case of Exercise 8 up (0 A/D = 2: K, = Soksivin; F,, = 80ksi Determine the frcurve for eracks emanating from a semi-eliptical edge notch with a depth d and radius r in a plate of width 1 subjected 10 uniform loading. Use $ afd values at increments of 0.1: dir = 5. ‘Acceack emanated from a hole with diameter D at a distance of e from the edge ofthe plate. The entre ligament has cracked. Determine the f-curve for the erack emanating at the other side of the hole. {A shouldered part witha stress concentration factor of k, = 1.Sat the filet tadius develops a through crack. What is fi for small cracks. [An elastic finite element analysis with elements of 0.1 inch at the erack tip and a ctack of Zinches produces the following results for the longitudinal stress in the plane of the crack: 48,9 ki in element 1, 10.0ksi in element 2, 25.2si in element 3, The applied stress is non-uniform. The highest stress ppled in the model is Sksi, Calculate assuming the given stresses act in the center of the elements, and taking highest stress as a reference. ‘A finite element analysis of a model with a crack is subjected to a point load P. The crack size is | inch and the crack tip elements are 0.1 inch. The calculated displacement of the loading point is @02inch. In another run the erack is extended aver one element, and the calculated displacement is D024 inch. Calculate K if the applicd load is 10000Ihs. Assume unit thickness and E = 10000ksi “Assuming ji = 04nd a fastener spacing s/h = 0.1, calculate the L.curve ‘and fastener load curve for 5 values of > in increments of 0.1 et F £8.25e apply and assume that 60% of the toad is transferred by the fastener, B = 0.2inch, h = Sinch; pis defined as A,/bB, where A, is stringer eross section, . 21 References IND Rooke and D4. Cartwright, Compendia of tres inensy cos, HM. Stationery Office, London (1976) 12} GC Sih, Hobo of testes firs, Lehigh Unversity (1973 [BL HL Tada ols The resent cracks hanook, De Research (1973, 1986), HHL D. Brock: GEOPAC:u preprnecn for geometry factor calealation, Frcuresearch stare cnn {99 1€. Newman ad 1. Raj Stes intensity factors equations fo crackin theeimensonsl Ante Boe, ASTAY STP 794 (1983) pp. F238-1265, (6) 15 Rajwam dC: Newman Stew neni actor fr cicumeremta cick in pipes and ods lander tension and bending adh. ASTM STP 908 (1986) pp. 789-15, Im JC: Newonsn and 1S, Rao, Anas face cack in fie pltecuner enon and hod duals, NASA TRISTE (19791 Is} Gee Teamtna et al, Three dimensional ite element analysis of smal surface cracks. Frcs Mec 18 (9RD) pp 925-938 [91 D. Brock, Frctre mechanics sofware, Pracutsearch (1987) NOL. Brock ta Anpcbily of fractre tough dato rfre flows andcomer cracks at bole Nat Atrpace an (Arter) NER-TR 71033 (1970 NIL 04. tows, Analyt nine pte containing radia craks oniginting a thé boundary ‘an inter ecuaeboe, J Math and Phy 25 (1980), pp 0-71 [2LDP Rooke eta, Smple mothe of deermming snes mens fecirs AGARDograph 287 (19a) Chapter 10 1) MIF, Buckner, A Novo princi for the eaaputation of sess intensity factor, 7. Angew ‘Mash. Mech 5 (1970) pp. 829-56, 04] PC. Pala ra The weight Seton metho for determining tes intensity factors. ASTM STP AOL (1976) pp 11-49, 5] Anon. Crack grocth aot fivare, Failte Analysis Associates Hol 11 Viger Residual stength of cracked slfened pels. Bg. Frat, Mech. § (1973p. ana, 711 Smit, Deslopment ofthe fail tf desen features ofthe DC 10, ASTM STP 446 (1074) pp. tet74 ix) T Swit, Design of rodandan seuctare, AGARD LSP 97 (1978). Chipter 9 U9] Ce Pac. Theft of ete and wafer paced ringers om theses new foto of ruked she, AFFDL-TRTD-144 (1970) pp. 207-216 20) V"Komare al An enpineringarprogck for eli plasic facture anal Elect Power Re. at NPLSST (19H, PIL V- Kumar Adve clap fractre aml, let Pover Res, lst, NPAT 984)CHAPTER 9 Special subjects - 9.1 Scope ‘This chapter covers a number of special subjects. Although the procedures liseussed in Chapters ¥and 4 for ressivat scngth analysis, and those for erack ‘rowth analysis discussed in Chapters $ and 7 remain unaffected in principle, slight complications arise eg. in the analysis of sutface flaws, corner eracks and riultiple cracks, or in the ease that residual stresses are present intentionally or vertently. In other cases such 2s leak-break analysis or in a situation where load transfer to other members: may set up ean merpretation of the analysis resulls may be somewhat different than Engineering procedures to deal with such problems are discussed in this cl Examples are presented ‘The final sections provide a brief review af mode loading cases and a short discussion 0 to mixed igincering solu amage tolerance of composites 9.2. Behavior of surface flaws and corner cracks ‘The classical solutions for stress intensity and geometry factors of elliptical surface flaws and corner cracks was discussed in Chapter 8, Section 3, More recently solutions for partthrough cracks of all kinds (surface flaws and ‘corner cracks in tension and bending, surface flaws in circular hars, corner cracks at holes, ele) were obtained by Newman et al. [1. 2. 3. 4), The latter solutions are generally acckiimed 19 te K, (or Ky) fracture will continue (break), but if K < K, (or K,) the through crack will be sub-crtical, fracture will stop and only a leak is formed: leak-before-break, 3 further growth would be by erack growth instead of fracture, tacture of the part-through crack is governed by plane strain. However, the constraint for the through crack is governed by thickness, and there is @ dlistinet pagsibility that plane stress or a transitional condition prevails for the through crack, In that case the toughness is higher (Figure 3.6) as wel, (leak only). Cleaely then, the chances for leak-before-break are better when the wall is thinner, The longer the through erack, the smaller the chances th K © K,, because K depends upon the through-erack size. Thus, a circular Naw hhas a better chance of leading to a leak than an elongated crack, the former causing a through crack of size 2a = 2B, the latter a through crack of larger Suppose that break through occurs by fracture ofa small crack. Also suppose that K of the ensuing through crack is less than the toughness, K, oF K;,. whichever is applicable. Fracture being 2 fast event, the question arses whether itis indeed arrested or whether it will continue even though K is less than the toughness, Although dynamic fracture analysis (S.8] in principle can answer this. ‘question, the analysis is rather involved and beyond the realm of day-to-day ‘engineering fracture mechanics. Besides, the solution depends upon the sumptions, Pragmatically, a simple assumption can be made. If K is equal to the toughness or slightly below, the dynamic effects will probably prevail and a bbreak occur. Accounting for # reasonable dynamic effect of 15%, a leak-before- break will occur when K is less than 85% of the toughness K, or K,, whichever applicable. Instead of using 0.85 K, (or 0.85 K;,) sometimes the so-called arrest toughness, K,, is used (ASME requirements). Even with the use of the arrest toughness [5] the question of state of stress ofthe through crack remains, as the through crack could be in plane stress. In view of the fact that measure: ment of the arrest toughness is dificult and data are scant, the pragmatic, approach of using a 15% dynamic effect may he the only possible avenue. An example of analysis is given in the Solutions to Exercise 1-4 It should be noted here that leak-before-break assessment requires est: ‘ment of complete residual strength diagrams (Chapter 3) for the part-through crack as well as the through crack, Various aje ratios must be considered. Considering only one crack size or stress can easily lead to erroneous —_— ‘conclusions, especially in this case, where geometry factors depend upon crack size as well as shape. 9.4, Fracture arrest Inall cases of load transfer to second elements, racture arrest is possible; arrest ‘occuring when the fracture teaches the second element. The resulting large damage is more casily detectable which is desirable for fracture control. Ifa structure has crack arrest capabiliies, superficial inspections for large damage might be adequate. Inthe literature both the terms ‘crack arrest” and ‘fracture arrest” are used. As the condition relates to fracture and not to crack growth ‘only the term “fracture arrest is used here Fracture arrest principles will be illustrated first on the basis of stiffened panels. Subsequently the idea wll be generalized to multiple element structures and “erack" arresters. Arrest analysis procedures were developed mainly by the craft industry [5. 9-11]. At present, most large commercial aircraft are designed with Fracture arrest capability. As a matter of fact, éamage tolerance requirements for commercial aircraft (Chapter 12) tend to make this almost inescapable, Thus, it is justified to start off with an illustration of aircraft practice. An application to ship hulls is shown in Chapter 14 I was shown in Section 8.10 (Figure 8.25) that load transfer to stiffeners causes a decrease of las shown in Figure 9.8. In turn, this load transfer causes a stress concentration Lin the stringer. As in al fracture analysis a complete residual strength diagram must be determined to avoid misinterpretation, The residual strength diagram is calculated (Chapter 3, on the basis ofthe criterion Fraciure if K = K, or fata = K, 94) [Nate thatthe use of the plane stressor transitional toughness is appropriate here because the pat wil usally be of too smal thickness fr plane stain, ik all eracks considered are through the thickness. For large W the geomet fetor mil be escnily equal to unity Tor rationed panel (B~ 1). Using Equation (34) te ractre sess forthe unsiflened pane! i calculated for 8 umber of rack sizes and plted asthe dashed ine in Figure Se, Appropriate204 © Pre 98. Residual strength of slfened rane (2) Gconcey factor: (9) Srngerstres conceit ions () Resaal sremgth cng corrections should be made for small crack sizes where a is close to F, as discussed in Chapter 3. Subsequently, the residual strength diagram for the stiffened panel is es biished, also using Equation (9.4). For small cracks (ff = 1) the residual strength curve is the same as for the unstiffened prancl (Figure 9.8), but for the case of ajb = 1. the ffor the unstiffened panel is much less than 1 (Chapter 8), IF for example f= 0.4 then the residual strength of the stiffened pane! will be 104 = 2.5 times as high as that of the stiffened panel. This defines point Bin Figure 9.8¢ (o be 2.5 times as high as point 4. For intermediate erack sizes the residual strength curve forthe stiflened panel will be as shown (see in Figure 98a). Apparently the residual strength diagram of the stiffened panel exhibits a relative minimum and a relative maximum at ajh = |. Paradoxically. the + BS residual strength under the presence ofa long crack of « = bis higher than for certain smaller crack sizes; the strength increases beyond the relative minimum when the crack is longer ; Consider a case where a (atigue) crack of size a, has developed (Figure 9.8). Shoulda igh sess (lnad) occur of magnitude a compet fracture wll ene Next considera case where the crack grows by fatigue unllsize oy Ata stress of magnitude 0; a fracture will occur, but it will only run toy and be a tested al a. Note that the curve ientifes all points where Ke K,. The fracture will proceed to ay, but then agnin (K <<); although the erack size increases, the decrease in is such that hf decreases, In order for facture to a), but sucha facture would be peril ony: it would be nested ator close othe stringer upon which the residual strength sill would tea Indeed, ifthe structure is designed such that & > o,, large damage can be sustained under all circumstances where 2 <0. s0 that facture contr i facilitated by inspection for large damage. For reasons of stiffs airframes are built of skin-stiffened structure anyway it then isa smal step to provide the above fallsfety by apropriae selection of stringer size and. spacing. (As discused in Chapter 8. the reduction of by load transfer depends upon Atingrsfeess and spacing, ie. upon the stiening ratio p Up this point thesiffeners and the fasteners have been ignored. When the fracture approaches the stringer the sresses inthe stringer wil increase sig canily because of load transfer from plate to stringer, the very reason for whieh the plate experiences lower crack fip stress and lower ft This was demon stated in Figure 825. The stringer siess concentration, Lis epicedin Figure 98b, Load transfer to the stringer has to be accomodated by the fasteners causing high fasener shear, As the stringer is uncracked it willfacture when ts local stress equal tits tensile strength oy = La = F,. Thus stinger fracture occurs when the remote sires is ay = Fe/l- Using fom Figure 9.80, ths criterion provides the stringer fracture line asin Figure 9.9 A fatigne crack on Figure 9.98 will cave a ratore at ties ay, and arrest at, The stress can then be increased 1 3 (Iracture proceding to} upon which ‘he singer will break. Then lo transfer tothe stinger is no longer posibe296 Fire 9. et of singer om residual suength dagen. (2) Stinger eat: (b) Plate eric (singer material with igher than in Figure and total fracture ensues. This intersection between stringer fracture Figures 9.8¢ and 9.9a). T that # i reduced 10 the level of the plate fracture line. (Compare an undesirable situation. 1 can be improved by selecting a stringer material with a higher F,.. The diagram of Figure 9.9a will then change into that of Figure 9.9b. Again. proper design will affect the erick arrest capabilities tothe degree require; other ways to change a stringer critical case into a skin eritcal case will be shown later in this section (Figure 9.17) Cis important to note here that the stringer fracture is delemined by Fy ‘because the stringer is uncracked. Should the stringer be cracked already (which is unlikely), ts strength would be much less and be determined again by the toughness, so that it would fail at lower stress, In such a case arrest may not be possible. 297 gon 9.10. Crack sting at fastener; with broken singe. (2) Geometry factor (6) Stes con ‘otration in sringers | and 2; €) Confguaton Fastener failer due to high shear must be considered as ell. The fastener forces can be calculated accurately or estimated as discussed in Chapter 8. In ‘an uncracked structure the fasteners carry hardly any shear, so that they are ‘often fight, their primary function being to hold plate any stringers together However, in the case of a crack the fasteners must transfer load to the stringer so that they are subjected to high shear. Thus, in order to make fracture arrest ‘possible, heavier fasteners may be necessary. Tay seem that the ease discussed above is nota relevant one, because cracks will not occur between stringers. If eracks develop they will do so ata fastener hhole at a stringer (Figure 9.10). The stringer crossing the crack willbe highly stressed, and develop a crack as well, It will not take long before this stringer ‘breaks. The plate then will have o carry the oad of the broken central stringer, so that Kis increased; ft > 1. However, from then on the problem is as before: there is a crack between the two adjacent stringers. By simple redefining the siringer spacing as b, instead of 2h as before, all previous arguments will hold (Figure 9.10). with slight modilications to fi and L; these parameters can be ‘obiained from handbooks as discussed in Chapter 8. Typically, the stringer oF frame spacing in a commercial jet is ftom 8-12 inches. These structures are designed to sustain a two-bay crack (16-24 inches) with the central frame or siringer broken. If stringers have higher stiffness, or are more closely spaced, they will ransfer ‘moe lal and e mote effective in reducing f. The smaller the fastener spacing.® © we ‘Prere 9.11. Behavior of sine plates with reo” fastener spocing (a ntept Ft eens stinger (b) Welded: facture seer stringer: (0) Adhsvely fed physic! parson Reece ‘ate and singer: (et elt (nh Seal seperation, the more effective load transfer to the stringers. This can be understood intuit- ively, because the closer the fasteners at cithes side of the crack tip the more the stringer wil prevent the crack from opening (low er K and lower fi) Hence, small fastener spacing improves arrest capabtiies, as long as the fasteners are of sufficient size to catty the shear. ‘The smallest ‘fastener’ spacing is obtained if the stringer is integral or con- ‘tinuously welded (Figure 9.11a. b) Unfortuvely, this does permit fracture to Proceed into the stringer, severing the kui. in the process. There being wo Physical separation between plate and stringer the fracture will include the stringer, upon which total fracture ensues. An alternative way to obtain “zero? IB¢ Fite 9.27 Pole ares fo piping Fige 9.1. Vatious kinds of second elements fastener spacing is adhesive bonding or fillet welding (Figure 9.11c, d). Because of the physical separation the running fracture cannot penetrate the stringer The bonds or welds must be capable of transferring the load to the stringer by shear. Although in the above discussions the word stringer was used, one need only replace the work ‘stringer’ by “arrester or ‘second element’, and the arguments will be equally applicable. For example, a crack arrester for a pipeline could be designed as in Figure 9.12, and the above discussions would apply. Similarly the yao dite > ARH hen dove (angen. 1 Onkcranie pe Bark th. Tate 109 LEFM: Recor analy, ‘Sable prowth and nai ‘aia Seon ® Ww Kin Nisin The 8100 O10 0000, 130 gow NTS aon, atok ak set nT) ts por Oth name oo orm etre onme 1686 Oise Sie ante wat im ony 4687 025 00% bs 07 ates On m2 tm om ae one om ex ome aes ane om es ows kw ota one ton oa ake Oa 10 rato intity. ily > Ria ma od ibe Oa) aR Oss Ose NO tom oa alse Os onan 2058 Oss aR se om Tor cocked pa pn atalino tion ti stayin fad con Note Vl eon = fron sep Collage st. = SOs ‘previous step Mos = roms Cot eh ns sre ets ee riceleece Oieee ter 0) eee War Thebes = Fines 355 ‘aRiae oon om on on on ons ons om 5100 0100356 [Numerical calculations by iteration, according to the logic of Table 10.8, are shown in Table 10.9 for LEFM (R), and in Table 10.10 for EPFM (Jp), So many iterations are necessary both in LEFM and EPIM that the procedure is best done by computer (4). Tables 10.9 and 10.10 are result of computer analysis. 10.7. Crack growth analysis ‘Theerack growth analysis procedure is shown in Figure 10.18. Fxtensive details, algorithms and examples were shown in Chapters 5 and 6, speclic problems illustrated with examples in Chapter 9. Crack growth analysis, it was pointed ‘out, requires appropriate computer sofiware [eg 4: in few cases an analysis by hhand is pogsible (constant amplitude). The following example may serve as a ‘general illustration. ig anticipated that cracks might occur al the edge of the reinforcement plate ‘of a nozzle in a pressure vessel Figure 10.19). The pressure fluctuates approxi Fire 10.18. Crack growth analyse 357 Figure 10.19. Examples of worl wed erick imately once per two hours between p= 1500 and 3000p. Under certain ios i the process stagnales the pressure may reach 3400p, but this is fan unlikely event, Failure of the vessel would cause explosion damage and involve oss of ie. The fracture conteol objective is to reduce the probability of failuce to essentially zero even if this wou ‘conservative analysis i indicated. Cracks are expected to occur ina plane perpendicular tothe hoop stress. As 2 fist approximation, the ves i treated a8 stress Figure 1.19) isa, = pR/B = px 20/2 ~ V0p. Hence, 0 GO ksi) at maximum operating pressure (and 36k at overpressure) with Ruc- tuations of 1Sksi witha stress ratio of R= 15/30 = 0.5. The yield strength of the steel is F, = 60 ksi, The toughness is 8Oksi /n at operating temperature. Fatigue crack growth data appear to be reasonably well decribed bythe Walker equation: daldN = 10" AR"/U-R) ‘Cracks will initate as surface laws. Several ofthese may init te simultaneousTable 10.10. Competstinsl slain in EPEM 2, § Tr = j ime a < : vo rd mae Et Oe . ne ‘me Be | oe i an asi tt sett 333 002 1560 sect e133 004 1999 soot Max loud instability 4710 > dye i935 1006 15789 ssn ess 108 ise sets e219%0 1010 sat soon 3 ca fice ante waa 