Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
THE FUNCTIONAL BUILD APPROACH TO INTEGRATING AUTOMOTIVE BODY
PROCESS VALIDATION
September 1998
Patrick Hammett, Assistant Research Scientist Jay Baron, Manager of Manufacturing Systems
Donald Smith, Associate Director Steven Geddes, Research Associate
Office for the Study of Automotive Transportation
University of Michigan Transportation Institute 2901 Baxter Rd.
Ann Arbor, MI 48109-2150 734-647-0046
2
1. INTRODUCTION
Many companies are successfully integrating their process engineering and manufacturing
functions into the design phase of product development. Using practices such as concurrent
engineering, multi-functional teams and design for manufacturability, companies are reducing
costs and lead time for new products by avoiding the foreseeable problems that typically occur
with sequential product development [Ref. 1- 2]. Unfortunately, this integration has been far less
common during the process validation phase of product development, especially for complex
products. Once designs are released and manufacturing processes are constructed, most
companies rely on a sequential validation approach. First, they validate that the processes for each
individual component within the final product are capable of meeting all design requirements.
After each component is approved, companies validate their sub-assembly processes, and finally
the completed product. This sequential approach subscribes to the basic paradigm that final
product quality will be maximized if each individual component meets all its performance
requirements.
Although sequential process validation appears logical, it has proven noncompetitive for
many complex products such as the automotive body [Ref. 3]. Vehicle manufacturers typically set
fixed dates for the start of a new production model. Thus, any delays for process validation of
individual components reduces the allotted time for final body validation. In certain extreme cases,
difficulties encountered trying to achieve process capability requirements for individual body
components have even delayed final process validation until after the start of production. With
sequential validation, manufacturers often focus more attention on meeting requirements for
individual components than on the principal concern of the customer, the completed product. In
response to these problems, several automotive manufacturers are adopting a more integrated
approach to process validation using a practice known as Functional Build.
3
Under functional build, rather than validating components solely to their part print
specifications, manufacturers also evaluate components relative to their mating parts and
subsequent assembly processes. Manufacturers treat original design specifications as targets rather
than absolute requirements. For example, if a manufacturer experiences difficulty meeting a
particular component requirement, they may be able to resolve the problem in a downstream
assembly process or change another related, mating component more expediently. By analyzing
components in their subsequent assemblies, manufacturers also may find that certain original
requirements are not critical to the final product build. Here, a modification to the design print is
less expensive than physically changing already-constructed manufacturing tooling such as a metal
stamping die or injection mold. Functional build manufacturers use the validation process to
dynamically set final component design specifications in accordance with the capabilities of the
constructed processes.
When functional build is used, manufacturers may realize substantial cost savings over a
traditional process and product development life cycle. These savings result from eliminating
unnecessary process rework during the validation phase. Under functional build, rework decisions
focus on meeting final vehicle objectives and not necessarily on conformance to all original
component specifications. To implement a functional build approach, manufacturers must
effectively integrate their product design engineers, process engineers, and manufacturers
throughout the three major body validation activities: die construction tryout, stamping process
PartDesign
ProcessDesign
DieDesign
DieConstruction Pilot &
Ramp Up
Weld ToolsConstruction
Weld ToolsTryout
Die Tryout
Production Validation
Production
Figure 1. Major Development Steps for Auto Body Dies and Assembly Tools
4
validation, and assembly process validation. This paper explores the use of functional build
techniques and discusses their implementation at several leading manufacturers, particularly
Japanese automotive producers.
2. RECURRING AUTOMOTIVE BODY MANUFACTURING PROBLEMS
The functional build approach has gained popularity in automotive body manufacturing
primarily in response to two recurring problems:
1. An inability to produce component mean dimensions at their nominal specification.
2. Weak correlation between component dimensions and their resultant assemblies.
The following sub-sections examine each of these problems. Supporting dimensional data
also is provided from case studies at several manufacturers (three North American, two European,
and two Japanese satellite plants). These studies involved assembling 36 body sides using
individual components with known dimensional measurements. The panels were sampled from six
different production runs in order to capture some long-term process variation.
2.1 Mean Deviations from Nominal
Automotive manufacturers typically collect measurement data for individual body
components at several locations or dimensions. Figure 2 illustrates several typical measurement
locations for a body side component and the center pillar reinforcement. This reinforcement panel
is welded to the body side component in a subsequent assembly operation.
