Anticlastic Action of Flat Sheets Analytical expressions for the stresses in bent elastic strips are obtained and found to agree very ‘well with experimental results by H. D. Conway and W. E. Nickola Austmact—A theoretical analysis is made of the stresses induced in thin elastic strips which are bent (by uniform bending moments) into ares with substantially constant radii of curvatures, "Stresses determined experimentally fon aluminum snd steel strips agree very well with the theory. List of Symbols e = strain A= strip thieteness steip width. strip deflection coordinate axis in direction of strip length ‘coordinate axis in dirvetion of strip width ‘coordinate axis in direction of strip thickness Aexural rigidity of strip Ei'/120 — »® ‘Young's modulus Poisson's ratio moment of inertia of strip evoss section bending moment radius of curvature span of teip a2 (Enya RD) Ba — yr) Introduction ‘Thin elastic strips or tapes are commonly used in many industrial applications, notably in computers. When such a strip is bent, the centra lportion of the ‘cross section usually remains essentially flat, whereas the edges curl up and away from the center of curvature forming an anticlastic surface. Since the stresses vary across the width of the strip, the ‘Bernoulli-Buler theory does not accurately predict ‘the stress. Tn this paper the analytical expressions for the stresses are obtained and compared with experimental results. "The present problem has been the subject of a number of investigations in the past, a substantially complete analysis having been given as early as 1891 by Lamb.’ That given here discusses the analogy with the axisymmetrical bending of a thin cylindrical shell and hence uniform beam on elastic foundation. 15 ms Pde of Mean Cm Ue: al ‘Rdced Mohn Ass Dipl 104i Bip Bias RE Papen rma 168 ESA Aunt Moving ald Calan Reprinted from EXPERIMENTAL in Bending Analysis Figure I shows a thin strip of thickness h and width L subjected to moments M and »M per unit length of the relevant sides. ‘There are no stresses ¢, normal to the larger faces of the strip and, since it is, ‘thin, it is proper to assume that there are noo, stresses anywhere in the strip. If we assume that the radius of curvature is R [Fig. 1(a)], it is easy to show that the direct strain e, at a height z above the middle surface is equal to z/R. This expression can be derived if we assume that the plane cross sections of the strip remain plane. If the edge tractions are properly distributed, this expression is an exact Solution of the equations of elasticity. It therefore Zz x} me ‘ee \ | “1 \o \ | | \I | \ _ = (a) Zz uM om G eS) o ta | ve to) Fig. 1Coordinates and geomety of a thin strip loaded on its tree edges with moments M and »M Mrcuaxtes, April 1905follows that the stress-strain equation in the x-di- rection is Ee ane ® If the strain e, in a direction at right-angles to the direction is zero, it follows that Be, = 6, ~ 90. = 0 ‘and henee on tee ® ‘The stresses c- are induced by the moments M since cn ® and the stresses 7) = ve, are indueed by the mo- ments »Mf which act as shown in Fig. 1b. Substitut- ing eq (2) into eq (1) gives Eo: ee es ae’ Ee where D is the flexural rigidity of the plate. Provided that the moments Mand »M per unit length are applied in a Yinear manner as dictated by eq (4), the strip bends into a cylindrical surface as shown in Fig. 1(a) and the cross section remains flat and is not deformed, Generally speaking, the problem of practical interest is the one in which only moments M are applied to the strip which, in the bent state, is actually a portion of a cylindrical shell. ‘Therefore, if we assume that the method of superposition is applicable, the problem of a cylindrical shell of radius R subjected to moments »M per unit length (on its circular edges) must be investigated. Natu- rally, the deformations induced by these moments are usually confined to the edges and are found to de- crease rapidly away from the edges. ‘This is the reason why the longer edges of a strip bend upwards and away from the center of curvature whereas the central region remains essentially flat. ‘This is true provided that the width of the strip is sufficiently large or radius R sufficiently small so that there will be no interacting edge effects. It is well known? that the radial deflections of axially symmetrically loaded cizeular shells are governed by the differential equation ew oa + aw = 6 where 48° = BR/RED ~ 20 — 8) /ReH A general solution