Concurrent Engineering http://cer.sagepub.com

Managing the Product Realization Process: a Model for Aggregate Cost and Time-to-market Evaluation Michael R. Duffey and John R. Dixon Concurrent Engineering 1993; 1; 51 DOI: 10.1177/1063293X9300100106 The online version of this article can be found at: http://cer.sagepub.com/cgi/content/abstract/1/1/51

Published by: http://www.sagepublications.com

Additional services and information for Concurrent Engineering can be found at: Email Alerts: http://cer.sagepub.com/cgi/alerts Subscriptions: http://cer.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.co.uk/journalsPermissions.nav Citations http://cer.sagepub.com/cgi/content/refs/1/1/51

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

51-

fIC0t4CURIIEb4T Et4G1b4EEk1b4Gi ReMOfch and App!imNons Managing

the Product Realization Process:

a

Model for

Aggregate

Cost and Time-to-market Evaluation Michael R.

Duffey* and

John R.

Dixont

of Engineering Management, The George Washington University, 2130 H St N.W., Washington, DC 20052, USA and † Department of Mechanical Engineering, University of Massachusetts, Amherst, Massachusetts 01003, USA

* Department

Received 4

September 1992; accepted

in revised form 29 October 1992

Product and process realization for innovative mechanical products often risks considerable cost- and schedule-related problems A tool to aid rapid evaluation of the aggregate development cost and time implications of alternate preliminary designs and design processes could therefore be valuable, particularly in a concurrent engineering environment. To this end, a model is proposed and demonstrated with a proof-of-concept computer implementation. There are several potential applications envisioned: as a "prospectus" for new product designs to evaluate development cost and time within a specific organizational context; to assist managers in concurrent scheduling of design, tooling, and other preproduction activities; as a vehicle for budget negotiation between engineers and financial managers during the design process, and as an aid for value analysis This research is intended as a starting point to investigate new computer-based management tools made possible in part by emerging engineering design representations and related methodologies for cost and time evaluation

Keywords: product realization, cost, time-to-market, design management.

1. Introduction Discrete

product realization is a complex, interdisciplinary process. At early design stages, decisions must be made not only about physical attributes of the design, but also about scheduling and resource allocation for many product and manufacturing engineering activities, as well as concomitant activities in purchasing, finance, marketing, etc. Typically, complex interdependencies exist among these disparate activities, and it is difficult to predict how decisions will affect overall organizational objectives of low cost, high quality, and short development time Many decision-support needs in this process seem to fall in a gap between emerging design-for-manufacturing (DFM) models (which can evaluate design attributes for relative cost of a specific manufacturing activity) and managementlevel models (such as gross cost and time estimates from PERT-type networks) DFM models generally neglect interdependencies between activities and organizational constraints (for example, strict adherence to a design-forassembly methodology may increase time-to-market due to long lead times for complicated component tooling [13]) At the other extreme, the lack of detail in managment-level models obscures many important issues, such as overall cost sensitivity to proposed changes in design

configurations This research addresses the

&dquo;gap&dquo;

agement and engineering models

between these

at the

man-

preliminary design

stage (prior to detailed design) for mechanical assemblies. In our proposed model, implemented as a proof-of-concept computer tool, relational matrices are first defined for the three distinct sets of design knowledge data: product attributes, realization activities, and resources An activity network is then generated using these data, and a simulation is run to obtain an aggregate cash flow of preproduction costs for a given design configuration. We will first discuss representational issues in the model, then show applications for decision making using the proof-of-concept computer implementation.

2. Product

elements, activities, and resources

Consider a comprehensive representation of the realization process as a triple (P, A, R), for which P= { p} is the set of product elements, A={a} is a set of activity classes, and R= {r} is the set of resources available for realization. Each element in each set has certain attributes (e g. p = {s} ). These attributes denote hierarchical ordering within each set (i e level of detail) and relationships between product and activity sets and between activity and resource sets (Figure 1), as well as other information used to assemble an activity network for simulation of aggregate cash flow We