5
Figure 2. Measurement Locations For A Body Side And Center Pillar Reinforcement
Ideally, a manufacturer would like to produce each stamped component such that the
mean value at each measurement location is at its nominal specification. Figure 3 provides a
histogram of mean values for 600 dimensions across ten different body side components. For each
dimension, the design print has a tolerance of +/- 1.0mm about the nominal specification.
Figure 3. Histogram Of Mean Values For Dimensions On Ten Different
Body Components
#4
#5
#8
#4
#6
#9
#3#1
#2
#3
#15
#12
#11#14
#13
#7
#8
#6
#7
#5
#16
#17
#1#2
Histogram of Mean Values
0%
5%
10%
15%
20%
25%
30%
35%
-2.5
to -
2
-2 to
-1.
5
-1.5
to -
1
-1 to
-0.
5
-0.5
to 0
0 to
0.5
0.5
to 1
1 to
1.5
1.5
to 2
2 to
2.5
> 2
.5
Deviation of Mean Value from Nominal (Nominal = 0)
% o
f P
art
Dim
ensi
on
s
(6
00 t
ota
l)
6
This histogram suggests that producing all component dimensions to nominal is extremely
difficult. In this example, 20% of the dimensions have mean values outside their specification
limits.
The difficulty with producing all mean dimensions to nominal is not unique for this
manufacturer. Table 1 provides a comparative look at mean bias deviations for three
manufacturers. Mean bias represents the absolute deviation of a dimensional mean from its
nominal value. This table suggests that lack of mean conformance is an industry-wide problem.
Other empirical studies also confirm the difficulties in producing sheet metal dimensions to
nominal, particularly for larger, non-rigid components such as a body side panel.
Table 1. Mean Deviations for Three Manufacturers
Mean dimensional deviations often are attributed to difficulties in predicting metal flow
during forming operations. These deviations also may result from variation in die setups, press
operating conditions, or material properties. Another, less recognized problem, is the difficulty
associated with measuring large, complex-shaped components. Automotive manufacturers
measure body components in absolute, three-dimensional space. For rigid structures, they
typically use a 3-2-1 fixture layout, which requires six degrees of freedom to locate a part in space
prior to measurement [Ref. 4]. For large, non-rigid parts, however, body manufacturers must use
additional clamps or locators to stabilize the part for measurement. One concern with these
CASE STUDY Mean Bias < 0.5 mm 0.5 < Bias < 1.0 mm Mean Bias > 1 mm
I.* 51% 34% 15%
II. 66% 24% 10%
III. 64% 17% 19%
Mean Bias = | Mean - Nominal |
* Body side panel in this study has an integrated quarter panel (more complex panel)
7
additional locators is they actively influence the location of the surfaces being measured. In other
words, the positioning of the locators, not the stamping dies, may cause mean deviations.
These limitations are further exacerbated by the lack of adjustment factors to shift
stamping dimensional measurements. Many processes have built-in mechanisms, or setup
parameters, which allow manufacturers to shift critical part dimensions to a nominal value. In the
case of sheet metal stampings, injection and blow molded plastic components; this is not always
true. These types of processes require dies that dictate the geometry of the product. To change
the dimensions on a component, one must physically change or rework the dies. Die rework often
is an expensive and time-consuming process. It may involve several iterations, especially for large,
complex-shaped parts. Even after several rework iterations, manufacturers cannot always correct
every mean dimensional deviation. A functional build approach allows some individual component
dimensions to remain out of their initial design tolerance provided the assembly meets its
specifications and functions properly.
Figure 4. Body Side Measurement Fixture Illustrating the Complexity of Many Clamps and Locators Than Influence Part Measurement
8
For companies using sequential validation, mean deviations severely impact the product
development process. Under sequential validation, companies typically use Cp and Cpk indices
[Ref. 5] to approve parts for the next validation phase. Both indices are related to the ability of a
process to produce outputs within their specification limits. (For example, Cp is the total
tolerance / 6*standard deviation). The Cpk index differs from Cp because it includes the deviation
of the mean from its nominal in assessing process capability. These indices have become widely
accepted in the automotive industry because they provide objective criteria to validate the
conformance of components to their design requirements.
