Small Volume Fabrication Cost Est Models PDF
Small Volume Fabrication Cost Est Models PDF
Small Volume Fabrication Cost Est Models PDF
Manufactaring cost estimation is un integral part Researchers gathered cost data in five different companies, as
of design for manafucturing. Cost estimation mod- documented by Mareet, Schuster2, Schreve3 and De Swardta,
from which the models were developed. The cost estimation
els are presented for fabricated mild steel welded
models are aimed specifically for use in the early phases of
assemblies, between a few hundred grams and Iive embodiment design, with the corresponding limits to informa-
ton, mude in small production volumes. Duta col- tion available and accuracy achievable. The design scenarios
lected in time studies at five South African compa- encountered in a small production volume environment often
nies was cast into a times and rates approach, us- includes that the potential manufacturers of a partor assembly
are only known after completing the detailed designs and
ing robast statistical techniques. The models nre
submitting them for quotations to manufacturers. Information
aimed at the eurly embodiment design phase, to help relying on the knowledge of a specific manufacturer could
the designer to optimise the design fo, fabrication therefore not be used in the cost models.
and to compore fabrication with alternative pFoc- The role of cost estimation and design for manufacturing in
esses. The application of the models to evaluate the the design process is considered in the following section. The
approach used in developing the cost models presented here,
redesign of a fabricated assembly is demonstrated.
and the resulting models are then described. The use ofthe cost
NOMENCLATURE models in comparing alternative designs is finally demon-
strated.
Amaterial S heet materi al ar ea[mm2]
Aoor, [-*']
Proj e cte d p art ar ea Manufacturing Cost Estimation in the
A*"u Cross section al areaof weld seam [--t] Design Process
B rheet Is the part a sheet fboolean]
Design for manufacturing starts with the selection of the
ca Critical alignment with external feature [bool
manufacturing processes most suitable for a part and then
ean]
adapting the part's design to those processes. Selection and
F Flexibility [boolean]
adaptation recurs in the design process, at increasing levels of
Lb Bendlength [mm]
refinement and detail. Cost estimation is an integral partofthis
L"r, Length of cut [mm]
process, even if only implicitly. In addition to the cost, other
Lj Length ofweldingj oint [mm]
factors not addressed in this paper also play significant roles,
ffihand I formanual handlirg, 0 formechanical handling
e.g. material constraints imposed by functional requirements,
ffip*t Part mass [kg]
time considerations, companypolicy (such as preference for in-
nangles Number of different angles per part
house manufacture or outsourcing), etc. Process selection
nb Batch size
therefore often involves finding a suitable compromise, in
fthol", Number of holes per part
which cost plays an important role.
frnozrles Number of nozzles on a flame cuffing machine
Pahl and Beitzs divide the design process into four basic
nnstd-angles Number of non-standard angles per part
stages as shown in Figure I . The overall manufacturing process
ns Number of parts per sheet
selection usually has to be done just before or during the
nstd-angles Number of standard angles per part
embodiment design stage, preferably in a concurrent engineer-
nw Number of welding j oints
ing context. Manufacturing process choices mad e atthis stage
,S Size[mm]
t Time per part [s]
t, Critical sheet thickness parameter: I if sheet
is thickerthan 30mm, 0 othenvise.
tld Alignment feature's principal dimension [mm]
t* Material thickness [mm]
e Symmetryangle ["]
eb Bend angle [radians]
lntroduction
The CAD Laboratory at Stellenbosch University developed
manufacturing cost estimation models for mild steel welded
assemblies in small production volumes in South Africa.