3 feng ites tn tin : ioe i ‘ono oa0 as ae E tom own Ses tsn tat as 7 tomo Ones} aS 2 Ma on mnt de> 0 2 toe60 "00080" S665 1570 1378 om os t F| toe ome Seas 3st i 3 ‘oie oor Seas stato aes 3 : Gintial wus eu 3 ending: thoegh cack; conftest: hentingo ain 23 ae 5 ‘Outie dancer = bach 2 2 srangg i Wall tckncs = 03 ches Bs ¥ hack st 2a = 2 chen go 3 Yd stength = 20k = ‘imate ete Scag Fe = Oks a3 eg Cotte stes Fra = Ok cae Be Referee ue p Oks ae ay Reece sain = 1919603 4 a 2S tae a dhs ar! Sain nprenng exponent x = 34 gy ig Young’ modes = Dood iz a3 Pst modus = 10066 407A 14 ba ed [Nominal siesta collapse = 4439756 ksi gE iB LE i | sii z et is i i | 7 a* Fiore 10.20, Rests of Tube 10.11 for ese of Figure 1.19 ly and join up to form a long slender flaw, or a slender flaw may form ‘mediately, For flaws with large aspect ratios, ¢ will approach 1 (F stress concentration factor, k, = 1.3 is included to account for the stress con: centration due to the weld. A hand analysis is shown in Table 10.11: the results are plotted in Figure 10.20. Cracks may of may not occur, but if they do, they will cause catastrophic failure, Therefore, the fracture contro isto be based upon inspection (Chapter 11), Since the maximuen permissible crack size (or the case of overpresste) is 0.54inch, much smaller cracks must be detected. Ultrasonic inspection (Chapter probably best suited for the purpose. Considering the circumstances surrounding piping) the probability of detecting cracks with a depth less in is considered low regardless of their length. In that case. the femaining life to failure would be 12600-92000 = 34G00hours or 3.9 years (Figure 10.20). The calculations were generally conservative, but nevertheless a factor of 400 life is considered advisable. Thus, the inspection interval is taken as 3.9/4 = 1 year. Further analysis should be performed to evaluate cracks of various depth-length ratios, Clearly, more refined analysis is possible with sophisticated computer software (4). t should be realized however, that this ‘would not necessary lead to better answers. (Chapter 12; Sources of error. ‘Most practical problems are more complicated than the above example, Questions of stress history, sequencing, clipping and truncation, retardation parameters and s0 on, must be considered, as diseussed in Chapters Sand 6. The ‘above merely shows the principles. Analysis is best performed using reliable software, but the reliability of software is largely determined by the user (input) ith a sound knowledge of the principles behind the software the effects of assumptions can be assessed (Chapter 12). 361 10.8, Exercises Given that Ky = SSksi i and that ¢, = 16k calculate a, using iteration, and! Equation (10.3), The configuration ism single edge crack (Figure 3.3) with WY = 1Dinches:F, = 100ksi B = O.8inch 2. Calculate the complete residual strength diagram for the case of Exercise 1 ‘and determine a,. How much time did you save and how much more information did you get, comparing Exercises | and 2? CCaleulate and plot the residual strength diagrams for three panels with center cracks; "= 4, 10 and 3Oinches respectively. Fy = Fy, = OOksi K, = 6Sksigfn, In each case determine the per permissible minimum residual strength is 20ksi, and if SU si, use 10 crack sizes in each ease 4, Repeat Exercise 3 for the case that J, = 0.7Skips/in; Foy = Oksi. Use H as in Table 10.5. Assume F = 1.2 E11, n= 5, and neglect clastic term of J. ‘Take erack sizes 80 that a/W in each case comes out at values for which H is known, so that you do not need to interpolate inch, How Calculate graphically the fracture stress for a fatigue crack of a using the information of Table 10.7 and the R curve of Figure 10.18; ‘much stable fracture will occur? 6, Repent Fxercise $ using the procedure of Table 10:9, assuming ff = Compare with results of Exercise 4 7. Repeat Exercises $ and 6 using Fand n of Exercise 4, Assume that J_ curve is same as R-curve in Exercise 5, and assuming W = 16inch. 8. On the basis of Exercise 7 estimate K, from J (E = 10000ksi) Then repeat Exercise 4 using LEFM procedures 9. Compare results of Bxercise 3-8 above and draw your oun conclusions References 11 MJ. trocthovenn MG, Roylebeck, avg crack extras nlc. ed SMIRT Cont. paper G4 97S) IGG. Che A procedure or incorporating thermal and residual stress into the concent of Flore analysis dnpram, ASTAY STP 8 (1979) 121 1M. Boon, Prditon of date tenting ing propoedsrain had i Prat 1 (10) pp RA Pave mechan tu eat (1987)CHAPTER 11 Fracture control - I1.A. Scope (Chapters 2 through 5 provided the concepts of damage tolerance analysis. whi ‘Chaptér 6 through 10 were concerned itl isin aad analysis practice. This chapter considers the use of the analysis f+ tra.ture control, Crack growth and fracture analysis is not an end by itself. Its sole purpose is to provide a basis for fracture contro. Fracturecontrol can be exercised in many different ways. Apart from a review of fracture “ptions, this chapter pcosides procedures for the use of ‘analysis results mn scheduling inspections, tejsir amt replacements, proof tests and 80 on. In view of the nature of the problem. the discussions do not provide clear-cut recipes. Even more so than ihe ‘vost, Sracture control measures ‘require engineering judgement and pragestnsi_ the onsiderations wywon which such judgement may be based are reviewed. Some damage tolerance require- ‘ments already specify the fracture control procedure. as discussed in Chapter 12. ‘These can be understood in the light of the possible fracture control measures presented here. After a summary of fracture control options, the selection of inspection intervals on the basis of analysis results is disussed. Fracture control by inspection is probably the most universal; safety depends upon the timely detection and repair of cracks. The sole purpose of the damage tolerance analysis is then to establish the inspection procedure and the inspection In view of the cost of analysis, itis important that this be done rationally. The analysis efforts are Futile ifthe inspection interval is still determined hapha7- ardly. The chapter is concluded with a survey of itemized and a discussion ofthe cost of fncture and frartre contrat 2. Fracture control options ‘Structural strength is affected by cracks. The residual strength as furs crack size can be calculated, using fracture mechanics concepts. A condition has 363 aaa i r Pare 11.1. Tie avilable for Gractre contre. (8) Reson strength dagiam provi 4: () ‘Crack pro cure providing H. to be set (generally by offical rules, regulations or requirements) as tothe lowest acceptable strength in the case of eracks, ie. the minimum permissible residual sirength. 9,.. When the residual strength diagram has been calculated (Figure 14a) the maximum permissible crack size, a,, follows from the minimum Permissible residual strength, The other information from analysis is the crack propagation curve, It shows how a crack develops by fatigue or stress corrosion as a function of time. The ‘maximum permissible crack, a, following from the residual strength analysis of Figure 11.12. can be plotted on the calculated crack growth curve as in Figure Mab, ‘There are several ways in which this information can be used to exercise fracture control. In all cases, the time period, H, to reach a, (Figure 11.1) is the «essential information needed. As no crack is allowed to grow beyond a,, repair ‘oF replacement is dictated by H. The following options are available for the implementation of fracture control (a) Periodic inspection; repair upon crack detection. ‘upon occurrence of partial failure. (c) Durability design: replacement or retirement after time H {(d) Periodic proof testing: repair after faire in proof test {e) Stripping: periodic removal of crack. Damage tolerance requirements sometimes prescribe the fracture control procedure. For example military airplane requirements prescribe methods (a) 8nd (c), commercial airplane requirements prescribe methods (a) and by their intent promote (b). Requirements will be discussed in Chapter 12. ‘The above fracture control options are discussed below,364 Table 11.1 inepection methods Meted Prince ‘Naked eye asd y Only a pce aly smagnitying els, ps ances tnd minors Penetrant Coloured guid is Only place ni brated on to sxcewte ener ito cack, {hen washed off Quickly dying as petion of cha is applied (Deveonen Penetant in crack crated by developer to pve coloured line Mageetic nly fr magnetic rete, materia Pate tru be dismount ‘ned inspected ole ght. Magnetic Special eatin ed es indicate rect xeny rays pass trom Method with sHroctore and are sersatity and on Bln sensi Smal (Cracks ate defineted arte awe by black line om ln, ‘iit, Unrate Univer mein satiety of probes High roquney wave ind input pes. into materia. The serve reflected by Crack, The between ple and releton Indias postion of ack dy cotent Coit induces ely Cheap meth (08 caren inthe meta caperive eqinment) ten hi indaces fed cay to apy ‘current in the el (Coils cam be made Under the presence of sll enough vo crack the induction into hoes, Sensitive. hana. Acoustic Measurement ofthe tion while ited in the atria toad. Coninvoue ‘de to pli deform suevellance pose Ilion at eack tip Totepetation ict 365 (a) Periodic inspection Safety is insured when cracks are eliminated before they impair the strength ‘more than acceptable: they must be repaired before reaching a. Therefore, any cracks must he discovered before that point by means of periodic inspection. Variousinspection methods are possible, vir. visual inspection (including toupes and magnifying lasses). peneteant, eddy current, ultrasonic, X-ray, and ‘acoustic emission, These methods are summarized in Table 11.1; for detaits on these procedures the reader is referred to the extensive literature on the subject. ‘Whatever the inspection method, there is certain erack of size, dy, detection ‘of which is questionable; during inspections before ay, the crack is unlikely (0 bee discovered. This implies that discovery and repair must occur in the time interval 1 between a4 and a, as shown in Figure 11.1. Should an inspection take pice at time 1, the erack will be