Empirical studies of stamping tryout suggest that although manufacturers may achieve Cp
requirements, they often fail Cpk due to mean deviations from nominal. In other words, their
processes have sufficiently low variation (small σpart-to-part) but are off target. Unfortunately,
shifting all mean values on a component to their original nominal specifications has proven an
infeasible task. Of the 600 body side component dimensions presented in Figure 3, over 50%
would not pass Cpk requirements using their original tolerances. Table 2 shows that of seven auto
companies investigated, the best one achieved a Cpk greater than 1.33 on just 39% of the body
side outer check points. Interestingly, all of these manufacturers are producing acceptable
finished bodies even with this lack of Cpk compliance.
Cp versus Cpk
Cp = ½ tolerance
3σpart-to-part
Cpk = ½ tolerance – Mean Bias
3σpart-to-part
(Note: Cp = Cpk if Mean Bias = 0)
Figure 5. Contrast Between Cp andCpk Quality Indices
9
Company Typical
Tolerance % of Points
|mean| > 0.5mm % of Points Cp > 1.33
% of Points Cpk > 1.33
A ± 0.7 mm 67 % 54 % 15 %
B ± 0.7 mm 62 % 63 % 10 %
C ± 1.25 mm 35 % 97 % 75 %
D ± 1.0 mm 58 % 90 % 23 %
E ± 0.5 mm 26 % 97 % 43 %
F ± 0.3 mm 16 % 92 % 29 %
G ± 0.5 mm 31 % 88 % 37 %
Table 2. Dimensional Conformance of Seven Auto Companies on Body Side
Outer Panels
Automotive manufacturers using Cpk criteria for sequential validation ultimately modify
original design tolerances to approve stamped components for production, and this is typically
done shortly before the production deadline approaches. This common expansion of original
tolerances to pass Cpk requirements implies a basic futility in the use of this index for approving
sheet metal components. This finding raises questions regarding the importance of achieving Cpk
requirements during tryout. Before addressing this question, we first must explore the
relationships between stamping component measurements and their subsequent assemblies.
2.2 Correlation of Components and their Assembly Dimensions
Another recurring automotive-body manufacturing problem is a lack of correlation
between stamping components and their welded assemblies. For rigid structures such as engines,
manufacturers assume that in the mating of two components, coordinated mean dimensions stack
linearly and component variances add. Based on this assumption, manufacturers try to produce
individual component mean dimensions at their nominal specification with minimal variance.
10
These manufacturers further assume that measurement values of components are correlated with
their assembly outputs. In other words, they may predict their assembly outputs based on the
measurements of the input components.
This basic assumption, however, does not always hold with non-rigid components. These
components may continue to deform during subsequent weld processes. Some components will
more closely resemble the geometry of the fixtures used to orient them at time of assembly than
their initial measurements. Non-rigid component dimensions also may conform to more rigid
dimensions during assembly [Ref. 6 - 7]. The net effect is that non-rigid component measurements
often poorly predict final assembly measurements.
Table 3 provides a summary of the dimensional relationships between stamped body side
components and their respective assemblies from six case studies. This data suggest two key
observations. First, the coordinated measurements (measurements taken in the same physical
location before and after assembly) shift unpredictably (> 0.5mm) during the weld assembly
process. Nearly half of the dimensions exhibited mean shifts of four and five sigma from stamping-
to-assembly (typical sigma values = 0.1 ~ 0.2mm). Second, virtually none of the dimensions
exhibited a strong correlation between stamping and assembly processes.
% of Dimensions
Company
No. of Dimensions
Correlation (R) > 0.6
|Stamp Mean – Asm Mean| Difference > 0.5mm
A 43 2 % 50 %
B 104 2 % 66 %
C 60 6 % 67 %
D 31 0 % 48 %
E 35 0 % 36 %
G 77 1 % 61 %
Table 3. Correlation of Part Dimensions Before and After Assembly
11
Several explanations exist for this lack of correlation between individual components and
their respective assemblies. Among them are:
• Deformation of metal during the weld process.
• Changes in the part locating schemes.
• Conformance of non-rigid component dimensions to other rigid areas of the assembly.
• Measurement system errors.