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

52

1. Product element, their inter-relations.

Figure

activity class, and

resource

objects,

and

will first consider representational conventions and requirements of P, A, and R individually, and then discuss issues related to inter-relationships between these three sets. Borrowing from emerging conventions in knowledgebased design, we can structure a product P hierarchically by defining each element as a subassembly, component, or feature (this last is described here as a subcomponent level element which has both form and functional or manufacturing intent). While representational properties in traditional CAD systems have typically included only low-level geometric information, knowledge-based representations are coming to include a much richer set of design and manufacturing data, including those useful for activity-specific cost analyses in assembly, stamping, injection-molding, etc. [6]. A resource representation R can also be hierarchically ordered similar to a company organization chart for departments of product and manufacturing engineering, tooling, marketing, finance, as well as available outside vendors, prototyping facilities, etc. &dquo;Leaves&dquo; in this resource tree may be people or machines or some combination of the two. Resource properties include direct and overhead cost rates, non-recurring costs, and other capital expenses. A classification scheme for realization activities A is of course more problematic than for product elements or resources. For purposes of this model, consider the partially shown (and certainly tentative) and/or tree of design, prototyping, and other realization activity classes in Figure 2. Product realization concerns the design, building, and evaluation of physical and analytical models prior to production runs. Process realization concerns the tooling and manufacturing ramp-up activities prior to on-going production. Much activity interdependence occurs between these two groups. Decision points are the periodic management evaluations which approve and allocate money for successive of in-progress project stages, and may determine

reworking

Figure

2. And/or tree of

product or process designs. Leaves of this tree are the smallest &dquo;units&dquo; of activity classes meaningful for development cost and time estimation purposes and for which specific machine and human resources, and capital expenditure, can be identified. In many cases, there are subsets of activity classes for which some generalized procedural relationships exist. These activity templates, identified from the field research, define graph segments for an aggregate activity network that can be created to estimate cash flow. In general, such activity templates are more standardized for process realization activities than for product realization activities. For example, the realization of progressive die tooling has a fairly routine ordering of activities (strip layout, selection of standard die components, layout for die assembly etc.) that varies very little from one component to another (though the work time required for an activity such as strip layout is highly dependent on part complexity, and may vary considerably). A template for proof-of-concept design, however, may vary greatly depending on company practice, functional requirements, technology and materials, designer experience, and many other factors. In summary, we demand of our &dquo;comprehensive&dquo; representation of activity classes that it define some taxonomic classification, multiple levels of detail, and whatever generalized precedence information is available for activity subsets. Obviously, many aspects of the above &dquo;comprehensive&dquo; representation are tentative-for example, hierarchical ordering of P, A, and R might be replaced by a more flexible network exposition. Certainly, the classification of activities needs further work-even a definition for some &dquo;activities&dquo; with finite beginning and ending can be arbitrary, particularly within the context of concurrent engineering. (See Dixon et al. [4] for a related discussion of design problem taxonomies.) Nonetheless, this representation serves as a starting point for presenting the larger model concepts and implementation described below.

3.

P, A, /9 relationships

By configuring relationships between P and A, and between R, a specific realization problem is defined from which an aggregate activity network and an estimate of cash flow can be generated. To do this, two relational matrices are used in a manner somewhat similar to quality function deployment [3] (Figure 3). In the first matrix, row and column A and

activity classes (only partially shown)

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

53

Figure 3. P, A,

R

relationships configured

in

two relational matrices

labels represent P and A elements,

spreadsheet format

in the

respectively. Using a implementation, the tree hierarch-

ies for P and A are &dquo;traversed&dquo; to find where the proper match in level of detail occurs, and elements in the matrix are assigned. For example, tooling design is an activity done at a component level of the product, but prototype testing might be done at its &dquo;parent&dquo; subassembly level. A proof-ofconcept design effort might be assigned to a subassembly, component, or even a feature of a component (e.g. some novel compliance design for a snap-fit). A high-level manager should be able to assign or view an activity at one level of detail (e.g die design), which in turn should point to multiple, lower-level activities (strip layout, die cutting, etc.) which are more meaningful as cost and time drivers for product attributes For example, a higher-level manager might need not &dquo;view&dquo; a snap-fit design, but it nonetheless has specific contribution to cost/time. When a product element is matched to an activity class an instance of that activity class is created, and the second matrix assigns relationships between the activity instance and the required resources. After this, probability distributions for activity times are assigned, as well as activityspecific costs There are three ways to do this: subjective estimates (e g. the PERT approximation of a beta distribution using &dquo;best case/most likely/worst case&dquo; estimates); using statistical data gathered from historical records within the firm; or using DFM-type methodologies which generalize data gathered from multiple companies Although subjective estimates of distribution parameters from participants in the field study were used for the implementation examples, this model was developed with the implications of this third method in mind: for example, Mahajan [11] has devised a method for automatic estimation of die design and build hours for a stamped component using a knowledge-based representation that includes attributes of material type and thickness, and its subsidiary features Other knowledgebased methods can evaluate a component design for downstream cost such as die materials (e.g. tooling steel stock and purchased die components such as punches and bush-