of eq (6) may be written in the form of w = Gysin ay sinh fy + Ce in ay cosh fy + sens ay sinh By + Cy 0s By cosh ay (7) Since the problem is symmetrical in the coordinate y of Fig. 1, C; = C; = O and C, and C, may be obtained from the boundary eonditions ew . De = Maty = 31/2 tw gay = 412 Gis = Oaty = 41/2 ® ‘The circumferential bending moment per unit length M, is then determined and is given by 2M My = 9D sam ag WA + B08 a cosh sy — (A = B) sin ay sinh y] (9) where A = cos asinh a, B= sin « cosh a, a ~ sl/2 (10) Tn addition, the circumferential membrane force is sven by Ehw heat No ~ “R ~ RDAteinh a + sim a) [CA — B) cos ay cosh sy + (A+B) sin ay sin ay in) ‘The resultant circumferential stresses in the bent. strip can now be computed. These consist of: (@) bending stresses 6M /h* which are tensile on the convex surface and compressive on the concave surface, () bending stresses GM,/h*—which are ‘compressive on the convex surface and tensile on the concave surface, and (c) membrane tensile stresses Ng/i. ‘The expressions for M, and N, are given in eqs (8) and (11) respectively. “When these equations are reduced, the resultant circumforential tensile stress on the convex face of the strip may be written in the form of a2) where — 2x Bem oy cea By) sinh sl + sin al a3) and A and B are given in eq (10). Hence at the edges of the strip where y = 41/2. inh 20 — sin 2a) 9 Goinai-Fan an)? = et 08 ‘The circumferential compressive stress ¢, on the con- cave face of the strip ean also be written in a form similar to that of eq (12). ‘The corresponding values of K are obtained from eqs (18) and (14) by making the last term negative. Graphs showing the variation of the stress factors K across the tensile and compressive faces of the strip are given in Figs. 2 and 3, respectively. ‘The membrane stress at any given“05-04 -03-02-0) 0 Of 02 03 O4 05 TRANSVERSE POSITION (4) Fig. 2—Varation of tensilestress factor ‘across width of strip. yh/6M POISSON'S. RATIO“03 +. | “05 Gt 03-62-01 6 Gt G2 3 04 a5 TRANSVERSE. POSITION Yj] Fig. 3—Variation of compressive-stress factor KS ach¥/6M across width of stip location adds algebraically to the bending stresses, and thereby causes a difference between the value of K on the tensile and compressive faces. Experimental Investigation Specimens and Test Fixtures Equations (12), (18) and (14) indicate that the Jongitudinal stress at any given transverse position is dependent upon the parameter a which is a function of the cross-sectional dimensions and radius of curvature of the specimen. Since it was desirable ‘that the experimental study cover a range of values {for ay, two specimens were prepared for investigation. Roth sides of each specimen were instrumented with metal-foilelectrical-resistancestrain gages. Figure 4 shows the specimons and strain-gage locations (A, B, C, BY, AY). ‘The two gages which were bonded at each location were oriented slong the longitudinal (x) and transverse (y) axes of the specimen. Poisson's ratio and Young's modulus were deter- see locations. (A) 2024-74 aluminum; € = 105 pai, » = 032. (B) AISI 65.80 carbon ste; E = 225 pal: + = 023 mined from tensile tests conducted on specimens obtained from the same sheet stock used to prepare the specimens. ‘The mechanical properties in the longitudinal and transverse directions differed by only a few percent for each specimen. Figure 5 shows the aluminum specimen (plate) mounted in the test fixture and subjected to pure ending moments, Each end of the specimen was clamped to a flat machined surface at the center of each supporting bar. A ball bearing at each end of the bar contacts a smooth, hard, horizontal support- ing track, Pure couples were applied by bolting a 44in. d am disk to ench end of the bars and applying dead woights to wires pinned to the circumference of the disks. ‘The ball bearings permitted free rotation ‘ag well as free horizontal motion at: the ends of the specimen, thereby satisfying the requirments of a simple support. ‘Restraint of the transverse edges of the specimen, introduced by clamping these edges to the bars, was Jound to extend only a relatively short distance from_ the clamp. This was demonstrated by loading a photoelastic model in the tost fixture and observing the stress pattern. A reflective polariscope was arranged so that the model was viewod from the topFig, 52—Aluminum ‘specimen mounted in test fixture Fig. Sb—Test fixture PLATE ~ Bale BEARING vise: "% ~sueponrme reAck (convex) surface. ‘The