This lack of correlation presents serious ramifications for the traditional validation
approach. Here, manufacturers rework dies during tryout at both the tooling source and the
production facility trying to meet Cpk criteria for all component dimensions. Manufacturers using
sequential process validation estimate that rework accounts for 20-30% of the die costs. The
above correlation analysis suggests that this rework may have minimal impact on the final body
dimensional accuracy. In another study of a vehicle launch, one manufacturer found that 72% of
all root causes of dimensional variation were due to assembly fixture failures [Ref. 8]. Thus, if
launch dates are fixed, delaying assembly tryout to rework individual components may not allow
sufficient time to resolve the primary causes of final body dimensional problems.
The effects of rework are not limited to additional die construction and tryout costs.
Several manufacturers maintain that numerous rework iterations for a set of component dies also
impact the reliability of the tooling. Constant grinding and welding of dies increases the likelihood
of subsequent tooling failure. North American manufacturers maintain that because of this
concern, they must design dies that are more robust, i.e. costly, than their Japanese competitors
who more readily use functional build [Ref. 9].
The difficulties associated with producing components with mean dimensions at nominal,
coupled with the lack of correlation between non-rigid components and their subsequent
assemblies, has lead several manufacturers to abandon the traditional sequential validation
approach which uses Cpk as the primary decision criteria. As with many processes, the use of
statistical indices such as Cpk without integrating product/ process knowledge results in sub-
12
optimal decision-making [Ref. 10]. To integrate manufacturing experience and knowledge into the
validation process, many automotive companies are utilizing a functional build approach. The next
section discusses how this process works.
3. THE FUNCTIONAL BUILD PROCESS
Under a functional build approach, manufacturers typically assemble body components
into prototype car bodies or “screw bodies”. The logical progression is to build up small sub-
assemblies (builds) as separate screw bodies, and then to assemble the individual builds into the
whole body. The individual components used to construct these prototypes are stamped using
their regular production dies. Manufacturers evaluate these components in relation to their mating
automotive body parts. Components do not necessarily have to strictly comply with their original
design specifications. If a part assembles into an acceptable car body, it is not modified regardless
of whether it meets original tolerance requirements. As builds are attached to each other, they
become more rigid and less tolerable to deviations, so the first level build provides the greatest
opportunity to allow for deviations. Allowing panel deviations at the first level build reduces
rework costs and lead time associated with individual production dies.
For example, Figure 6 considers the mating of the center pillar reinforcement and the body
side panel. The center pillar reinforcement is a structural component and thus will have greater
influence on the final assembly. If the body side panel is 1mm outboard from the centerline of the
car, but the center pillar is at nominal, the overall assembly will likely shift toward nominal. This
shift occurs because the mating surfaces are in parallel. Thus, the less rigid body side panel will
conform to the rigid inner structure [Ref. 7]. Under the traditional approach, a manufacturer
would likely rework the body side panel because the outboard stamping condition would cause
this part to fail its Cpk requirements. In contrast, a functional build manufacturer would assemble
these two components and make any rework decisions based on the resultant assembly and not
necessarily on Cpk compliance. In some cases, the resultant assembly might still deviate from
13
nominal, but this manufacturer may find it easier to adjust an assembly process locator than
physically alter a stamping die.
Figure 6. Parallel Assembly Of A Non-Rigid Mating Surface To A Rigid Reinforcement
Functional build evaluations typically occur in two phases. In the first phase, evaluations
are made using parts off regular production dies but at the construction or tooling source. This
evaluation process provides a mechanism to approve the shipment of dies (“die buy-off”) to the
production facility. In this process, manufacturers consider both the actual dimensional
measurements of the components and their relationship to mating components. The primary
objective of this phase is to correct those problems known to affect subsequent assembly
operations, while delaying rework decisions for those dimensions with unknown impacts. The
second functional build phase occurs after the dies are shipped to the production facility. The
primary objective for this evaluation is to produce a dimensionally acceptable finished body. A
manufacturer even may choose to rework certain in-tolerance dimensions if the changes will
improve the overall manufacturability or appearance of the final body.