are expensive to change later, often requiring subs tantialeffort to Eskilandert. The data was gathered by Mareet, Schustel,
in redoing detail designs ofparts and assemblies, and may even Schreve3 and De Sw ardta, through time studies using a video
extend to reworking process and production planning. It is camera or stopwatch. It was explained to the artisans that the
therefore extremely importantthat sound process choices be datawould be used for academic pu{poses. Still, the taking of
made as early as possible. times certainly affected the productivity of the artisans, espe-
Cost estimation during the early design stages is compli- cially when doing video recordings. For one batch of welded
cated by the lack of product information available. Typically, assemblies, the last assembly was recorded with video camera
afterthe concept development stage and during the preliminary and the welders worked notably faster- maybe to show offtheir
embodiment design (or layout design) only rough sketches skill! Unfortunately this can distort the cost models. However,
with the principal dimensions are available. In spite of these the data was gathered over long periods and at four different
challengos, the role ofmanufacturing cost estimation within the facilities. It is the authors' opinion that the sheer amount and
design for manufacturing (DFM) paradigm is well recognised. diversity of the data countered the effect that the measuring of
Leibl et al.6 showed that designers not using a cost estimation times had on the productivity. Therefore, the authors conclude
tool during the early design phases, created products that are that any effect that the recording of times had on the produc-
up to 80% more expensive than the design oftheir counterpart tivity of the artisans does not significantly decrease the accu-
who used a cost estimation tool. They also observed that the racy of the cost models, particularly when the wide variability
designer not using a cost estimation tool took 40-50o/o longer in measured costs is taken into account.
than their counterparts using the cost estimation tool. This Due to the nature of a small production environment, with
clearly indicates that cost estimation during the early phases of much manual labour, the recorded times are very variable. There
design can help to eliminate infeasible concepts quickly, so that often were a few parts in a batch that took significantly longer
designers can focus on the more cost effective concepts. to manufacture than the rest of the parts. During the assembly
A very important aspect ofcost estimation is that it must give of the parts this was often caused by parts that did not fit
the designer feedback to optimise a part' s design for particular properly in the assembly. These times tend to pull the average
processes (Feng et al.1 and Eskilander8). The designer must time ofthe batch unreasonably high. This presented difficulties
know what the cost drivers are, which features of the design in the analysis of the dataand development of the cost models.
carries the most cost and what, if anything, can be done to After consultation with an expert statistician, it was decided to
reduce the manufacturing cost. This information, combined use median values rather than average values and a robust
with the appropriate design rules, must guide the designer to regression statistical method (Hoaglin'') rather than least
minimise the manufacturing cost and meet the cost target. squares regression. In this way the cost model does not
Despite the convincing arguments for advantages of cost penalise the design for bad production practice.
estimation in the design phase, there is a remarkable lack of In the cost models presented here, reasonably efficient
published data and models. Some of the studies relevant to production is assumed. It is the authors' opinion that a design
fabrication processes that were published are the following: must not be penalised for inefficient production settings. A
Boothroyd, et al.e published cost estimation models for various design should however be penalised if some features cause
high-volume processes, but with little attention to fabrication inefficient production. One ofthe reasons for doing a thorough
processes. Esawi and Ashby'o developed a system that helps cost estimation is to identify such aspects of a design. The
designers to eliminate infeasible manufacturing processes and researchers therefore used their judgement to discern situations
it also prevents them from overlooking potential processes, but where cost increases could be attributed to poor production
works at a high level ofabstraction. Farkas and Jilrmairr describe practices and to lessen these effects on the cost models.
a cost model for welding that they use for product optimisation, Preliminary cost models were published by Maree and
but they ignore the other fabrication processes such as bend- Basson'*, schreve et al.'s, and Basson and De Swardtru. The
ing, cutting, etc. Maropoulos et al.12 describe a system for dataand models from this research were combined into one set
fabrication process planning in the early stages of design, but of models and presented here. The most challenging part of
they do not describe the model used. developing these new models is the integration ofDe Swardt'sa
Publications often do not state what production environment models with the rest of the data. He did his time studies on the
their models are aimed tt, even though cost estimation is manufacture ofheavy earth moving equipment, typically dragline
strongly production context sensitive. For example, a cost buckets with acapacity ofup to 168 m'. The rest ofthe datawas
model developed for a mass production environment cannot be collected for assemblies weighing less than 200 kg.
used for products manufactured in job shop or small lot size
envlronment. Cost Models
The value ofcost estimation models during early embodiment Models for flame cutting, bending, tack welding and welding
design is therefore clearly established, but no models applicable assembly are presented here. As indicated above, a time and
to the small production volume fabrication environment were rates approach was used. The cost models therefore relate the
found in the literature. This lead to the development of the manufacturing time to cost drivers, or manufacturing features,
models presented here. on the parts. The estimated times have to be multiplied by
suitable respective rates to give cost estimates. Since the cost
Gost Model Development Approach models are aimed at small production volumes, set-up times (i.e.
