See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/292871943

Early cost estimation for injection molded components

Article · January 2000

CITATIONS READS
31 1,769

2 authors, including:

David Kazmer
University of Massachusetts Lowell
289 PUBLICATIONS   1,671 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Model-Based Polymer Processing View project

2018 Polymer Processing Society Meeting (Boston, MA) View project

All content following this page was uploaded by David Kazmer on 19 September 2017.

The user has requested enhancement of the downloaded file.

 Early Cost Estimation for Injection Molded Components
ADEKUNLEA. FAGADE and DAVID 0. KAZMER

Mechanical and Industrial Engineering Department

University of Massachusetts Amherst
Engineering Laboratory Building
Amherst, MA 01003

Cost and performance estimation are frequently used at early stages of product
development to determine feasibility and drive critical design decisions. Early cost
estimation has been hampered by the unavailability and uncertainty of informa-
tion. This paper derives cost estimates from a complexity metric as defined by the
number of dimensions that uniquely define the part geometry. Dimensionality and
other critical design variables can be automatically assessed within modern com-
puter aided design systems throughout the product development process to provide
continual feedback regarding tooling, process, and material costs. The complexity-
based models were developed and tested with empirical data for thirty injection
molded parts from different suppliers and was found to have a highly significant
correlation with mold costs and tooling lead-times. Models for estimating material
and processing costs and yield at the early stages of design are also developed. The
developed methods enable real time evaluation of the effects of a product design
on its tooling cost, tooling lead time, processing costs, and yield at the early stages
of design.

INTRODUCTION applications. Tooling lead-times are also very important

he injection molding process is increasingly being factors in today's very competitive market environments.
T used in the manufacture of complex net shaped
parts of industrial and domestic electronic and electri-
The early stages of product design provide a good
opportunity for the optimization of these factors. How-
cal appliances. Designers are taking advantage of im- ever, in industry cost estimations are usually done
provements in the process capability and engineering when the design is well detailed, and by departments
materials by consolidating multiple parts and functions that are extemal to design. At this late stage of de-
into complex parts. One of the most frequently used set sign, there is considerable inertia against any drastic
of guidelines for parts consolidation is the Design for changes. Tools for evaluating altemative design config-
Manufacture and Assembly (DFMA) guidelines devel- urations for costs and tooling lead-times, from the com-
oped by Boothroyd and Dewhurst (1). One significant puter-aided product data, will greatly facilitate optimi-
benefit of DFMA is the considerable savings in assem- zation of component consolidation.
bly cost from fewer parts that need to be assembled.
Other inherent benefits of DFMA are the encourage- RESEARCH OBJECTIVES
ment of teamwork between design and manufacturing,
and the improved product reliability from the reduced The research vision is to develop real-time design
probability of system failure because of component evaluation techniques that are available at very early
failure. stages of design. Advanced analysis techniques have
In spite of expected savings in assembly costs, com- been developed to provide many estimates of design
plex parts have longer tooling lead-times, higher tool- performance. Typical types of analyses used in molded
ing costs, and potentially lower production yields. part design include structural (stiffness, impact. creep,
Hence, the net benefits of parts consolidation may be fatigue), manufacturability (pressure distribution, cooling,
uncertain. However, the effects of parts consolidation shrinkage, fiber orientation), and economic (amortized
on mold tooling cost and lead-time have not previously tooling cost. material costs, machine costs). However,
been quantified. The amortized tooling costs of injec- these numerical simulations may require complex
tion molded parts constitute a very significant portion meshes and boundary conditions to be built on top of
of their manufacturing costs, especially for technical detailed geometry. As such, advanced analyses tend

JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3 97

 Adekunle A. Fagade and David 0. Kazmer

to be performed at the end of the design cycle, after MOLD COST ESTIMATION
the majority of critical design decisions have been
Related Research
completed.
Similarly, manufacturing cost estimates are made Two methods published in injection molding plastic
after design detailing. The detailed cost estimate is design literature that address the problem of estimating
normally evaluated by an experienced cost estimator mold tooling cost at the design stage are the method
and results in a binding quote. The early cost estimate, of Dixon and Poll (2) and the method of Boothroyd and
however, is essential for the economic evaluations of Dewhurst (1). The two methods agree that a part's
design alternatives. If the early cost estimate includes geometric complexity is a significant contributor to its
cost factors that can be controlled by the designer in tooling cost. However, they evaluate part complexity
the early development phases, then the designer would differently.
receive valuable early feedback towards an optimum The Dixon and Poll method estimates the relative
design. Thus, the cost models developed in this re- tooling, material, and processing costs of an injection
search are to be used for real time cost analysis at the molded part from look-up tables. These costs are esti-
early stages of design. Another objective of this re- mated relative to the cost of tooling for a simple refer-
search is the development of guidelines for optimum ence part. The reference part used is a flat disc with
consolidation of multiple components of a product into outside diameter of 72 mm and inside diameter of 60
fewer but more complex components. The component mm. The approximate tooling cost for this reference
consolidation can be subject to different desired objec- part, based on 1991-92 costs, is $7000 and includes
tive functions such as minimization of tooling costs or about $1000 in die material costs. Seven attributes
time-to-market, or maximization of profit. The injection that can be determined from the part at the design con-
molding of plastics has been chosen as the domain for figuration stage are used in evaluating a part's basic
this research because of the its usage in the manufac- complexity, Cb, from a look up table. These attributes
ture of complex net shaped parts. The procedure devel- classify the part by its size, shape, number of walls
oped here are however applicable to other net-shaped with undercuts etc. Two multipliers of Cb, are also eval-
and near net-shaped parts producing processes such uated from look-up tables. They are the subsidiary
as metal die casting, forging, and stamping. complexity factor C5 and the tooling and tolerance fac-
tor Ct. C5 is a function of the number of form features
in the part's cavity and number of undercuts. Once
PART COST ESTIMATION the design has been assessed, the relative mold con-
The cost drivers of manufacturing an injection molded struction cost, Cdc, and the total mold cost, Cmold• is
plastic part are expressed in Eq 1. The material cost evaluated as:
contribution, Cmat, is very significant, typically 50% to Cctc = CbCsCt
80% of the total part cost. Tooling and processing (3)
Cmold = 0.8Cctc + 0.2CMB'
costs are also significant cost drivers. The processing
cost, Cproc• is dependent on the hourly rate charged where CMB is the mold base cost. These estimates are
for the usage of the injection molding machine as well relative to the earlier mentioned reference part. It
as the processing yield, Yproc• which is the ratio of should be noted that the relative weighting between
good parts to the total number of parts produced. The construction and material costs may not be universally
tooling cost, Ctooz• is amortized over the estimated pro- correct. Moreover, the determination of subsidiary com-
duction quantity N for the life of the tool. plexity, C5 , requires the judgement of the estimator in
the classification of some cavity detail features as reg-
ular or irregular, and evaluation of undercut complex-
(l) ity as extensive or not extensive.
The Boothroyd and Dewhurst (B-D) method uses
empirically derived formulas and estimated manufac-
Equation 2 is an expression for the assembled product
turing parameters to estimate the times, t 1, for the dif-
cost. The m parts that constitute the product include
ferent tasks that are carried out in transforming a
both injection molded and standard purchased parts.
purchased mold base to a finished mold. The sum of
The cost of assembly is the product of the assembly
these times are then multiplied by an average shop
shop hourly rate, Ra,;sy• and the total time required to
rate, Rroc,1, to estimate the tool construction cost. The
assemble the m parts constituting the product. Thus the
mold base cost, CMB, is a function of the area of mold
assembly cost decreases as part-count m decreases.
base cavity plate and the combined thickness of the
The overhead cost per product Con includes both the
cavity and core plates. Mold tooling cost is then the
shop and the administrative overheads.
sum of mold-base cost and mold construction cost:
m m-1 n
cproduct = L c :xm + Rassy ~ t:xm + CoH (2) cmold = Rassy ~ t 1+ cMB, (4)
t=l t=l t=l