bottom surface was made reffective by spraying it with a silver enamel paint. It can be shown that the photoelastic signal is proportional to the membrane stresses developed by ‘the anticlastic action at the edges of the model. ‘Figure 6 shows the half-order fringe observed near clamped edge. It can be seen that there is dis- tortion only in the immediate vicinity of the clamp. "Thus, the effects of clamping decrease rapidly with distance from the clamp; therefore, the strain-gage readings are unaffected by clamping. Results Figures 7 and 8 show plots of the average experi- mental strains vs. bending-moment relations ob- tained for the top and bottom surfaces of the aluminum specimen. Measurements at locations 4 and A’ agreed closely with one another as did measurements at B and B’. Similar data were obtained from tests on the steel specimen but these data have not heen included in this report. Analytical considerations show that the parameter a is related to specimen geometry and radius of curvature by a= laa - 7 e/arayy ‘The radius of curvature R was determined experi- mentally by measuring the mid-span deflection 3 and determining R from the relation R- (a where is the free span of the specimen, Once the value of F is dotermined the value of a is calculated, by using the first of these two relations. ‘Tho experimental values for the constants K in eq (12) were obtained by first calculating the longi- ‘tudinal bending stross ¢, at each strain-gage location. ‘From Hooke's Law: measured longitudinal strain measured transverse strain Substitution of #, into eq (12) gives the experimental values of K: BN = 9, teed -— eae K ‘Table 1 compares the experimental values of with the predicted values given by eq (13).Fringe Produced By Curling At Edges ~~ Fig. 6—Photoelastc pattem showing transverse restsaint effect of end clamp 400 ‘000 8 600 3 | & 200 a a | RS, Creagense 5 222 | ovemnowan reais : Sra =i 8-600 2 § 1200 coun || ‘400 Fim _| #0 0 0 2 x APPLIED BENDING MOMENT iN. LBS: Fic. 7—Exparimental strains on ton surface of aluminum plate vs, applied bending moment (Strain gages located on lop of specimen) Each experimental value of K in Table 1 was obtained by calculating an average value for both applied positive and negative moments of equal magnitude. The sign convention is indicated in Figs. 7 and 8. TABLE 1—COMPARISON OF EXPERIMENTAL AND PREDICTED VALUES OF K [SEE EQS (12) ANO (13) Location Sursce & Convex Concave B Convex Concave © Convex Concave a=57 a=0.%5 Al. Spec. Steel Spec. A. Spec. ——wocarov a | tocarowe -600 MEASURED STRAM HL W/W 1000. “1400 APPLIED BENDING MOWENT- MCS, Fig. 8Experimental strains on bottom surface of aluminum plate vs. applied benging moment. (Gtrain gages located on bottom of specimen) cussion and Conclusions ‘The small nonlinearities near the zero load points of the experimental curves are attributed, at least in part, to initial curvatures existing in the specimen at the nosload condition. ‘These do not manifest themselves at higher load values. Referring to Table 1, it can be sven that generally there is very good agreement between the theoretical and experimental results. Even though the elastic constants # and » for the plate materials are only known to be accurate within a range of 2 ta 3 percent, the trends of the experimental and theoretical re- sults are the same. For example, referring to the convex side of the steel specimen (x = 0.95) and proceeding from the edge to the center of the strip, both experimental and predicted results decrease whereas on the concave side they both increase. ‘The other theoretical and experimental tronds are in general agreement. It can also be sen that the theoretical and experimental maximum resultant stresses are near the edges of tho convex fave. In conclusion, the tests seem to support the theoretical results given by eqs (12) and (13), and these equations may be used with confidence. ‘It is seen that the flect of anticlastic action on the usual stress caleu- lations is not great. In many cases it will be the distortion of the strip which is important and it is clear that eq (7) may also be used with confidence. Achnowledgments ‘The authors are grateful to D. W. Skinner, 8. J. Skocz, and J. W. MacBeth for their assistance during the experimental portion of this investigation. References lg Re 0h he Flare of «Flt Blac Sorin Oe Timackeno and Warrowsy Krier, "Them of Plates and Ska ‘MeGreHtl Book En Ine Chapler 13,458 (1838) Prieapnicat