The models presented here follow a times and rates approach costs incurred once per batch) play a significant role and are
to cost estimation. This method is intuitive to develop and given separately.
interpret and it is a very popular approach in industry according
Flame Cutting
Flame cutting is the operation where parts are cut from sheet
(4)
metal using a blowtorch. The material is mounted on the
machine bed and not moved until all the parts are cut. All the data
gathered for flame cutting operations integrated very well. The unloading time is the time it takes to remove all the parts
In the model presented here, dvariabl e ffihn,dis used. This is and scrap from the machine bed. For the large parts, consider-
simply a boolean variable indicating whether the part is handled able time is spent to remove scrap material from the bed, since
manually or with a lifting aid such as an overhead crane. this is also done with a mechanical aid. The unloading time is:
According to Corlett and ClarckrT parts weighing less than 2lkg
can readily be handled manually. Note also that all the models t : 13 .9m,,o,,t+ 165+ 98( | * nro,",)]( I -ffi no,) (s)
for flame cutting operations are valid for mild steel sheets up to
200mmthick. A grinding operation normally follows flame cutting. De
The set up time includes the time to prepare the machine for Swardta provides a separate model for this.
operation and clean it afterwards. The set up time per part is:
Bending
The principal difference between the models ofDe Swardta and
t = L[os8(l - ffihnn,t)+ aaomnonal (1)
Schreve', is that the former models were developed for a 1250
nb ton press, bending material between 6 and 50 mm thick, while the
latter models are for a4tonpress bending material no more than
The time to load the material is a function of the number of 6 mm thick. By integrating the models, handling operations with
parts that fit in one sheet and the area of the raw material. The mechanical aid is now included. The cycle time forthe two press
raw material area is normallynot determinedbythe designer, but brakes differ significantly. The gap is bridged by incorporating
by the production planner. However, sheet metal normally data published by press brake manufacturers. They provide
comes in standard sizes and it should therefore not be difficult approach speed, bending speed, return speed, stroke length,
to make a reasonable estimate ofthe material area. Note that the motor power and bending force for each machine. The bending
time must be reduced not by selecting smaller sheets, which will force required to make a specific bend can be calculated in terms
result in more handling time, but by designing parts for the of the length of the bend, the material thickness and the bend
optimum nesting, since the time is divided by the number of angle. Although bending angle is included in this equation, it
parts per sheet. The loading time is: must be remembered that bending is done in standard dies, so
the designer can minimise the angle, but should keep to the
standard angles. Schreve3 did observe the bending of non-
t = Lfz.sz(t - m hanct)* (r 5.27 A,,,o,o,io, *3 standard angles. In these instances, the stroke length of the
n"
-59)m ha,ct] cl press was changed so that it did not press to the bottom of the
die. In these cases the set up time is longer, since the adjustment
For each cut made with the flame cutter, the flame must first of the stroke length is a trial and effor process.
pierce the material before it can start cutting. This is called The equation for the set up time includes the time to do a die
piercing time. Obviously the designer must try to keep the change (1467 seconds). If all the bends on a partis for the same
number of cuts a small as possible. One piercing operation is angle, the die is not changed. This quantifies the design rule that
done for the outside profile of a partand one per hole. This time the number of different bend angles must be minimised. The
and the cutting time both increase with the material thickness. additional terms in the set up time equation are for operations
Some flame cutters used in the factories where the case studies such as setting the stops and preparing the machine. This is
were done, have four nozzles, therefore the piercing and cutting added per bend. The set up time rs:
time per part decreases if two or more parts can be cut
simultaneously. However, this is a production parameter over
which the designer may not have control or knowledge of. It is ,:L( .1467 1565 n _angl", +g3
- rft,,,d
nsla _u,",""
therefore recommended that, unless other information is n, n ouglu,
stu
stct
-urtgtes -ungt"t ] (6)
I )
available, the designer must assume that the parts are cut one
at a time, thus assuming that h,o,"r",:l . The piercing time is:
The loading and unloading time must be added per bend that
is made. Schreve' found that for parts handled manually, the
r%(t .v ot *+1 4)+1 tt
"(t+n or^) (3)
time depends mostly on the part's projected area. The loading
fl nuul", and unloading time is:
The equation for cutting time is a least squares regression t - l(25 .8Apo,t+ 5. S)mnona+ 98 .5(* oon - I )] (ftrta_onsus* frnstd_and"r)
Note that the bending operation times are only valid for Parameter Value
sheets up to 50mm thick.
a1 0.0003
Tack Weld Assembly
a2 -0.1716
A new model for tack weld assembly had to be derived from the
datacollected by De Swardta and Schreve's3. Note that Schreve's as 0.0043
model already incorporated the data of Mareer and Schustef.
a4 0.1358
Simple integration of their models is not possible, since the
models use different design parameters for cost estimation. The cr-c -6.3877
parameters are summarised in Table l.