98 JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3

 Early Cost Estimation for Injection Molded Components

The B-D method calculates part complexity as a obsolete. The impractical alternative would be to con-
sum of inner and outer surfaces complexities. The stantly update the cost data for custom design fea-
surface complexities are estimated with an empirically tures. Problems of feature recognition or extraction
derived formula that sums the number of holes, de- from either blueprints or CAD data also arise. Identi-
pressions and surface patches. The generalization of fying and classifying all the geometrical features of a
all possible design features into the three categories part correctly from its blueprints or even from a phys-
limits the sensitivity of the B-D complexity index. In ical sample is not a trivial task. Problems such as
addition, the enumeration of surface patches is diffi- whether to classify a set of parallel protruding features
cult and uncertain for moderately complex designs that as ribs or grooves existed. Automatic feature recogni-
may have blending surfaces and many protruding rib tion has been reliably implemented only for a restrict-
features. ing set of feature profiles. Thus an alternative approach
Complex systems are known to consist of finite vari- to costing was sought.
ety of interacting elements. According to Scurcini (3)
the number, variety, types, and the organization of Proposed Approach
elementary components drive the complexity of a For the purpose of cost estimating, we surmised that
technological system. Since form and shape features the number of dimensions that are used to define a
constitute the basic components of a plastic part, an feature is a measure of its complexity. Difficulty in
enumeration of the features in a designed part could manufacturing the product will tend to increase as
be functionally related to its complexity. more dimensions that are required to define uniquely
In order to overcome the low sensitivity of the previ- define its features. Every dimension represents an
ous two methods of tooling cost estimation to changes additional point to check or a setup to make in the
in part complexity, our initial approach was to enu- manufacturing of the mold or the electrode that that
merate and assign a cost to every type of design fea- will be used to electric discharge machine the mold.
ture. The cost would be proportional to the difficulty of This reasoning is then logically extended to the total
reproducing the design feature in an injection molded number of dimensions required to completely define
part. These relative costs could be estimated through the parts' model. This information is readily available
literature review of standard machining times and the within constrained-based type modelers, which include
interview of mold makers. As shown in Fig. 1, injec- most 3-D modelers.
tion molding form features were classified to conform The total number of dimensions, D. is the number
to the Form Feature Information Model (FFIM) (4) es- of parameters required to unambiguously define the
tablished for the Standard for the Exchange of Product part. In the current work, these were enumerated by
Data (STEP). counting all the dimensions on all blueprints that ac-
However, since the designer has the freedom to de- companied the request for quotes (RFQ). All dimen-
fine application type features, as long as they fall sions in all views; elevations, sectional, and detail,
within the FFIM classification, cost estimation models were counted. When a view represents a repeated fea-
built on a fixed number of features are soon rendered ture the number of dimensions is multiplied by the

l FORM FEATURES I
EXPLICIT FEATURES
List of two dimensional IMPLICIT FEATURES
shape elements

I
I I I I I
DEPRESSION DEFORMATION PROTRUSION AREA PASSAGE TRANSITION
TYPE: TYPE: TYPE: FEATURE TYPE: FEATURE
·Groove • NA ·Boss TYPE: ·Hole TYPE:
• Hinge ·Disk • Thread ·Slot ·Edge blend
·Pocket ·Flange ·Marking ·Window • Corner blend
·Step • Gusset ·Gear teeth ·Others
·Others • Projection • Surface finish
• Rib ·Texturing
·Snap fit ·Others
• Tab
·Others

Fig. 1. A classljica.tiDn of injection ITWldingjeatures.

JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3 99

 Adekunle A. Fagade and David 0. Kazmer

R Z.O
TYP Z PL.AC£S

DETAIL A
TY P -4 Pl.AC£ S

st£ DCTAIL A

ALL SURF"AC£S TO
SPl B3 nNlSH

:OPCl. ~ SANPLE
~--------81~~--------~

Fig. 2. Sketch of a low complexity part.

number of times the feature is repeated. Usually, such makers in various parts of the U.S., Canada, Spain,
views, if labeled in accordance with ANSI Y14.5M di- and Taiwan supplied the mold quotes. It was the nor-
mensioning and tolerancing standard (5), show how mal practice of the custom injection molder to obtain
many times the feature is repeated by a number and quotes from three or more different mold makers in its
an X as in Fig. 2. Table 1 shows the procedure used in cost estimation process. The job was awarded to the
enumerating the dimensions of the part shown in Fig. 2. mold maker based on cost, lead-time, and past perfor-
mance. The size of the part, qualitative complexity,
CoDecUon of Empirical Data number of slides, gate type, surface finish, and ejection
A custom injection molder in western Massachusetts system are some of the factors that are considered in
assisted in this research. Original equipment manu- the estimation of a part's tooling cost.
facturers (OEM) submit requests for quotes (RFQ) to When the quote from a toolmaker falls outside a
this company. The company in tum sends out requests reasonable range estimated by the tooling engineer, it
for tooling quotes to moldmakers locally and overseas. could be due to one of three reasons. lf the mold quote
Seventy-five mold tooling quotes of single cavity molds is too low, the toolmaker may have failed to consider
for thirty of the parts that the company has quoted for the need for slides or other factors not apparent from
in the past three years were selected for analysis from the blueprint or CAD model. ln this case, the molder's
its records. The thirty parts vary in size from a small tool engineer tries to confirm that the toolmaker con-
reset-button with basic envelope size of 22 cc and 17 sidered all design specifications. If the quote is too
basic dimensions to a large sewage pump enclosure high, the moldmaker may be at capacity and would
with size 136,282 cc and 153 dimensions. only accept the job at a premium. Finally, the tooling
The origin of the seventy-five tooling quotes for the engineer at the molder could have misunderstood some
thirty parts were also geographically diverse. Mold design specifications.
There is often the post-design stage cross communi-
cation among the three parties: the design engineers
Table 1. Counting Dimensions of Gasket Disk.
of the product developer, the tooling engineer of the
Envelope Size: 160 x 160 x 5 mm3 128 cc molder, and the toolmakers. Engineering changes, usu-
Type and number of dimensions
ally minor, that may reduce tooling cost and/or facili-
Circular hole features 8 (1 X 8) tate molding, are suggested to the product designers
Angular spacing of holes 8 and are either accepted or rejected. However, there-
Dimensions of slots 8 (2 X 4) cent trend is towards simultaneous engineering among
Diameter of center hole 1 these three parties. This trend is facilitated by improved
Radial distance of holes 1
Ref. angle from center line 1
communications and CAD data protocols. Prototypes
Envelope dimensions 2 or preliminruy designs are being sent via the internet
Chamfer radii 16 (4 X 4) to injection molders and moldmakers for their imme-
diate feedback. This practice significantly reduces prod-
Total number of dimensions 45
uct development time and product cost.