ings [7]). 4.

Activity network

for the realization

process output of this model is an aggregate cash flow, and to generate this we must first configure the multiple activity

The

instances into an aggregate activity network. This is done partly using predefined precedence relationships within the activity templates, partly using heuristics (e.g. placement of decision points), and partly interactively (e.g. &dquo;over-lapping&dquo; of templates). The explicit relational model for P, A, and R enables some limited automation of scheduling constraints and activity assignment This includes precedence of activity templates (e.g a heuristic such as &dquo;tooling design of a component cannot begin before layout design of its parent subassembly&dquo;) and scheduling of decision-point activities for budget approval and go/no-go decisions as prescribed by existing company policies. Other &dquo;indirect&dquo; activities which are often not explicitly identified in the realization process (e.g. purchasing and inventory planning activities) can also in some cases be automatically assigned and scheduled according to how many and what type of product elements have been defined. In addition, the completeness of the problem configuration can be checked: for example, (1) at a level not higher than the component level, every product element must be linked to tooling design and fabrication activities (or vendor purchase), and (2) every leaf in the product element tree must be linked to one or more nodes in the concept-design activity subtree either directly or indirectly by an ancestor product element node. In the Introduction

we

stated that traditional

use

of PERT-

type networks for modeling the realization process has been limited due to

of detail and disassociation from data There are other reasons as well. PERT has been generally criticized for statistical inconsistencies such as a time-to-completion calculation based solely on critical path [8] (this might be a particular deficiency for large-scale concurrent engineering projects for which there are many parallel paths) Perhaps most relevant to concurrent engineering, however, are types of activity interdependencies identified in the course of the field research that preclude PERT network assumptions These include: (1) design iteration during proof-of-concept and production prototyping phases of engineering design, and (2) changes to or cancellation of manufacturing process activities concurrent with product design when the redesign iteration in (1) occurs. For the first type, design is often informally described as an &dquo;iterative&dquo; process, and in fact many sequences of activities in the realization process are repeated one or more times The proof-of-concept &dquo;template&dquo; in Figure 4 has two iteration loops identified by design engineers as typical at one of the companies studied. The inner loop (minor-proofchange) typically occurs when minor dimensional changes are proposed after engineering analysis of test results for a proof-of-concept model. In such cases, detail drawings must be revised, tolerances again reviewed by manufacturing, and the other activities inside the loop are &dquo;repeated&dquo; In some cases less work time is required for iterated activities, and their values must be reassigned in the corresponding network simulation. The outer loop (major-proof-change) represents redesign efforts required after a review of the proof-of-concept model by upper management, including marketing, production, finance, and the other non-engineering evaluations. Often the redesign for this outer loop requires considerably more effort for affected activities (a new form or functional concept may be required, or new layout of the subassembly with respect to other subassembl-

product and

a coarseness

resource

ies, etc.).

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

54

Figure

4.