missed, and should the next inspection be at ty. alter an interval H, the crack would already be too long (having reached a,). Hence, the inspection interval, I, must be shorter than H. Itis often taken as 1 = H2, but a more rational procedure to determine inspection intervals can be employed, as discussed later in this chapter. Damage tolerance analysis, to obtain residual strength and crack growth ‘curves. is performed solely to determine H and from this the inspection interval Safety is maintained by providing a sufficient number of inspections (atleast 2) during 1, to ensure crack discovery before a,. Naturally, a crack once discavered must be repaired at the operator's earliest convenience. Since a is @ permissible and not a critical rack, and since detection will commonly occur at ‘Sizes less than o,. immediate repair may not be necessary, hut any complacency will defy ll analysis and inspection efforts. Regardless of how long or short H (the inspection interval} or the inspection. ‘procedures used, safety is maintained, with some reservations as discussed later in this chapter. Whether inspections must be performed every day (e.g. H = 2 days} or every year (H1 = 2 years), there will always be two inspections between ‘a,and a,. Although a daily inspection might be cumbersome, the achieved safety is not really different in the eases of daily or yearly inspections. Ifa crack is missed in daily inspections a potential fracture will occur sooner, but if'a rack ised in yearly inspections fracture will occur nonetheless before the year is lover. Ifshort inspection intervals are undesirable, one has the option of selecting. ‘a more difficult but more refined inspection procedure with 1 smaller ay. Then Hand hence the inspection intervals, will be longer. Fewer inspections are necessary, but the cost of individual inspections may be highe It does not matter either at which time the crack initiates, as illus Figure 11.2. Inspections scheduled at e.g. H/2 interval will always give two ‘opportunities for detection regardless of when crack growth begins, provided that inspections ate scheduled at H/2 interval starting from hour zero (even if initially the chance ofa ernck is small). Similarly ifthe interval is chosen as H/3,366 Few 11.2. Two detection posibies with intra 12, males when crack sat, there will always be three inspections between vy and u,. whether the erack ‘occurs early of late, or whether simple visual or other inspections are used (different ag and #1). (0) Fail safety Providing fail safety by means of crack arresters of multiple load paths, is ‘essentially a variation of method (a): cracks or failed members must be detecied ‘and repaired. The only difference is that the struciure is designed for tolerance of large damage which is more readily apparent For example, crack arresters can be designed specifically for such a large permissible crack size (Chapter 9) that cracks will be obvious in superficial (but, frequent) inspections. Alternatively, a pipeline or pressure vessel can be designed so that cracks will cause leaks rather than breaks. As a leak is presumably obvious, no special inspections are necessary other than frequent checks for leaks (leak-before-break is slso disrussed in Chapter 9). In either case, the structure must be designed to provide fot targe and hei s revinisible crack or leak, otherwise the effort i in vain and one of the ctl tn tere control methods must be employed, The same can be accomplished by providing multiple load paths (parallel ‘members) as discussed in Chapter 9. When properly designed, the structure can 367 I sustain «, when one member fails. Inspection for cracks would not be required, but regular checks for failed members would be, OF course member {failure must be obvious, otherwise there is no advantage; a second member will soon develop cracks when it must carry additional load. (©) Durability {no inspections can or will be done, a small crack, o,, could be assumed to exist initially in the new structure (Figure 11.1). The time H, forthe crack to grow from a, to a, is then the available safe life. In that case the structure or component must be retired or replaced after e.g. H/2 hours. This is called the durability approach. This may prove a wasteful method. Since H must be very long, heavier and ‘more costly design may be necessary. If no inspections are performed, there is ‘no other choice than replacement afler H/2 hours. Inthe caseof inspections, the structure can essentially be operated forever (Figure 11.2). If no cracks should ‘occur, this would be evident; if they do occur they are detected and repaired and so-climinated. In the durability approach without any inspection, cracks of size 42, could he present after #1, but there would be no way of knowing their presence ‘A major problem with the durability approach is the necessary assumption for the size of the initial crack. This is an odd problem for structures components that are essentially defect free. In that case the iritial law may j ‘epresent an equivalent crack; at best itis the size ofa flaw that can pass initial quality control, In welded structures the assumption of an initial law is more realistic. Welds often contain defects such as porosity or lick of fusion. In Particular the latter isa sharp defect equivalent to a crack of about equal size. (8) Proof testing Whe toughness is very low the maximum permissible flaw, o,, may be smaller than the detectable crack, a, (Figure 11.3). This can also be the case when the structure isso large that inspections for cracks are impractical (e.g. a 1000 miles Pipeline), so that the “detectable” crack size is effectively infinite, In such Situations proof testing is another fracture contral option, Let at a certain time a component be subjected to @ proof StS, Opn Fracture would occur if a crack day Were present, as shown in Conversely if no fracture occurs, the maximum possible crack is djeu, 8 that 1 safe operational period 1 (For growth from dau to 4) is ensured (Figure 11.31). IF the proof testis repeated every H hours, a period of atleast H hours of safe operation is available after each successful proof test, Should failure occur during the proof test (gma Present) then a repalr oF replacement is made. ‘The life can be extended forever if no proof test failures occur, provided proof {ests are always conducted at the proper interval, H.368 Pure 11.2. roof testing. (9) Crack gromth: 1) Pro ess mo eck IE Ha 8 hy pccxaot lie of ppctine or presoe see () Lower pon to! hd afer onus SAT approach. A lie or vessel normaly filled with gas or dangerous chemicals can be proof tested (hydro-tested) with water, A failure during the proof-tes! would happen under controlted circumstances, causing a water leak only ‘cases hydro-tests ae alrendy performed anyway. Selecting the proof stress level and interval on the basis of fracture mechanics analysis, 1, would give these a rational foundation, Proof tests on structures other than pressurized containers are often hard to perform. However, if the component can be removed and easily loaded, the option is available and has been exercised (wing hinges of F-111 aircraft, Cooling the structure or component during the proof test causes a drop in {oughness. This permits the use of lower proof stresses to ‘detect’ the SMe dant, as shown in Figure 11.3c. After the test and warm up, the original toughness and residual strength are automatically restored (©) Stripping Stripping is another option for fracture control in components with permissible ‘rack sizes so small that they defy detection. If ata certain time the crack has reached the permissible size a,, machining (stripping) away a surface layer 8 ‘would reduce the crack size (0 a, = a,~5 (Figure 1b). As it would take 7 om ee eae 369 o : gare 114. Siping oe race conta a) Crack growth: (B) Siping) Practice a et) Practice in oversizing holes hours for the fperation would be ack to grow from a, to a, (Figure 1140), H hours of safe Inble after stripping. After these #7 hours the crack could be of si we that cracks are too small for detection), hence the ripping of 5 would haye to be repeated every H hours. {It would seem thal stripping cannot be repeated too many times, but it should be realized that the stripping layer 4 is very small indeed, the cracks not being. detectable, Furthermore, it should be known where the cracks occur, as for ‘example ina filet radius (Figure 1.4e, d). Machining the fillet radius by a small, mount is repeatable many times without affecting the general stress level. Shot pening after the operation, to introduce residual compressive stresses, would be beneficial, Another caap where stripping is feasible and often performed is fastener hole, Oversizigg the hole and using an oversize fastener can be repeated several times. A safe pagiod 17 is available after each such operation 11.4 The probability of missing the crack All information needgg to determine Hand the inspection interval can be calculated, except for Us detectable crack size. The later has tobe obained from Inspection experience, which fs not very well documented inthe open literature. Yet, Hor the length of the inspection interval is very sensitiveto the (choice of) detectable crack size, because the slope of the erack growth curve is small for(oe = oa ae I Figure 113. Probability of crack detection in one inepction a8 function of crack ize (0) Probability of detection in one inspection: 6) Fest of cecunsiaaces on dtetaity small cracks. Consequently, an ASSUMPTION with regard to the detectable ‘rack size, may have more weight in the determination of the inspection interval than the painstaking and costly damage tolerance analysis. This i unsatisf tory, Amore rational procedure for establishing inspection intervals is desirable, as discussed inthis and the following two sections Detection of eracks larger than the ‘detect sire" not a cert lected by many factors: () the skill ofthe inspector; (b) the specifi ‘one secific location, as oppmsed toa whole wing or bridge; (c) and viewing angles () exposure: part ofa rack may be hidden ‘other structural elements; (e) posible coreosion products inside the crack, and 