100 JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3

 Early Cost Estimation for Iryection Molded Components

Mold Coat Drivers Table 2. Quotes and Attributes of Observed Parts.

The thirty parts, their mean mold quotes (MMQ), # Cmold ($K) Tmold (wk) S (cc) I D A
I HF HT
mean estimated tooling lead times (ML11. and their
geometrical attributes are as shown in Table 2. Only
1 67.27 15.5 27349 250 2 I y N

RFQs accompanied by blueprints that have adequate 2 27.50 14.0 327 64 0 N N

detailing for tooling were selected. Only three quotes 3 25.38 13.5 352 99 1 N N
that were much higher or much lower than the aver-
4 35.70 13.7 4199 181 0 y y
age quotes for the same part were discarded due to the
probability of over or under estimation, as mentioned 5 17.22 12.5 17 22 1 N N
previously.
Some significant mold tooling cost drivers such as 6 19.35 12.0 344 153 1 N N
part size, part complexity, number of walls with under- 7 38.00 15.0 675 108 2 N y
cuts, surface finish and tolerance level were identified y
8 20.10 12.5 450 53 0 N
through literature review, the industrial experience of
the authors, and interviews with mold makers. The 9 68.50 18.0 3334 141 3 N y
methods used for determining part complexity here 10 25486 y
63.89 18.5 289 1 N
differs from any previously published method. Prime
consideration were given to parts' attributes measur- 11 41.93 15.0 855 152 4 y y
able from its blueprints or CAD model such as size, 12 56.00 18.5 9997 495 0 N N
number of dimensions, part projected area, material
volume of part, number of critical-to-function dimen- 13 66.90 17.7 35928 286 1 y N
sions, and dimensional tolerances. Multiple regression 14 57.82 16.0 1371 172 4 y y
analyses were performed with the mean mold quotes
15 67.43 17.3 16453 372 1 y y
and mean lead-times as dependent variables and a
systematic combination of the other attributes as in- 16 143.86 21.0 108023 613 2 y y
dependent variables. Low correlations were found be-
17 47.97 16.5 14839 137 0 y y
tween the dependent variables and some independent
variables such as part material volume and part pro- 18 127.00 21.0 136282 153 2 N N
jected area which were thus omitted from Table 2. -
19 84.80 19.0 60853 337 0 N N
In Table 2, the envelope volume, S, measures the size
of the part in cubic centimeters. This is the volume of 20 31.00 12.0 1524 164 2 N N
a rectangular box that completely encloses the part y
21 29.90 12.0 2927 123 0 N
(Fig. 2). Even where a long projection is isolated, the
envelope volume still determines the size of the mold 22 22.70 11 .0 284 40 0 N N
base and to some extent the manufacturing work re- 23 14.90 11 .0 127 57 0 N N
quired to make the mold. The number of actuators, A,
is the total number of separate mechanisms that have 24 111.74 20.5 75821 28 1 N N
to be constructed into the mold to permit molding of 25 40.55 14.5 9176 31 0 N N
internal and external undercuts, and screw features
on the part. Undercut features that lie on the same 26 36.00 13.5 3722 101 0 N N
wall of the part and that are within 75 mm distance of 27 37.15 14.0 421 93 1 N y
each other are are assumed to require one slide mech-
28 45.47 14.5 2949 126 3 y N
anism. Evety screw feature is assumed to each require
a separate unscrewing mechanism. The parts with Y
(Yes) under the columns labelled HF and I-IT require
29 97.87 16.5 54919 46 \ 0 I y y