&dquo;Templates&dquo;

for

proof-of-concept and stamping (tooling only) activity

The second type of activity inter-relationship (manufacturing process activities concurrent with product redesign) relates to what Clark and Fujimoto have termed &dquo;overlapping problem solving&dquo; [2]. A template for tooling of stamping dies is shown concurrent with the proof-of-concept template in Figure 4. When product redesign is required, concurrent process activities such as tooling may also need to be repeated. In some cases, rework for tooling may be relatively insignificant. For example, minor dimensional changes to injection molds are often anticipated, and can be performed with minor remachining and welding operations For more significant changes of part form, however, entire tool sets may have to be scrapped and the process begun again. Much of the risk evaluation for concurrent product and process design involves assessment of possible redesign activity, and the resulting additional cost and time for tooling rework. (For conventional, well-tested designs with low-cost tooling, concurrence here makes sense; for more innovative designs and high-cost tooling, end-to-end sequencing of product design and processing activities may actually yield a shorter expected development time.) By allowing a manager to explore alternate &dquo;overlapping&dquo; of templates, and assign time and branching probabilities based on the best available estimates, this model explicitly embodies some of these decision-making trade-offs. Uncertainty of aggregate cost and time in this model is determined by two properties of the activities: (1) parameters of the probability distributions for duration and related cost of each activity, and (2) probabilities for iterative branching within design templates where redesign may occur. For (1), activity duration in the model is separated into two parts the lead time and the work time, each represented as a probability distribution function. The former represents the delay time that often occurs for the many &dquo;overthe-wall&dquo; hand-offs of design information within an organization. This can be an important time component, particularly in organizations where new product design vies for resources and priority with on-going routine design work for minor product variants and production. The latter (work time) represents actual productive time, which also determines associated labor and machine costs when multiplied by the cost rates of the related resources. For (2), branching probabilities are obviously quite subjective, but can help to assess &dquo;worst case&dquo; scenarios, and reflect the fact that the required number of prototype builds is uncertain in many discrete product industries [1].

5. A

Proof-of-concept implementation

proof-of-concept computer tool implemented using activity and

based

on

resource

this model was data from field

sequences

studies and some simplified product design examples. Only the preproduction process (i.e. configuration design through tooling) and related cash flow was modeled, extension to production and later life-cycle activities obviously would be desirable in a full-scale implementation. In Figure 5, a preliminary design for a blade subassembly of a battery-powered beard trimmer is represented by its product attribute hierarchy (the subtree BLADE-SUB-ASSY in the Product Attributes window in the upper left corner) and the partially obscured graphic in the lower right-hand corner. Activity classes are shown in the Available Activities window After selecting the subtrees TOOLING (an activity class) and BLADE-SUB-ASSY, their respective children elements become columns and rows for the relational matrix in the lower left corner of Figure 5. In the matrix, instances of STAMPING template activities are then assigned for two components (MOVING-BLADE and STATIONARY-BLADE); INJECTION-MOLDING activities are assigned for the CAMTRANSFER and BLADE-PLATFORM components; an outsourcing process with a VENDOR is assigned for the RETAINING-SPRING component. The &dquo;T&dquo; elements assigned here create instances for all the &dquo;children&dquo; of the specified activity class, along with predefined precedence relationships (e.g the STAMPING template in Figure 5). Other element types denote different product-activity relationships. Some product-activity relationships can be created automatically, such as for management reviews for

capital expenses specified by existing corporate policy. Elements in the activity-resource matrix can be assigned automatically by searching the resource database (e.g a certain tooling-engineer resource instance may have striplayout as a value on its list of performable activity classes) Multiple resources can be assigned, after which cost rates for labor, machines, and overheads, etc., are aggregated and default time distribution parameters are assigned as attributes of the instantiated activity. Alternately, activityresource matching can be done interactively by perusing the resource hierarchy [As one potential application, a manager might want to compare tooling and ramp-up costs at two different plants within a multinational organization After selecting the plant (a high-level resource), lower-level human and machine resources are automatically assigned ] In this manner, a proposed product configuration can be very quickly matched to a large, diverse set of activity and resource instances for proof-of-concept design, production prototyping, assembly design and tooling, and other activity classes. The result at this point is not yet a single, unified activity network, but rather a set of independent activity subsets with precedence relationships. As discussed in Section 4, an initial configuration of these subsets into an aggregate

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

55

Figure

5.