50 01 ‘Typically, the probability of crack detection depends upon crack size in the manner shown in Figure 11.5. There is» certain crack sie, y, below which detection is physically impossible. For exanuple, for visual inspection this would be determined by the resolution ofthe eye, fr ultrasonic inspection by the wave length, and so on. tn realty, ays larger than these physical limits. The probability of detection is never equal to 1 even for large cracks: any crack may be missed. It follows that the probability curve must have the general form as shown in Figure 1.5, which can be described by the equation p= ae te ay where ay is the erack size for which detection is absolutely impossible (zer0 probability of detection), « and A are parameters determining the shape of the curve, The equation gives the probability. p, that a crack of size a will be detected in one inspection by one inspector ‘This inspector may nat detect the crack. The probability of non-detection is |p. Note that Equation (11.1) is curvesitting equation; it does not make any statement about the slatistics ‘volved, despite the fact that a similar equation is used for certain statistical procedures. Equation (11.1) is merely « convenient format for describing the general shape of the curve. Any othe equation providing a similar curve slipe would be equally useful am ‘A crack is subject to inspection several times before it reaches the permissible size, At each inspection there isa chance that it will be missed, At successive inspections, the crack wil be longer, and the probability of detection higher, but there is stil a chance that it goes undetected. ‘Consider 100 cracks growing at equal rates, (same population) all in the same stuge of growth (same size). Let the probability of detection at a certain inspection be p = 0.2, The probability that a crack will be is then 1~p = OR. That means that 80 cracks will go undetected. At the next inspection the cracks are longer: let the probabilty of detection then be p = 0.6, so that q = 0.4. Thus of the remaining 80 cracks 0.4 x 80 = 32 cracks will go, undetected, etc. Apparently the cumulative probability thet @ crack will be missed in successive inspections is Q = q, x qx... x q.. In the above example Q = 08 x 0.4 = 0.32: of the 100 cracks 32 cracks remain undetected after two inspections. The cumulative probability of detection is P = 1—Q. In the above example 2 = 0.68: of the 100 cracks 6% were detected after two inspections, but 32 were missed. ‘The cumulative probability of detection is then: pe t-flo-n nay where p at exch crack size follows from a curve such as in Figure 11.5 oF from an equation such as Equation (11.1). Figure 11.6 shows what happens if inspection intervals are determined as 1 = H)2, where His the time required for crack growth from a, to a,. The > & om Figure 116 Inspection interval flloing fom 1/2. (a) Probab of detecion in oe inspection ‘oF crack se a, with two methods: (6) Crack growth cure and inspection Intervals Tor Iwo, inepection methods.selected as a crack with a certain probability of detection, . detectable crack size could be defined as ay, « crack with 50% probabilly of detection. Such acrterion cert it stl leads to inconsistencies. For the case of Figure 11. either method Zor method 11 could be prescribed The detectable crack sizes, ay lad to different inspection interval, 1/2 and H,[2-Anspections would take place a indicated by arcowsin Fignre 11.65. The cracks would be inspected forthe fist time when they sill have a size smaller than ag. At this first inspection, the probability of detetion is not zer (unless @-< a). There a distinct probability, 7. that the erack is already detected during that frsinspecton: the probability that itis missed is q, = 1 p,.Atthe next inspection the rack is targer. andthe probabilty of detection i ste. By the time thé'crack reaches, it has been mspected 1 tines. The cumulative probability of the crack having been detected at any one of these inspections follows from Equation (11.2) ‘The probablity-of detection curves are different for the two procedures i Figure 11.6, but either method would be satisfactory on the bust ofthe criterion 1 = H)2, where Lis the inspection interval. The more involved method 1 (for ‘example ultrasonic) inspection with a conveniently long interval, andthe easier hod 1 (for example visual) inspection withthe more cumbersome shorter inly has appeal, because it seems consi Ack S126 mm 3 Figs 11.7. Akecotive cinch growth ese. | Of detection certainly would be Figure 11.6 and Equation (11.2), The effect is even more striking for the case of different crack growth curves, as illustrated in Figure 11.7. Such different growth curves are eg. the result of ferent #(dlferemt geametries) Let the method of inspection envisaged be one with a “detectable crack size ag” of Smm, and let in both cases the maximum permissible cragk size be a, = 33mm. Then H would be 4800 hours for both ‘racks. This would lead to an inspection interval of | = H/2 = 2400 hours in both cases. Consequently, both eracks would be inspected equally often. But crack 1 would have a nauch better chance of being detected than crack 2, the latter being much shorter at any of the inspections. Although the same criterion was applied for both eracks, not the same level of safety i achieved. A better rationale for determining inspection intervals (1) is discussed in the following. ferent, as can be readily appreciated from ILA. The physics and statistics of crack detection A sireable body of inspection data was generated by Lewis e al. (2} for several large and small structural components inspected by many inspectors. An ‘example of the results is shown in Figure 11.8. The figure shows the detection defined as the number of detections of a crack of a certain size divided by 7 7 1. wim Figs 11K Detetion sa ro Lewis ea [2 a fonction a eack iem4 the number of inspectors. Ifthe number of inspectors is large, as it was, the detection ratio will approach the true probability of detection, The number of inspectors covered in Figrue 11.8 is 96. For convenience, consider a situation where 100 inspectors would look at the same crack for which the probability of detection is known tn be p. The results of these 100 inspectors are known beforehand for any probability p, because the distribution is binomial, as if these 100 inspectors were piching marbles out of a vase with white detect) and red (miss) marbles. They sie pick a white ota red one. For certain probability, the band in which e.g. 90% of the observations must fal can be constructed a priori For example, for p = 0.8 and g = 0.2. the standard deviation of this binomial distribution is 5 = pam = ORK OTK TIM = 4, aay where 7 is the probability of detection, 4 = 1p the probability to miss, and im the number of inspector Since ttc KIM hand ie piven by 6.4 the results ure pre-determined. The expectation is for 80 inspectors to detect and for 20 to miss. There isa ive percent chance that only 8-64 = 73 inspectors would sletect, unl a five percent chance. that 80 + 6.4 = 86 inspectors would detect the crack. Conversely, if 73 inspectors ut ofthe 100 detect a certain crack, the probability of detection could still be as high a5 0.8; if 86 inspectors would detec, the probability of detection could still be a low a 0.8. The 90% band for the detection ratio of 100 inspectors MUST be between 0.73 and 0.86 if the actiil probability of detection is 2 = O.8, Hence, ifonly 20 out of 100 inspectors detect « erack. the probability ‘of detection is definitely less than 0.8. 41 follows that if 100 inspectors look at two different cracks ofthe same size, and seport 75 detects (25 misses) for one crack, and 30 detects (70 misses) for the other, then the probability of detection of these two eracks is different. rough they may have the same size, they belong to different categories with different probability-of-detection curves. For any such curve, the 90% hand for 100 inspectors is pre-established by Equation (11.3) as shown in Figure 11.9. 1f ‘some data obtained by 100 inspectors fall outs: this hand, the cracks involved belonged to different categories (populations), For example, the data in Figure 11.8 show that certain cracks of approxi ‘mately 9 mim size were detected with a ratio of 4, while others of the same size were detected with a ratio of 0.9. The above discussion shows tit the spread for 96 inspectors cannot be that large, so that these cracks of approximately ‘9mm were not of the same population. Freo. the ata set in Figure 11 R containe ‘more than one population and more than one probability curve, The ‘probability-of-detection’ curve for cracks of any type in any structure is a5 Pewe 119. Probability of detection in oe inspection; 90% band or 10 inspector low and fallows the lower hound of the data in Figure 11.8, Indeed, this tow curve should he used if inspectors were assigned just (o look for cracks ina large ie somewhere. That is not the way inspections are specified in pract ‘The data of Figure 11.8 are repeated in Figure 11.10 to show categories of dificulty of inspection. Figure 11.10 also contains data (3} for a case where Pine 1110 Data foo Fine 11.8 apt int dierent fe,376 inspectors were assigned one particular location in different aircraft: they knew ‘where to inspect and what type of crack to look for (high specificity), as is ‘normally the case; thei results are much better. Had all inspectors represented in Figure 11.10 been given specific assignments their resulls would have heen better as well. As the assignment was for any eracks in large areas the results are varied, namely the way they came out, ‘The 90 percent bands, known « priori as discussed on the basis of Fyuatio (11.3) and Figure 11.9, ean be drawn for the 96 inspectors involved for certain, probability curves, Three such bands are shown that the data cover at least thee different populations, determined by specificity (some components were small, others very large) and accessibility (some eracks were easily accessible, others were not). ‘A category or population can be defined qualitatively only. A Population is formed by the increasing crack sizes during