The integration led to the regression formula given in equa- a6 0.0006
tion (9). a7 223.8
1
a concept process plan, suitable for manufacturability assess-
1 10 100 1000 10000 ment during early phases of embodiment design). The input
Measured Assembly Time [s] parameters for the cost estimation models are calculated and
entered by hand. This program was used to do the case study
Figure 2: Comparison of Assembly Time Estimate and Measured Times reported here. The program uses the simplified assembly model
(Note that logarithmic scales are used on both axes) presented in equation (10).
80000
70000
60000
50000
o
o 40000
E
i=
30000
20000
1 0000
0
Original Redesign
Gonclusions
In order to validate the cost estimation models presented In this paper cost estimation rnodels, using a times and rates
here, the fabrication time oftwo side loader assemblies (Figure approach, are presented for fabricated assernblies made in small
3) are compared with the tirne estimated using the new cost production volumes. The models are intended for use in the
rnodels. The first side loader consists of 99 parts and weights early embodiment design phase and therefore only parameters
l30kg. Mareer redesigned the side loader and applied the readily available to a designer at this stage are used in the
Design for Assernbly rules. The redesigned side loader has 4l rnodels. The rnodels can help the designer to optimise the
parts and weights 121kg. He also ffleasured the fabrication times design for fabrication cost since it links the cost to the design
for both assemblies. parameters. It can be used to help the designer in selecting an
The total predicted fabrication times are colnpared with the appropriate manufacturing process if the cost estimated with
ffreasured times in Figure 4. The times are over estimated by 28% these models can be compared to the manufacturing cost of
and 7o/o respectively. It is also very interesting to note that the alternative processes. The fabrication time of two assemblies
reduction in fabrication time observed for the redesigned side were estimated with accuracy colnmensurate with a small pro-
loader is proportional to the reduction in the number ofparts of duction volume environment. This result was achieved with a
the assembly. very simplified assernbly model, which can be applied to parts
It took 4h27min to do the cost estimate ofthe original design weighing between a few hundred grams and five ton.
and t h 4Ornin for the redesign. It took less time to estimate the
cost of the redesigned side loader because it consists of fewer Acknowledgements
parts and because solne of the parts are common to both The authors wish to acknowledge the financial support of the
assemblies and thus it was not necessary to repeat their Foundation for Research Development and the Stellenbosch
estimates. Most of the time goes into calculating the input University Research Committee, as well as the invaluable con-
parameters, such as the mass of the part, section areas, etc. tribution of the various industries that gave access to their
Many of these parameters will be readily available if CAD manufacturing facilities and support during the data gathering.
drawings ofthe parts were available - which may not necessar-
ily be the case during embodiment design. This was not the case References
here. Also, ifthe cost estimation software were linked to a CAD 1. Maree WG. Design -fo, Fabrication: A Cost Estimation
system, the datainput tirne can be reduced significantly. Liu and Modelfor Small Production Volumes. Masters Thesis, Univer-
Bassonre did such an implementation. sity of Stellenbosch, South Africa, I997.
Although the side loaders were assembled using SMAW 2.Schuster HR. A Computer Aidfor Specrfication Develop-
and MIG welding, the assembly model still gives reasonable ment, Conceptual Design and Manufacturing Cost Estima-
results. The assembly model was developed from datafor MIG, tion in Mechanical Design. Ph.D. Thesis, (Jniversity of
SMAW and FCAW. In this instance it was a safe simplification. Stellenbosch, South Africa, 1997.
However, the validity of this must be investigated for a large 3. Schreve K. Cost Estimating Welded Assemblies Produced
sample of case studies. The reality is that the designer may not in Batches. Masters Thesis, (JniversiQ of Stellenbosch, South
know which process will be used and will therefore anyway Africa, I998.
have to make an assumption. A case study reported by Farkas 4. De Swardt AV. Designfo, Manufacture Through Fabrica-