Defining product-activity relationships

network is automatically generated using some simple heuristics about design/tooling precedence and slack time considerations Activity subgraphs are displayed as Gantt-type &dquo;blocks&dquo; (Figure 6), and branching probability values for design iteration loops can be assigned At this point, a simulation can be run to evaluate preproduction cash flow (Figure 6, upper right corner) using either a single run with nominal values (e g mean values from the work and lead time p d f s) or with multiple runs using a random number generator to collect cost and time uncertainty information, displayed as

frequency plots. The time and cost implications of possible design &dquo;iteration&dquo; and changes in activity concurrency can be examined by varying the model parameters For example, a tooling activity subset for a stamped component can &dquo;overlap&dquo; testing for a production prototype of its parent subassembly (such as in Figure 4, with increased cost implications if design iteration occurs), or these activities can be assigned sequentially Because of the explicit relational model for design problem data, the cost and time results can be inspected in many useful ways by product element (for value engmeering), by activity class (for activity-based costing and resource (for departmental usage scheduling), and by particular cost category (labor, non-recurring, overhead, strategic capital investment, etc )

overhead assignment), by and allocation,

6.

Applications

for

design management

Field research for this study took place in several older, relatively small-sized manufacturers of metal and plastic

in

implementation

assemblies which have attributes that seem common to a large class of US companies Marketing, manufacturing, design, and finance are &dquo;under one roof&dquo;, but in distinct departments which have traditionally passed product information &dquo;over the wall&dquo; between personnel with distinct functional responsibilities There are strong skills in product and manufacturing engineering, but these skills serve primarily to support on-going manufacturing within an organizational structure that has long viewed &dquo;design&dquo; as small, incremental changes to existing product lines. Increasingly, they have found their traditional markets eroded by foreign competitors who aggressively introduce new, innovative products with low cost and comparable or even superior quality Managers, engineers, and line workers within this manufacturing &dquo;culture&dquo; are slow to respond to or even acknowledge this challenge; however, there are typically at least one or two top managers who are trying to spearhead innovative product and process design. Existing costing and planning methodologies within these companies offer little to help managers evaluate and communicate the profit potential of a new design, in terms of uncertainty of cost, development time, and market acceptance. This model was constructed specifically with the needs of this class of company in mind; four of the observed &dquo;decision scenarios&dquo; for design management at these companies are briefly described below as addressed by the model implementation, using several small example design problems Decision scenario 1: development cost estimation with uncertainty Budgeting for design and prototyping activities is notoriously difficult Typically, finance managers want &dquo;hard&dquo; numbers similar to those available for production operations. However, as discussed above, realization cost

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

56

Figure 6. Network configuration (as Gantt chart) and cash flow output

and time is inherently uncertain, any many engineering managers either artificially pad their budgets for unplanned costs (sometimes due to design iteration) or risk criticism for cost overruns. Figure 7 shows frequency plots for cost and time from a simulation for the beard trimmer subassembly example in Figures 5 and 6. Such output might be used better to communicate budget uncertainty between engineers and financial managers within the firm, as well as compare distributions of aggregate cost estimates between alternate design configurations Decision scenario 2’ schedule negotiation between early design and &dquo;downstream&dquo; engineering groups. Because of intense pressures to reduce overall time-to-market, minimal time is often allocated for early engineering design and analysis activities prior to prototype build and production. However, additional time spent &dquo;up front&dquo; in the process may in fact preclude later redesign and tooling rework problems. Typically, conflicts arise over schedule allocation: manufacturing engineering demands early delivery of a complete, detailed design to facilitate complex tooling and process design activities, while product engineers demand additional time for design refinement. In Figure 8, a total realization time vs branching probability for a &dquo;major&dquo; design iteration is plotted for a small hinge assembly of four stamped parts. Using the slope of a &dquo;best fit&dquo; line, a manager might decide, for example, that an additional 2 weeks of early design effort is justified if it would reduce the possibility of a major design iteration by at least 30%.

in

implementation.