high polish finishes and tight plastic tolerances, re-

30 20.95
I 14.0 210 67 1 0 I y y

spectively. Parts with surface finish specifications of

SPI A1, A2, and A3 or that are textured on more than
25% of their entire surface areas are classified as hav-
tolerances, while those with greater values have nor-
ing high polish finishes. Parts with surface finish of
mal tolerances. The decision was guided by a table of
SPI B 1 or less on more than 75% of their surface areas,
dimensional tolerances allowed to mold makers (6).
are classified as having normal finishes. Plastic toler-
ances are specified as percentages of overall lengths.
Regression Reeultll
Owing to shrinkage characteristics of polymers, longer
parts are normally specified with larger tolerances. A In the summaty outputs of the regressions, the sam-
cut-off value of 0.07% of absolute percentage toler- ple coefficient of multiple determination, R2 , is the pro-
ance per unit length was used to classify the observed portion of the total variation in the dependent variable
parts as having tight or normal plastic tolerances. Parts that is explained by or accounted for by the regression
with absolute percentage tolerance per unit length model that is formed by the independent variables. R2
less or equal to 0.07% were classified as having tight can take on values between 0 and 1, where a better fit

JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3 101

 Adekunle A. Fagade and David 0. Kazmer

is obtained as R2 approaches 1. The regressions were preferable to develop and parallel-tool simple compo-
done at the 95% confidence level. The R 2 values ob- nents for automatic or manual assembly than to com-
tained with the mean mold quotes as the dependent bine components into a complex piece. The single
variable was greater than the value obtained with the complex tool may take longer to tool and may cost
individual seventy-five mold quotes. This is because more than the individual tools put together. Consoli-
the mean mold quotes provided a degree of central dation may later be done when demand is stable and
tendency towards the ~actual" mold costs. The result- new sets of tools are being ordered for large produc-
ing cost model derived using just size, 8, and number tion runs.
of dimensions, D, as the independent variables is: The models described can be easily developed by any
organization that has historical data on mold costs.
cmDid = 28300 + 0.818 + 45.6D The regression coefficients will differ with different
(5)
R2 = 0.869 data set but their proportion will be approximately the
Equation 5 shows that size and number of dimensions same. The accuracies of the models are higher than
explain 87% of the variation in mold cost of the sam- the accuracies of estimates from human cost estima-
pled parts. The intercept, 28,300, represents on the tors, that may vary within 50% of actual costs based
average the lower bound on the mold costs. Three other on Malstrom (7) as well as the empirical data from this
part attributes (number of actuators, A, high surface study.
finish, HF, and high tolerance, HT) can be included in The mold costs and lead-times estimated with Eqs 6
the regression analysis, with the latter two having only and 7 are plotted against observations in Fig. 3. Esti-
0, 1 states. The model now explains 91.1% of the vari- mates for aluminum molds for some of the thirty parts
ation in the mold costs: are also plotted. It can be observed that the mod-
els overestimate the tooling costs and lead-times for
Cmold = 22500 + 0.828 + 30D + 2940A aluminum molds, indicating the need to adjust the
+ 7630HF+ 5470HT (6) model coefficients down for aluminum molds. It is rec-
R 2 = 0.911 ommended that a chi-squared statistical test should
The mean tooling lead-time has a lower but still very be performed to check that the actual costs are not
significant R 2 value when regressed against size and significantly different from their estimates at a siginifi-
total number of dimensions (complexity), as shown in cance level of 10%, that is a/2 = 0.05. If it is different,
Eq 7. The imperfect correlation may be due to other a re-evaluation of the multiple regression coefficients
molder specific factors, such as maching availablity or using the new quotes should then be implemented.
willingness to expedite a job to gain a customer. The
minimum of 13 weeks can be considered the minimum Model Comparison
lead-time that molders would normally take to tool a Mold estimates of two test parts were made using
simple part. Historical data of these internal produc- B-D, D-P, and the proposed model in Eq 5. There-
tion parameters were not (and are not typically) avail- ported quotes for each of the parts were received from
able to molders, and thus could not be used in devel- three different mold makers. The first part is a 73 mm
oping the following predictive model: X 29.4 mm deep, end-cap-base of a water filter. It has