In the functional build process, manufacturers first evaluate components by constructing
screw-body subassemblies. They then build these subassemblies into full screw-body prototypes
(usually 2-3 per car model over the course of a year or so). Manufacturers usually assemble both
Body Side Center Pillar Surface
clamp
AssemblySurface1.0mm
SIDE VIEW
WELD GUN
Body Side Center Pillar
MatingWeld
Surfaces
14
the subassemblies and the full screw-body prototypes with screws or rivets instead of the normal
welding operations. They use screws or rivets to minimize the distortion of components caused by
welding. Thus, screw-body prototypes help to determine whether individual components may
feasibly produce an acceptable subassembly or final assembly. Manufacturers assume that if the
screw-bodies are acceptable, they may eventually setup weld tools to match them. Thus, the
screw-body process provides a technique for setting-up and tuning-in welders.
The first screw-body prototypes for sub-subassemblies usually are constructed between
twelve and fifteen months prior to the start of production. Additional screw-body prototypes are
assembled based on need and strategy. For example, manufacturers may construct additional
prototypes if a part significantly changes during tryout as a result of a design change or a
manufacturing problem. The final screw-body prototype typically is constructed four-to-six
months prior to the start of production.
The screw-body process allows significant changes to the original design without
impacting customer perceptions. For example, although engineers may symmetrically design the
Figure 7. A Screwbody Build Has Several Sheet Metal Pieces Attached With Screws (Replicating Spot Welds) So The Assembly Can Be Evaluated
15
right and left side of a car body, a functionally built body may not have this characteristic. The
right side may build several millimeters outward, while the left side is inward from design intent.
As long as this lack of symmetry does not result in structural or appearance problems such as
inconsistent body gaps, the customer is unaware of the lack of conformance to the original design.
The argument for this approach is that manufacturers should not commit resources to correct
deviations from the original design unless the deviations affect customer perceptions.
When making an engineering change under the functional build approach, manufacturers
search for the least costly alternative without sacrificing product quality. For example, if the
quality of the body is unacceptable because an overall dimension of several parts in a subassembly
is too long, only one die may need rework. In many cases, manufacturers choose the least
expensive die for rework even if it produces a part with dimensions inside specification limits. For
example, a dimension with a measurement of 0.5mm out-of-tolerance could mate with a
dimension at its nominal. In some cases, altering the dimension at nominal is less expensive than
reworking the out-of-specification condition. The objective is to produce an acceptable final
assembly. This solution might differ for a manufacturer using a traditional, sequential validation
approach where each part is evaluated independently against its design specifications. Here, a
more expensive die or possibly several dies might be modified.
4. INTEGRATING PROCESS VALIDATION RESOURCES
Specialization in the development of component parts and processes results in a narrow
focus by development engineers and manufacturers toward their individual tasks. Functional build,
however, shifts the development focus from optimizing individual components to the entire body.
This approach represents a major paradigm shift for product and process design engineers. Final
specifications for the components within an assembly are determined concurrently with the
validation of the component and final assembly processes. Whereas concurrent engineering
requires more integration of downstream functions into the design process, functional build
involves a more active role of design engineers into the validation process.
16
One concern among design engineers is that functional build simply shifts development
problems from die construction and metal stamping to final assembly validation. Assembly
manufacturers and process engineers argue that if stamping plants would produce parts at
nominal, then they could efficiently validate their tooling. Proponents of functional build certainly
agree that greater conformance of stamping dimensions to specification is desirable. They,
however, recognize that once manufacturing processes are constructed, drawing changes to
design prints or minor modifications to assembly tooling locators may be less expensive than
physically reworking tools. Functional build does not imply that the original design would not
produce a high quality body, rather that alternative designs also could lead to the desired result.
Thus, a basic tenet for functional build success is a commitment from product engineering,
stamping, and assembly to work toward improving the overall vehicle development process and
not necessarily optimizing an individual activity.
Engineers and manufacturers may also simplify the validation process by designing
flexibility into the assembly-build process. One example of greater flexibility involves the use of
slip planes. Here, manufacturers overlay two mating surfaces in parallel prior to welding. With a
slip plane, a manufacturer may adjust or slip one mating surface such that the resultant assembly
meets its desired specification. The use of parallel welding surfaces is particularly effective at
setting cross-car distances and for joining outer-body components to rigid structural components.