successive st growth ofa single crack. These cracks of diferent length are indeed of the same 18 of difficulty (access). Similarly, cracks of diferent sizes atthe same location ina number of identical components would he in the same category. Rut two such crack types would be indifferent populations if one were located ina niche with a difficult viewing angle, the other in smooth flat surface. The data reported in {3}, shown in Figure 11.10, belong to more than one populatin, ‘A comprehensive damage tolerance assessment will identify types of cracks 1d their locations. Inspections will be prescribed for specific locations with known access. The probabilty-of-detection curve can be defined for the specific circumstances of the inspection envisaged. Clearly, probability-of-detection curves obtained in the laboratory are not very relevant. In the first place the inspectors know that there are cracks, otherwise the experiment would not he conducted; this introduces bias. In the second place laboratory specimens are ideally accessible under comfortable Circumstances, Third, the assignment is very specific: small specimens and usually one specific location, Results of such investigations can provide data for the most ideal circumstances only. Cracks in the tension bars of a suspension bridge will be in a different category. In each case a different probability of detection curve applies. ‘Available data obtained under realistic cireumstances for structures [2} are useful, provided itis realized that they cover more than one population. For inspections of high specificity and/or easy access their upper bound applies, while for general inspections and/or poor access their lower bound applies. As Tong as inspection assignments are accounted for, the relevant probability curve can be determined from those data. Thus 2 re-evaluation of the data would be highly worthwhile. fa the mean time, the dats set is still useful, beenuse of ite extent, Not only eddy current inspection was covered, but also X-ray. penetrant and ultrasonic inspection an 11.5. Determining the inspection interval Inspection must be preseribed Th ith due account of accessibility and specificity ‘can be done if the critical locations are properly identified. Specificity and cvessbility determine the applicable probability-of-detection curve. Categories of accessibility a {1}, and for each of these the probability lable data (1, 4] The length of the inspection interval should be established such as to provide lative probability of detection) independent of of the crack growth curve, the accessibility, and the specificity of the inspection. The aimed for cumulative probability of detection could be set for example at 95 or 98%, and be specified in damage tolerance requirements Given the calculated crack growth eueve and permissible crack size, and the probability-ofdetection for the relevant specificity and accessibility, the ‘cumulative probability of detection can be calculated for diferent lengths of the hy means of Fquation (11.2), When the results are plotted. the Al forthe desirable probabilty of detection can be obtained from the curve U1 5}, The interval will be different for differen inspection methods, different crack growth curves, accessibility and specificity. but the cumulative probability, ‘of detection is always the same (equal safety). The problems discussed in Section 11.3 are then eliminated automatically A computer [5] can perform the calculation for different interval lengths, provided the erack growth curve calculated in the damage tolerance analysis and the applicable parameters to Equation (11.1) are provided as input. For a certain inspection interval it finds the crack sizes, a which the inspection will take place, from the crack propagation curve. At each inspection (crack size) the probability of detection follows from the probability curve with the parameters appropriate for the inspection method and category. Equation (11.2) is then applied to obtain the cumulative probability. An example ofa hand calculation {or 2 inspection interval-lengths is shown in Table 11.2. For a complete analysis more inspection intervals-engths must be considered. In a computer program the procedure can be further refined by accounting for the fact that the crack at the first inspection may vary in accordance with the time of crack initiation (Figure 11.2) Typical computer results [1,5] are shown in Figures 11.11 and 11.12 for the two crack propagation curves in Figure 11.7. As discussed in Section 11.3 the criterion 7 = 11/2 would assign the same inspection interval to both cracks. According to Figures 11.11 and 11.12 this would lead to different cumulative probabilities of detection. In order to ensure the same probability in both cases intervals must be shorter in case I. Figure 11.13 shows how the probability of detection ofthe case I crack would be affected by accessibility and specificity. Although refinements can he made, the above provides a rational procedurea9 me i i ' pager a i Ff ' 7 i | 4 ; — ej | 5858 2: ' rene u d - u i a i } ed i 1 v4 Prune HAH. Cumulative peobabityof detection: Cte Veeck promth carve ofFiguce 1.7 (ema sea: good access Is for which the probability of detection + dependent ‘of the inspection technique, the crack grow curve, and the \ ment, The procedure is finding acceptance in the aircraft industry. As s expensive damage i, calculation of the crack growth curve and a, requi a i tolerance analysis, using the results for determinetion of inspection interval as + | se | His unsatisfactory indeed. The above procedure is a much better BE se ngs Gon Fae anea Gas Ro rn Sietemeiiiiecantuay cas 11.6, Fracture control plans The optimum fracture control plan depends upon the’ consequences of a Fracture. If the number of fractures experienced is considerd to be at an acceptable level with a certain fracture control plan at accepiable costs, the plan is close to optimum. Before implementation of a fracture control program the ‘objectives must be identified. Ifa structure can sustain assumed damage under an assumed loading co is not necessarily safe despite all analysis. Fefore defining the permissible residual strength or permissible crack size, the desired level of safety should be established, even if only qualitatively. It will appear that every component and structure tets different fracture control re quirements, Table 112. Hand clelation of cumulative probability of detection fortwo inspeion interval2380 a 5 ge ie ——mt Fare 11.12. Camlative probity of dteton: Cine crack growth cave of Fae 17 (smal Fiqe 1118, Cumulative pratt of detection Case 1 erck growth curve of Figure 1.7. Cat ‘A: One loation: Good aecessibty Cat E: Large aes aezest na et. 381 The consequences of a fracture must be acceptable, the fracture control measures in accordance with the acceptable risk. The structure must have adequate damage tolerance to meet this risk. Designer or manufacturer prescribe the details of the fracture control plan, the operator implements this plan through maintenance, inspection, repair, replacement, proof testing, and [possibly load monitoring. The plan must be suitable for a particular structure, component of part: it also must be suitable for the potential operators, Pro- fessional operators of pressure vessels, airplanes and the like, can implement ‘more complex fracture control measures than the general public operating automobiles. When fractures can be adequately controled by selecting ‘materials of sufficient toughness, he fracture control plan is indeed simple. But here the concern is with those cases where fractures can have serious conse- quences and! where material selection alone does not provide adequate ssfeguards against such fractures. etectable cracks ble HA shows the rlure control plans for sir ‘racks are detectable by inspection. I initial defects will not grow during service. Phan 1 is applicnble, If defects, whether initial or developing later, may grow ander service loading. a crack eventually will become critical, unless itis timely discovered and repaired, If the permissible crack is large and readily apparent Plan Ila is applicable. otherwise inspections should be scheduled in accordance with Plan It. (b) Cracks nor detectable by inspection ‘Cracks may not be detectable, either because their permissible size is so small that it defies inspection, the location not accessible, or the structure so large that inspections are wot feasible, Plans Hla and IIIb are applicable in such cases (Table 11.4) If stripping (Section 11.2) is possible, Plan Ille may present an aliernative, This could be for cracks at fastener holes (oversizing of holes and use of oversize fasteners) or for cracks at fillets ifthe component can be easly removed. Plan IV in Table 11.5 involves proof testing to show that no cracks larger than dy are present (Section 1.2) I larger cracks are present, a failure will occur during the proof test, but this failure must not be eatastro consequences. The latter should be ensured by the use of waler instead of gas for proof testing pressurized containers, testing small sections of a pipeline at a time, evacuation of surroundings, and if possible, cooling so that low pressures are suficient (6) After erack detection All fracture control calls for immediate repair or replacement when a erack is discovered. This is not always convenient. Large savings may be realized if32 Table 11.3. Prac contol pans for soit sacks tt ae dst by as lon 1 For tia defect on expected Wo pow by fatipo Calculate permisible nize of df. tres conrosion cn oeu,elelte which se st ‘an be ssa lately given the Ry af the ara Inspect nce using a eh tha ca rehiably deat dees of shove see Eriminae all detected defects larger shan above Pon ol defects nit ating ber erate Ha ‘Show hy anus or et thatthe tute ca atin without ire sh args det that he damage wil be ebro (eg eaiy ape leak one component al Repaie when damage x diner Uae Hh aon can sow, case pee cack re {stain cock tht cm eds hy wih option cee Catculate time for crack growth, " a Implement perc peton based on rack promt ck, wing ae ator ve rete of Secon 8 Sit mpeton my ax ie fc Repro ec when te