Tmold (weeks) = 13 + 0.0000558 + 0.007D

an outside circumferential thread split diametrically
(7) in two halves. The mold dividing surface is aligned with
R 2 = 0.7
the thread for easy ejection. The required surface finish
These results are surprising and useful. Increases in is SPI A3. This part has an envelope size of 154 cc and
complexity, as measured by the number of dimensions, is defined with a total of 73 blueprint dimensions. The
have a greater impact on tooling cost and tooling lead- second part is the top housing of a medical laboratory
time than similar size increases. Equation 7, shows that analyzer. The envelope dimensions are 375 mm X 200
every 100-count increase in number of dimensions, mm X 56 mm. The part is defined by 181 dimensions.
which is a normal phenomenon when parts are con- The features include three big and two small square
solidated into complex parts, increases tooling cost by windows for assembly and accessing internal com-
$4560, and tooling lead-time by 5 days. A comparable ponents, as well as structural ribbing. The exterior
increase in mold cost due to size increase is only pos- surface is textured while the inside surface needs a
sible if the size of the part is increased by 5600 cc, a regular SPI B 1 finish. The results are summarized in
six-fold increase if starting with a 1000 cc part. Table3.
The results show that consolidation of parts is prefer- The results indicate that the B-D model under-
able when the parts to be combined have low complex- estimates the costs of the molds and underpredicts
ity. Consolidating two already complex components the relative sensitivity between the two designs. The
into a more complex piece may increase tooling cost D-P model exhibits greater range than the observed
and tooling lead-time drastically. The cost incurred in mold quotes, but is likely the best predictor for these
higher tooling cost and lost sales due to late market two test parts. The proposed model overpredicts the
introduction may surpass the benefits expected from mold cost and does not exhibit adequate sensitivity. It
the parts consolidation. When timely market introduc- should be noted, however, that the model utilizes only
tion of a product is critical to its life cycle profit, it is three parameters and requires assessment of only size

102 JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3

 Early Cost Estimation for Injection Molded Components

4 Mean Quotes 1111. Estimated Tooling Costs

X 10
15~========~==~------~--------~
*o *o Steel Mold (P 20)
Aluminum Mold
•

10
~
.,"'
0::1
0
...,
• ••
c:

::!i
5

Tooling Estimates ($)

Fig. 3. Comparison of rrwld cost
and. rrwld lead-time estimates.
Mean Lead-Times vs. Estimated Lead-Times For Tooling

*o *o Steel Mold (P 20)

Aluminum Mold

..
...
120
. ...
,§
... ......
.
ID
...J
• ... ... 0

... ...
((\
"'
c:
~
0
......
...
.
1-
c:
15
......
...
::!i

.... ...

15 20 25
Estimated Tooling Lead-Times (wks)

and complexity. Moreover, these two design assess- p, is obtainable from polymer handbooks such as the
ments can and have been easily automated within Modem Plastics Encyclopedia (8) or from resin ven-
CAD systems and modem product development proc- dors. The runner and sprue weight contribution to total
esses. material consumption is significant for small parts but
negligible for large parts. For most thermoplastic ma-
MATERIAL COST ESTIMATION
terials, moreover, runners and sprues can often be re-
Material cost per part, Crnat, is the cost of direct ma- cycled without significant loss in final part quality. ln
terial that goes into making the part. For injection practice, after four cycles of repeated recycling, the
molding, this includes the cost of the plastic polymer, thermoplastic is significantly degraded that it is com-
additives, and fillers consumed per part. The following pletely different from the virgin material. Hence, in
equation expresses Crnat as a function of material vol- practice up to 15% recycled material from reground
ume, V, density, p, and polymer price per unit mass, P: runners, sprues, and second class quality parts are
blended with virgin material. (One notable exception
VPp
Cmat=-- (8) is the prohibition of recycled resin in medical, and
1-j food related applications by the U.S. Food and Drug
The part volume, V, is easily computed from a 3D Administration.) Since at the early stage of design, the
model of the part at the design stage. Polymer density, optimum number of cavities in the mold and hence

JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3 103

 Adekunle A. Fagade and David 0. Kazmer

Table 3. Comparison of Cost Estimates ($K). It is infeasible to assess the multi-dimensional prob-
Test Parts
ability density function across the process domain,
even if the variance and relationships between proc-
Source of Estimate End Cap Top Housing essing variables and quality attributes are determi-
0-P 14.69 47.2 nistic. As such, one approach is to assume Gaussian
B-0 5.43 17.52 distributions corresponding to measured process capa-
F-K 30.6 39.9 bilities. Hunkar Laboratories Inc. in Ohio (11) has de-
Mean Quote 17.3 38.5 veloped a classification of injection molding machines
from its survey of hundreds of machines over many
years. Deviations from the set of optimal process pa-
the runner volume are un.lmown, a conservative esti- rameters required to obtain the quality characteris-
mate for fis 10%. This agrees with a promotional lit- tics of a part are due to complex interacting variations
erature from DuPont (9). of noise variables, represented by a vector a = {ry},
where j = 1,2, ... ,m Frey and Otto (12) argued that
PROCESSING COST ESTIMATION though functional relationship between noise variables
The processing cost per part, Cpart• constitutes 400/o and quality characteristics are in general non-linear, a
to 80% of the part cost for both commodity and engi- linear relationship can be assumed in the neighbor-
neering plastic parts. Efforts to reduce the processing hood of a target vector, t. Equation 13 shows that the
cost at the design stage easily translate to significant normalized deviation of quality characteristic, By1• is
savings per part and to very large. Cproc is a function directly proportional to the deviation of the noise vari-
of the machine hourly rate, R,... production yield, P, ables from their target value, given the assumption
and the cycle time, t,. required to mold the part: of linearity in the neighborhood of the target noise
variable. However, the values of constants ku are not
R.nat.: known.
Cproc = 3600P (9 )
1 m
The cycle time has been estimated by performing a By1 =
USL 1 -
L
LSLt J=l
kiJ • (n1 - t1). (13)
transient thermal analysis to model the structural
rigidity of the part required for ejection (10). The rna- The matrix of constants, ku· relating changes in each
chine rate, R.na. is the amount charged per hour for noise variable to changes in the quality characteristics
the usage of the injection molding press. It is a conve- can be determined by experimentation, by analyzing
nient way of summarizing the direct processing cost historical data. by complex deterministic computations,
that is traceable to the part as well as the indirect proc- or by simulating the process. This last approach, using
essing costs that is allocated to it. The direct labor random event simulation and relative machine capa-
content of R,.. is the operator wage(s). while the indi- bilities was used to predict process yield for each class
rect costs include the costs for the consumption of of machine.
utilities and consumables by the press as well as a The results are shown in Fig. 4. The results indicate
depreciation charge. The machine rate ($/h) charged the trade-off between machine capability, number of
in the custom injection molding shop in western Mas- critical to function specifications, the passband of the
sachusetts has a linear correlation with the machine specifications, and defect rates. Figure 4 clearly identi-
clamp force, Fc1, measure in tons. Equation 10 show fies that machines with low class factors (highly capa-
the linear function that closely fits this data with a re- ble) will produce consistent moldings independent of
gression squared value of 0.986. This function is com- the number of critical dimensions specified. However,
parable to a similar relationship used by Boothroyd average and poor machines may present significant
and Dewhurst when adjusted using 4% inflation as quality problems, especially when multiple dimensions
shown in Eq 11. are specified to tight tolerances. While the methodology
has been developed and validated from a statistical
R.na = 31.33 + 0.725Fc~ 20 s Fc1 s 1500 (10)
perspective, it is impractical to believe that the yield
R.na = 32.00 + 0.631Fc~ 20 s Fez s 1000 (11)
predictions will be quantitatively accurate, especially
under development uncertainty when future defect
types may not be identified. However, the developed
ProcessiDg Yield E.umatioa method can provide qualitative and immediate feed-
Part quality attributes may exhibit some inconsis- back regarding the effect of design complexity and
tency due to manufacturing process, material, and specification tightness on the potential processing
operator variation. The probability of producing an ac- yields and cost.
ceptable product, P, is a function of the probability
IMPLEMENTATION OF MODELS
density function, pdf. and the product specification
IN CAD AND INTERNET
limits, LSL and USL, for each i-th quality attribute, y 1:
The models developed in this research have been
USL,
implemented within the SolidWorks CAD system. The
p = f pdf(y,)dy (12)
application evaluates a CAD model of a plastic part for
LSL,

104 JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3

 Early Cost EstimatiDnfor Injection Molded Components

Percent Defects for a Tightly Toleranced Part

100

90
... ...
)( )(
1
3
Critical
Critical
Dimension
Dimensions
+ + 5 Critical Dimensions
80 0 0 7 Critical Dimensions