5. CURRENT APPLICATIONS OF FUNCTIONAL BUILD
Functional-build type practices have existed for several years throughout the world at
different automotive manufacturers. The principal evaluation tool, or screw-body process, may be
traced to the early 1970s to a process known as “screw and scribe”. Here, manufacturers would
screw mating components together to check for assembly interference. One Japanese
manufacturer iterated on this process and began using it as an evaluation tool for die rework
decisions. By performing functional evaluations, this manufacturer has been able to eliminate
unnecessary rework and reduce overall validation time for new vehicle launches.
17
Functional build requires an additional step to the validation process. It requires resources
to construct screw-body assemblies and delays potential time to rework dies waiting for functional
build decisions. In those cases where a component dimensional deviation is known to cause an
assembly problem, manufacturers should probably rework the die rather than spend time
performing a functional evaluation. Manufacturers gain little if they evaluate a component and/or
try to modify assembly tooling only to ultimately rework the stamping die. Thus, a challenge with
functional build is determining when to rework dies versus when to perform functional
evaluations.
The various approaches used to implement functional build exist across a wide spectrum.
On one end, several leading Japanese manufacturer utilize a relatively subjective approach to
functional build. They rely on manufacturing experience in determining whether to rework a die.
They have recognized limitations of this approach due to the reliance on technical experts and the
absence of objective data. In contrast, several North American manufacturers are attempting to
develop more quantitative approaches to apply functional build. As shown in Figure 8, both
North American and Japanese companies appear to be heading for a compromise that employs a
criteria-based application of functional build.
One approach to developing objective criteria for functional build decision making is to
separate the criteria for the mean and variance, rather than combine them, as the Cpk index does.
The principle behind this approach is that for non-rigid sheet metal components, controlling
variation about the mean is more critical than the relative location of the mean to a design
nominal. The reason is that manufacturers may correct certain component mean deviations in their
assembly tooling at a lesser expense than to rework the stamping dies.
18
Figure 9 presents the basic tool-buyoff decision model. Manufacturers must decide
whether reworking a stamping die is more cost effective than making an adjustment in the
assembly process. This figure relates the mean deviation for a non-rigid stamping dimension to the
cost of producing an acceptable assembly by either reworking a stamping die or making an
assembly adjustment. In the case of assembly adjustments, this figure presents two cost curves to
represent the greater uncertainty in compensating for a stamping deviation. In some cases, large
stamping deviations may require only minor tooling adjustments. Some large deviations, however,
could require significant changes to the tooling to prevent undue stress in the assembly.
Manufacturer may even find that correcting certain deviations is infeasible in which case they
ultimately would need to rework the stamping dies.
Pure NetBuild
Reliance onStatisticalCriteria(Cpk)
Typical NorthAmerican
Companies
SeveralJapanese
Companies
Criteria-Based Functional Build(Combine Objective Criteria and
Technical Expertise)
PureFunctional
Build
Reliance onSubjective
Criteria
Figure 8. Net Build and Functional Build Companies Are Beginning to Compromise by Combining Less Rigorous Criteria With Technical Expertise for Evaluation
19
Figure 9. Decision Model For Tooling Rework
Figure 9 suggests manufacturers are unlikely to benefit from reworking dies for small
mean deviations (less than X1). They should be able to either ignore these deviations or make a
simply adjustment in the assembly tooling. In this region, manufacturers should clearly not rework
the tools regardless of Cpk compliance.
At some level of mean deviation (greater than X2), however, a manufacturer may be
unable to make a correction to the assembly process. In this region, manufacturers should rework
the die before constructing screw-body prototypes because they likely will have to ultimately
rework the die. The region between X1 and X2 represents the unknown or functional build region.
Here, reworking the die may or may not be less expensive than making a compensation in
assembly. Furthermore, simply using a measurement index such as Cpk or mean deviation may not
result in the best decision. In this area, die-construction, stamping, and metal assembly personnel
must integrate their knowledge to determine the best approach. Table 4 shows approximate
values for X1 and X2 for two classifications of sheet metal.