ree. Tite 114, Vest cont plans ot atcpte cracks tt by ington ar too al Pont avs whee so smal Vat dco peso AMernatie a alee ie Repleeetie afer escalated Wie expe (ong aya ato Merce He Mate fest estimate of eal defect Cae permis rack ize CGaeaate crack pron fe fom tial ec iret 4, Replacetetire afer eakulated We expires ong ae qa Fs Mermaioe Mle A siping pone (Calelate permite erack stabi fee ping ph, 6 (se Section 112) Repeat sipping of 5 at itera 12) " 383 Tuble 11 Fracire contol plan fr amped etcks no detectable by inspection because Inepecton ic aot fensibe, but prof teing tpt ‘Plo 1 Vor components ov srctres ha an Be pool teed nnd where ale dutog roo? testing 8 mot 3 eaaseonhy Determine Feasible proof leat presi or lod (Cae maxim cr 76 png that cou be pein alter proo es (ee Section 1.2, ‘Gaeuate maximum peminble rack * (Cae crack growth tne, H 0m fag 04 Repeat pot let before hae expired ong adequate lor Vor detect cracks fr which ao analy B owe Repair wr replace uncool Ph Horeca erachs far which alysis de (immedi replacement impractical Menage Show that ger det cam he satan, ‘Check prt day: dil sop hole if posible ‘repne for cpr or placental Comveiens ind materia data pone cu est pie from trace, ‘Ohi etal dant ses nto. Cake a, Galea tne, for growth 4, Prepare fr vepai tepicemcrt fine H (with gute ator) capes. (Chack proth aay: dl stp ole poste, Werack rows ater than eee, update prognosis and speedup replacement oF epwir IT pie relace operational oa Rue epics aes pe Phan 11 or snares teptial thote in which crack was detected oe pt ofeach Li srtne to obtin marl proper Inaplement one of Mas Ua be ab, 1 remedial action can be scheduled for the next major overhaul or shut down, or ‘when at least operations can continue until « new part or component has been ‘manufactured and received. Whether or not this is possible depends upon the fracture control plan in force. A wellconceived Plan Ma a nformation on erack growth and residual strength. Using thi information as an initial safeguard, operation can be continued but the analysis should besed updated and Plans Via or Vib (Table 11.6) be put into action. As it is often dificult to measure the exact size and shape ofthe erack, the more stringent plan ‘Vib may be indicated. ‘A crack may be discovered accidentally in a structure not subject to a fracture control plan, When no analysis is to be done, Plan V is the only possible e Otherwise, Plans Via of b can be used. Recurrence of the incident ean be prevented using Plan VIL 11.7, Repairs ‘The sole objective of damage tolerance measures s0 that cracks can be eliminated before they become dangero cither repair or replacement of the component, The objective is ot to detern whether a erack appearing in service pre be repuired: there is no exeuse for a fracture resulting from kiown cra regardless of what analysis predicts, Naturally, replacement or repair is always convenient immediately upon crack discovery. but dimage tolerance alysis is not intended to show hiow long “one cam live with’ cracks. The above are repetitious statements of the same issue, hut repetition is justified as the objective of damage tolerance analysis is (00 often misintespreted: i is to prevent fractures, not to evaluate how long discovered 1m be sustained. ‘When crack discovery demands repair a new ctamage tolerance analysis problem arises. Not only must the repair be adcquate to restore strength, it must be analyzed for damage tolerance again. A simple cover plate usually docs not suffice (Figure 11.14). In view of Figure 10.2 such a repair may rather aggravate the situation and cause new cracks in due time. The stiffness of the over plates may introduce a stress concentration, the solution may be "worse than the disease. (Consider the fact that the increased stiffness will attract Toads to the bolt holes; tached parts must undergo the same displacement ~ strain ~ 90 tha the stiffer part will take most ofthe load/stress.) Repairs must be designed to cause gradual transfer of loads and stresses as discussed on the basis of Figure 10.2. A new damage tolerance analysis must be performed forthe repair, It should not be assumed thatthe repair isa permanent Solution. Fracture control measures must be recinstated for the repait. ‘The above may seem trivial at first sight. However, severe accidents have occurred as a result of inappropriate repairs; ina recent case more than 400 people lost their lives owing to inadequacy of a ‘so-called repair. Upon dh Covery cracks must be repaired at the earliest possible convenience. Repairs fe not a “final solution’, Damage tolerance analysis of the repair must be performed, and fracture control measures taken for the repaired structure. Frstes discussed in Chapters 11 and 12 must be accounted for. Efforts to reduce stresses by including a (stiff) load bypass or second wike the 385 ies 11.14 Umsatatory rp for eter ston se Fig 10.21 situation worse than it was, because the load willbe attracted to the stiff element (see Figure 10.21, Statistical Aspects Many of the parameters and variables playing a role in fracture control vary beyond control of human beings. Usually, the statistical variabilty is dealt with ina deterministic way by assuming that estimates of the average values provide adequate answers to engineering problems. The answers are factored to account for variability. Sometimes variability is accounted for by aking 90 or 95 percent cexceedance values ‘All material properties, including ultimate tensile strength and yield strength, show variability (scatter), Fracture toughness and crack growth properties do too. A scatter in fracture toughness of 10 to 15% is not unusual; variability by 4 factor of about 2 of fatigue crack growth rates is normal. In most cases the Structural loads are statistical variables, The pressure in a vessel may be well ‘controlled, but fluctuations may occur. The loads on bridges vary Jely depending upwn traf; they can he estimated but cannot be known until After the fact. Finally, erack detection is governed by statistical variables. There is a non-zero probability that a crack will be missed. In spite of sophisticated386 fracture control, the probability of fracture will never be zero, Ideally the fracture control plan should be based upon the acceptable probability of failure Aecause ofthe variability described. a safely fuctor is necessary fa deterministic analysis is performed In addition, there are errors due to shortcomings and limitations of the analysis, due to the limited accuracy of loads and stress history, and due 10 simplifying assumptions (Chapter 12). The magnitude of the necessary safety factor then depends upon the total uncertainty’. There isa natural tendency to cover every uncertainty when it appears by taking conservative numbers highest estimates for loads and stresses, low estimates for toughness, upper ‘bound growth rates, worst crack configurations, and so on. This amounts to a compounding of ‘safety factors’ of unknown magnitude which may lead to conservative answers, but the final conservatism is unknown. For the effects of all these assumptions see the discussion on accuracy in Chapter 12. It is preferable to use best estimates and average stata und to apply a factor of known, ‘magnitude atthe end ofthe calculation. Ideally, regulating societies or authories should establish rules and recommendations for safety factors, as they do for general design. Otherwise safety factors must be decided upon on a case-by-case basis, ‘As an example, consider a crack growth analysis to determine an inspection I. The rate of crack propagation may depend upon AK to the 4th power. IC there is a possible uncertainty of 1% in the loads, 10% in the stresses following from these loads, and 10% in ji, the potential error in AK may approach 30%. The effect on da/AN will be a factor of (1.3) = 2.86. I growth rates can vary by a factor of 2, the calculated life might be off by a factor of 5.72 ‘One could then apply a factor of 5.72 on the calculated life, by scheduling 6 inspections: 1 = 1/6, Statistical fracture mechanics have been developed, in which all variables are accounted for by the rules of statistics. Such procedures ate of great interest, provided the statistical distribution parameters could be known. If these have to be estimated the more complicated technique may not lead to more reliable answers. The simple way of applying statistical fracture mechanics is to determine the statistical distributions of all input variables. By employing a Monte Carlo technique a value for each input parameter can be selected and a deterministic analysis performed. Subsequerily, new input values are selected, again witn the Monte Carlo technique, and another deterministic analysis performed. This process is repeated many times, s0 that eventually a di bution of answers is obtained. The utter exn then be analyzed statistically to determine the probability of failure, given that certain fracture control measures ‘are implemented. The problem with the latter procedure is to establish the statistical dist bution of the input. This can be done only if assumptions are made with regard 37 ‘o interdependence. For example, the statistical distributions of such input as F,. Ky. and da/dN are dependent, because these properties ate intrinsically ‘elated to the material. Both the toughness and F, fora certain alloy may show variations by 15%. However, if F, falls at the low end of its range, itis more than likely that Ki, will fall at the high end of its range, and vice versa. By assuming that these properties are independently variable, the physics of the problem are violated no matter how elegant the Subsequent statistics and ‘mathematics. The problem could be analyzed if the physics of the dependence were known, but they are not. Determining this dependence on the basis of data ‘would require many more test data than are usually available. Although statist are of interest, it would seem that much more develop- is needed for general engineering applications. 