70

60
Cll
Fig. 4 . Defects rates for an uyec· 0
tion nwlded part with tight speci.fi· .S! 50
CD
cations. 0
"/! 40

30

20

10

i 3 4 5 6 7 8 9
Machine Class Factor

its basic envelope size, complexity, and number of A world-wide web input interface has also been
cores. The user inputs information on surface finish, developed to make the cost estimator available to
tolerance level, and estimated production volume, N. the public. Through a drag and drop interface, users
A typical output screen is shown in Fig. 5 . can transfer their CAD files to this site and receive

.N • I I i•1 p "' 1 f!ltr.J ~

~""' ~' tl!=U ' -
... -'...:~ .-~Y;t_' J~ '5" " ~ ~ ::... e + ~ li:l :.0~ ~..i

' ,., eo..

8hMiil~­
t..h11761..-,
!if.·'~
Co« a ~Mit
I ,.p, S2illll15TI
L.~·.r-Mittclillual;»er1W11
. ;.p, so.w.
Hh-......-.....
LP,SQ15'!
=._P.M41idc:ou"'"'
""111104
(:!hT...,..,_
· ""IO.l'S5
h t..... l ...

~
(f

Fig. 5. Output screen for CAD implementation of cost estimator.

JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3 105

 Adekunle A. Fagade and David 0. Kazmer

immediate estimates of mold costs, processing costs, REFERENCES

and lead-times. The system utilizes the cost models 1. G. Boothroyd, P. Dewhurst, and W. Knight, Product
presented in this paper and currently evaluates with design for manufacture and assembly, New York, NY,
single components rather than assemblies. Further Marcel Dekker, Inc, 1994.
research is required to develop improved, application- 2. J. R Dixon and C. Poll, Engineering design and design
for manufacturing, a structured approach, Conway, MA,
specific cost models that leverage data and capabili- Field Stone Publishers, 1995.
ties from specialized industry suppliers. 3. G. B. Scurcini, "Complexity in Large Technological
Systems," presented at Measures of Complexity, Rome,
1987.
CONCLUSIONS 4. J. J. Shah and M. Mantyla, Parametric and Feature-
based CAD/CAM: Concepts, Techniques, and Applica-
This research has developed an automated costing tions, John Wiley & Sons, Inc., 1995.
methodology that designers of plastic parts can use 5. ASME, ANSI Y14.5M, American National Standard Engi-
when comparing alternative designs for cost and time neering Drawings and Related Docwnentation Practices:
to market. The method evaluates a part's complexity Dimensioning and Tolerancing, New York, American
Society of Mechanical Engineers, 1982.
at the early stages of its life cycle using the number of 6. D. V. Rosato and D. V. Rosato, Iryection ITIDlding hand-
dimensions from its geometric model. Validation was book : the complete 111Dlding operation technology, per-
performed using seventy-five different mold quotes formance, economics, 2nd ed., New York, Chapman &
across thirty different molding applications, indicating Hall, 1995.
a high correlation of part complexity with the mold 7. M. Malstrom, Manufactwi.ng Cost Engineering Hand-
book. New York, NY, Marcel Dekker Inc., 1984.
tooling cost and lead-time. All the independent vari- 8. MPE, Modem Plastics Encyclopedia 96, vol. 72, 1996.
ables in the models developed can be easily evaluated 9. DuPont, "Concepts in Engineering Plastics," in ProiTJD-
from feature-based CAD data. This enumeration of tional Literature of E.I. DuPont de NeiTIDurs & Co. Wil-
number of dimensions is a pratical alternative to the mington, DE, 1978.
10. H. Xu and D. 0. Kazmer, "A Stiffness Criterion for Cool-
use of complex algorithms for extraction and enumer- ing Time Estimation," International Polymer Processing,
ation of constantly changing design form features. The vol. 13, pp. 249-255, 1999.
results of the research are unique in their simplicity 11. Hunkar, "The Injection Molding Machines Class Factor,"
when compared to related work. Hunkar Laboratories Inc., 7007 Valley Avenue, Cincin-
nati, Ohio 45244 USA, 1998.
12. D. D. Frey, K N. Otto, and J. A. Wysocki, "Evaluating
Process Capability given multiple Acceptance Criteria,"
MIT Design Research Report, 1997.

106 JOURNAL OF INJECTION MOLDING TECHNOLOGY, SEPTEMBER 2000, Vol. 4, No. 3

View publication stats