Cos
t ($)
to B
uild
Acc
epta
ble
Ass
embl
y
Deviation of Stamping Dimension Mean from Target
Compensate in Assembly
Region B(functionalevaluate)
Region A(do not rework die)
Region C(rework die)
Rework Die
X1 X2
LBUB
UB= Upper Cost BoundLB= Lower Cost Bound
20
Part Type X1 “Gray Area” X2
Non-rigid < 0.5mm 0.5 ~ 1.5 mm > 1.5 mm
Rigid < 0.3 mm 0.3 ~ 1.25 mm > 1.0 mm
Table 4. Estimated Values for Lower (X1) and Upper (X2) Deviation Thresholds
For Sheet Metal
Current functional build research efforts involve trying to define criteria for X1 and X2
based on various panel properties. The approach is to utilize historical data for certain types of
components to determine values where manufacturers minimize their decision-making risk.
Functional build manufacturers must identify when it is value-added to construct screw-body
prototypes versus reworking dies or adjusting assembly tooling.
6. CONCLUSIONS AND RECOMMENDATIONS
This paper presents a new paradigm for the validation phase of product and process
development. Manufacturers that continue to embrace new methodologies and business practices
will not only remain competitive in the global market but also set the benchmark for competition.
By using an integrated validation approach like functional build, manufacturers may accelerate the
product development life cycle while saving costs in manufacturing process development.
An important issue facing functional build manufacturers is its future application.
Functional build involves an additional validation step to construct screw-body prototypes and it
requires greater coordination between development functions. Several manufacturers using
functional build hope they will eventually develop the manufacturing knowledge necessary to
identify when rework is value-added without having to construct screw-body prototype
assemblies. The goal is to replace these prototypes with process simulation or math-based
functional build. Even with this approach, two fundamental principles behind functional build will
remain. First, manufacturers should never rely solely on statistical indices to make tooling and
validation decisions. Incorporating process knowledge with dimensional data results in more
21
effective decisions than those based purely on quantitative results. Second, development functions
must integrate their product/process knowledge and focus on the final customer product and not
necessarily individual components.
22
References:
1. Nevins, J.L., & Whitney, D.E. (1989). Concurrent Design of Products and Processes. New
York, NY: McGraw-Hill Publishing Co.
2. Hoffman, D. (1997). “An Overview of Concurrent E Proceedings Annual
Reliability and Maintainability Symposium, IEEE, pp. 1-6.
3. Clark, K.B., & Fujimoto, T. (1991). Product Development Performance-Strategy,
Organization, and Management in the World Auto Industry. Boston, MA: Harvard Business
School Press.
4. Liggett, J.V. (1993). Dimensional variation management handbook - A guide for quality,
design, and manufacturing engineers. Englewood Cliffs, NJ: Prentice Hall.
5. Kane, V.E. (1986). “Process Capability Indices,” Journal of Quality Technology, 18.
6. Takezawa, N., (1980). “An Improved Method for Establishing the Process-Wise Quality
Standard,” Reports of Statistical Application Research, Union of Japanese Scientists and
Engineers (JUSE) Vol. 27, No. 3, September, pp. 63-76.
7. Liu, S.C., Hu, S.J., & Woo, T.C. (1996). “Tolerance Analysis for Sheet Metal Assemblies,”
ASME Journal of Mechanical Design, Vol. 118, March, pp. 62-67.
8. Ceglarek, D., and Shi, J. (1996). “Fixture Failure Diagnosis for Autobody Assembly Using
Journal of Engineering for Industry, Vol. 118, pp. 55-66.
9. Hammett, P.C., Hancock, W.M., & Baron, J.S., (1995). “Producing a World Class
Engineered in Japan - Japanese Technology Management Practices,
Edited by J.K. Liker, J.C. Campbell, and J.E. Ettlie, N.Y.: Oxford University Press.
10. Schwarz, S.A., and Lu, S.C-Y. (1992) “Representation, Acquisition, and Manipulation of
Probabilistic Attribute Values to Support Engineering Decision-Making,” Transactions of
NAMRI, pp. 261-267.