119. ‘The cost of fracture and fracture control The acceptable consequences of failure form the basis forthe fracture control philosophy. These consequences must be weighed against the probability of failures other than by fracture. Establishing the acceptable consequences of fracture is an economic as well as an ethical problem; they must be considered the light of other circumstances endangering life. From a technical point of view. the problem can be dealt with only ithe consequences cf a failure in terms ‘of economic, ecological, and human loss can be quantified (expressed in cost) ‘and compared to the cost of fracture conttol. Then the cost effectiveness of fracture control measures can be compared with their effect, Ifthe probability of fracture is low and the consequential cost of fracture manageable, costly analysis and a costly fracture control plan cannot be justied. It is morally difficult to assign a cost to a human life, but practically itis not. An individual buying life insurance, in principle assigns a value to life, although courts of law ‘may ignore this personal assessment and appropriate higher values. Be that as it may, a monetary value is assigned. Let the total cost ofa single fracture be S and probability of fracture P, then the expected cost of fracture is PS. Obviously, if P were equal to eg. 10°*. it would not be wise to use a fracture control measure costing 10S. This would be insurance against a loss of 10 *S at a premium of 10S. ‘The potential costs of fracture include; (2) Loss of human lives (b) Impact on environment, including {€) Litigation expenses, (2) Replacement of structure. (€) Damage to buildings and surrounding structures. (1) Down time (loss of production). (@) Goodwill loss of sales and contracts, tural habitat38 ‘The total potential cost of fractureis the sum, cost is PeS. ‘The costs of fracture control include: (a) Damage tolerance analysis (20000-50000 man hours for an airplane), (b) Coupon tests and verification tests. {€) Inspections (or stripping or proof tests) (@) Repairs ot periodic replacements ‘Some of these are incurred by the manufacturer, some by the operator, but the manfuacturer’s cost (including those of fracture) are obviously calculated in the price, so that eventually all costs are incurred by the operator. The costs of fracture control as listed above, can be easily assessed, but etermining those of fracture is more difficult. Some items can be estimated, others can be ‘guessed’ only. Besides, the an ‘upon the probability of fracture, which is the most di theless, the principle applies, whether the numbers are To facilitate the discussion, consider a qualitative Fracture Control Index (FCI), a higher FCI signifying more extensive fracture control measures. The probability of failure decreases within increasing FCI (Figure 11.15a). The _ i Ke ee Tica EO) FRO (9 mcm cra ee (oy ne Soa nee Prune 1.18. Cost of factor ad frctre cont () Probab of Fineture:(h) Cot of acre ‘conto ) Total eet 5. oF the above. The anticipated 39 decrease is faster for high strength materials, because low strength (high toughness) materials have a lower probability of failure in general. The cost of fracture control increases with FCI, both for design and operation (Figure 1.15b: the curves show trends only) The probability of failure can be translated into an expendivure, and the costs plotted versus the FCI (Figure 11.15e). The minimum of the total-cost curve indicates the most economic fracture control. For high-cost structures and high-strength materials. the minimum shifts to the right so that more extensive warranted (Figure 11.15d). only parts of the structure are I. the cost of fracture control would pertain mostly to those parts, and fracture-control costs could be much lower. (If fracture of a given component would eause loss of structure a higher FCI is warranted only for that specilc part), Should the cost fines be different than assumed, @ minimum may riot be achievable. The probability of fracture of various components of a system may be different; then the probability of fracture can be made the lowest for those components for which fracture control is the easiest, This permits acceptance of a somewhat higher probability for components for which fracture control is more difficult, while the total probability could remain the same The above is but a qualitative assessment of the problem. Nevertheless. it {touches upon the relevant issues. When the logical process of decision-making. leads to a fracture control plan involving analysis, information on loads must be available, The cost of obtaining load data must be expended. Any analysis without detailed information on loads, load history and stresses is wasteful. The decision maker should be aware of the obtainable accuracy inanalysis (Chapter, 12) and of the statistical aspects of fracture control as discussed. The decision maker, if aware of the above considerations and of the sources of inaccuracy, will not embark on finite element analysis to obtain geometry factors when loads and load history are not known accurately. Cheaper, approximate analysis will sullice in such cases; uncertainties should be covered by safety factors, Where fracture control calls for inspection, the decision maker should appreciate that even detectable cracks may be missed. Inspection intervals should be determined rationally as discussed in Section 11.5, otherwise all analysis, regardless of accuracy is futile. Ifthe cost of fracture control (including analysis) far exceeds the cost of fracture, a simple fracture control plan should be selected. Analysis then may serve as « guideline: it may bound the problem. But in such cases rough assessments should slice. 11.10, Exercises 1. Determine the inspection interval ox the basis of the criterion f= 11/6 for structures with erack growth curves asin Figure 11.7, assuming the ‘detectable crack size’ is Sram, and the permissible crack size 33 mm,390 2. Using the results of Exercise 1, determine the cumulative probability of detection for the two cases, assuming that the middle probability of detection ccurve in Figure 11.10 is applicable. 2 for the case that mm, 2 = 8mm, 05.4 4, Select three inspection intervals, $00, 1000, and 1500 hours. Determine the ‘curnulative probability of detection foreach of the cases of Exercise 2 using the ‘upper curve of Figure 11,10; then estimate the required inspection interval for a cumulative probability of detection of 95%. Compare the results with those ‘obtained in Exercise 2 5, Assuming the crack propagation curves of Figure 11.7, determine a proof test interval. Acsume that K, = SOMPa Jin, f= 1, and select proof test conditions that would eliminate ceacks larger than 15mm, What is the required proof stress? 3. Repeat Exes 6.A lnrge component made of a material with Ky = 30ksi /fi_and ZOD! is subjected to service strc of 1ODks, Crack growth from (0.01 inch toa, hasbeen calculate to cover two years of operation given that o, = I50ksi The crack occurs ata ill. Determine a stipping depth assume f= | 7. Fracture ofa certain structure isassessed at S,. Rephacement of the would cost 5. No other costs are antcipat, Analysis costs are S, per hour. Load data are availabe, Extensive analysis including finite element evaluation would require h hours. A total T of these structures are anticipated to be in ‘operation. The probability of a fracture is estimated to be P. Which course of action would you recommend? part References 11] D. Broek, Feature contol by pesodi inspection with xed cumulative probably of etek Sirucal fale, produto am tecaeal rane. Rossmnith Fa pp. science Enterpies, Lid (1987) Relaityofnor-desirative bspetions SA-ALCIMME 16-6381 (1978) yo he detection of fiaws and of the Setermination of a sve, AGARDo- te] UGovencm, Paper presented a ICAF meeting, Toulouse (198) 15] D. Broek 2POCRE, Software ractatEsarch Ine (1985) CHAPTER 12 Damage tolerance substantiation ‘Scope Previous chapters dealt with analysis procedures, (Chapters 2-5), the weredients needed for the analysis (Chapters 7-10) and with the use of the results for fracture control (Chapter 11). This Chapter concentrates on the general scope of the analysis, its relationship to tests (verification and substan- mn). the assumptions and sources of error, and the éesign options for improvement of damage tolerance. In shor, it considers the analysis in the framework of damage tolerance provisions. ‘Damage tolerance analysis substantiation is governed by damage tolerance Fequirements if any are in effect. Indubitably. the damage tolerance requi ‘ments for commercial and military aircraft are the most widely enforced, and presently there is considerable experience with their use. Thus, a discussion in some detail ofthe aircraft requirements is certainly worthwhile even for readers not concerned with aircraft, as it will bring out the good and bad aspects of requirements in general. Although incidental rules may be in effect here and there. the only other requirements addressing damage tolerance directly are ‘embedded in the ASME boiler and pressure vessel code, These will be reviewed aswell. Requirements forthe use of arrester strakes in ships ate discussed in one Oo the examples in Chapter 14 ‘Compliance with requirements is an issue in this chapter, but the discussions concentrate on the damage tolerance substantiation in general In particular the fects ofthe assumptions on the accuracy of the analysis - and thus on fracture control and safety ~ will be discussed. Damage tolerance requirements may enforce certain assumptions, and so be of greater effect on accuracy than the analysis itself (Section 12.8) 12.2, Objectives ‘The objectives of damage tolerance provisions have been discussed at various ‘ ' 1 owe a t ‘32 == o Few 121. Th aging pele, (2 Resi strength curve: by Crack grat cunse phices in this book. They are briefly summarized here, in the context of tolerance requirements. ‘A new structure can mnaxinrumn expected service fond, because of safety f higher than the Toads or allowable stain the design load, whieh ors stresses. The probability of occurrence of the design load is small, but finite for ‘many structures, so that the probability of failure is ot zero. When cracks are ‘present, the strength is less than the design strength so that fracture may occur