23
THE AUTHORS
PATRICK C. HAMMETT, Ph.D. Pat Hammett, PH.D., is an Assistant Research Scientist in the Manufacturing Systems Group of the Office for the Study of Automotive Transportation. He is involved in research developing and implementing world-class manufacturing, quality, and management practices in the production and development of the automotive body for Chrysler, Ford and General Motors. As part of the Body Systems Analysis Task Force of the Auto-Steel Partnership, a joint partnership consisting of General Motors, Ford, Chrysler, and twelve North American steel manufacturers, Pat conducts research to assist the North American stamping industry in achieving world class performance levels in automotive body development lead time and dimensional quality. Pat is a principal researcher and supervisor for several research projects to optimize the automotive body development process at General Motors Corporation. He is also Adjunct-Assistant Professor of Industrial and Operations Engineering teaching a course on Total Quality Management for the College of Engineering for senior engineers and graduate students. Pat has a Ph.D. in Industrial and Operations Engineering from the University of Michigan, a Master of Science in Engineering from the University of Michigan, and a Bachelor of Science in Engineering from Purdue University. JAY S. BARON Jay is the Manager of Manufacturing Systems for the Office for the Study of Automotive Transportation. He holds a Ph.D. and Master's Degree in Industrial and Operations Engineering from the University of Michigan, and an MBA from Rensselaer Polytechnic Institute. Dr. Baron has been researching techniques for achieving process and dimensional control in sheet metal stamping and auto body assembly. As part of his doctoral research, he has developed new analytical methods and procedures to support the dimensional analysis and control of stamped metal assemblies. Dr. Baron has also been researching evolving technologies in the auto industry. His interests are to conduct manufacturing analysis and planning for the automotive industry, concentrating on productivity, product quality, and system economics, and to assist in the development and application of new manufacturing technologies such as computer-assisted manufacturing, laser welding, hydro forming, sheet metal adhesives, and process control systems. Dr. Baron’s primary interests are researching and improving product development and manufacturing process capability through effective design, development and implementation. Prior to becoming a researcher at the University of Michigan, Dr. Baron worked for Volkswagen of America in Quality Assurance. He has also been a Staff Engineer and Project Manager at the Industrial Technology Institute in Ann Arbor, and Rensselaer Polytechnic Institute's Center for Manufacturing Productivity in Troy, New York. While at these organizations, he managed consulting projects for over twenty different companies in applying new manufacturing technology. Several projects were concerned with manufacturing systems and economic modeling to evaluate the strategic and tactical deployment of technology.
24
DONALD N. SMITH Don Smith, who was Director of the University of Michigan's Industrial Development (IDD) since 1964, became Associate Director of the University's Office for the Study of Automotive Transportation in July 1993. His major responsibility continues to include the organization and management of research projects dealing with technological and economic changes in major U.S. industries. His research emphasizes these industries: automotive; machine tool, tool and die; robotics and flexible automation. He has a long-standing interest in factory automation and related technologies for the factory of the future. Along with Larry Evans, Mr. Smith is co-author of a textbook entitled Management Standards for Computer and Numerical Controls. He has also published several technical papers. Mr. Smith has participated in metalworking industrial development study missions in foreign countries, including Israel and South Korea, under the sponsorship of the United Nations' Industrial Development Organization and the World Bank. He has performed two studies for the National Science Foundation identifying productivity research requirements. Together with several executives from U.S. machine tool companies, Mr. Smith was a member of the National Academy of Engineering's committee which evaluated machine tool technology in relation to national and economic trade issues. He was a member, again with several machine tool industry executives, of the Midwest Governors’ Commission on the revitalization of the U.S. machine tool industry. Mr. Smith has received awards from foreign governments, including Sweden and Israel, and from several technical societies, including Society of Manufacturing Engineers, NCMS, and the California Society of Professional Engineers. He is a Fellow in the Society of Manufacturing Engineers. In 1988 he was awarded the SME Joseph A. Siegel Service Award. He is listed in Who’s Who in Finance and Industry, and also in Who’s Who in Science and Technology. STEVEN W. GEDDES Steve Geddes is a Research Associate in the Manufacturing Systems Group of the Office for the Study of Automotive Transportation. His current research includes data collection and analysis of decision processes related to automotive stamping and dimensional validation for stamping dies and assembly weld tools. The objectives of his research are to develop a knowledge base of stamping and assembly process capability that supports the synthesis of a world-class process validation strategy for sheet metal components. He has conducted research for several automotive companies and performed analyses for plants in the United States, Canada, Europe and Brazil. Steve holds a Master's Degree in Industrial and Operations Engineering and a Bachelor's Degree in Mechanical Engineering and Applied Sciences, both from the University of Michigan.