Decision scenario 3: concurrency of product design and tooling. Management would like to &dquo;overlap&dquo; the tooling processes concurrently with design as early as possible to meet a perceived time-to-market window for the previously discussed beard trimmer. However, past experience indicates the possibility that if a major redesign is required after tooling has begun, the tooling will have to be scrapped or reworked. The graph in Figure 9 plots development time vs expected cost, where each curve shows varying degrees of design/tooling &dquo;overlap&dquo; for a given design iteration probability value P. This additional expected cost of compressing the development time might then be compared against the expected benefit of increased revenues from an earlier timeto-market. [Aside from this concurrency-related cost, there are two types of costs which are diametrically affected by time-to-market; (1) if the development cycle is compressed then, in many cases, resource usage costs will increase due to overtime, extra charges for early delivery, etc.; (2) if the development cycle is delayed, then the costs of maintaining production set ups, idle work force, and general overhead costs must be considered.] Decision scenario 4: project priorization and lead time sensitivity. In at least one company observed, a prioritized list of on-going projects is circulated to design, manufacturing, and other departments. For some support activities such as drafting and engineering change order (ECO) requests, new product projects directly &dquo;compete&dquo; for priority with on-going minor product variant efforts. If, for example, an

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

57

Figure 9. Expected cost vs. development tion probability values (p).

7. Histograms of total simulation.

Figure

Figure

8.

Design

development

time and cost from

time/iteration risk trade-off.

time for four

design itera-

ECO manager has a stack of different product change requests on his or her desk, the one with the highest priority is worked on first. This prioritized list may change as often as every week in response to management perception of delivery commitments. In a system geared towards processing of paper work related to on-going production, it would be useful for a manager to show quantitatively how changing the prioritization for a new product effort would effect its total development time. This type of prioritization is related to lead times in the model for affected activities such as drafting, process planning, vendor specification, and other operations. Parameters of lead time distributions can be varied for successive simulation runs to show the potential effect of a change in prioritization. Figure 10 shows a plot of total development time vs. a per cent change in the expected lead time for affected activities of a proposed injectionmolded component in a small firearm. In this example, the expected total development time could be shortened by almost 50 days if increased project priority could reduce each affected lead time by 50%.

Figure 10. Total development time vs. lead times of affected activities.

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

per cent reduction in

expected

58

These and other examples from the implementation were demonstrated to several industry product managers at companies which participated in the field studies. Their responses were generally encouraging, though several obstacles to practical application were noted which helped form our conclusions described below.

7. Conclusions The research described here explores a vexing design management issue for many US manufacturers: estimating the aggregate cost and time-to-market implications of decisions made at the early design stage After studying product realization data and related decision-making issues at several manufacturers, a model was developed which integrates several straightforward concepts (knowledge-based design representations, activity networks, relational data matrices) to help evaluate preproduction cash flow The model’s tractability as a computer-based decision support tool was then demonstrated in a proof-of-concept implementation. We conclude that this model is potentially viable as a basis for a management support tool and identify several remaining conceptual and implementation problems Most concerns relate to identification and appropriate representation for product realization activities While product and resource representations for this model are fairly simple, and many process realization activities and their procedural relationships (e g tooling activities) can be relatively easily classified, design-related product realization is not always easy to define as a sequence of discrete activities, particularly within the context of concurrent engineering For example, tolerancing of production part designs may be done partially by a design engineer, partly by a draftsperson, and partly by manufacturing personnel at different stages prior to fabrication Also, defining product realization activities for this type of model can be exacerbated by the tightly compartmentalized culture of many manufacturing firms, particularly the smaller, mature companies such as those studied for this research While much data for product attributes and resources is readily accessible (e g labor and machine rates, dimensioned blueprints, company organization charts), activity data is more difficult to obtain due to a lack of definition for many engineers and managers of how their individual roles relate to new product development within the company Also, as project managers have pointed out, despite the variety of established techniques for subjective estimation of probability distributions for cost/ time in network models [12], their consistent application for practical use can be difficult. There is a clear need for quick, preliminary cost and time estimation methods that capture uncertainty information, since most current practices for manufacturing costing typically require detail drawings in hand (or comparison with &dquo;just like&dquo; parts already in production), which precludes meaningful early feasibility studies for innovative designs Despite-these limitations, this model serves as a starting point to investigate new computer-based management tools made possible partly by new engineering design methodologies It is well worthwhile for management researchers to note advances in DFM cost methodologies and knowledgebased CAD representations, and the long-term implications

for

design management decision-support tools For example, the potential integration of feature-based design methodologies and activity-based costing techniques, particularly with respect to the sensitivity of indirect costs to product variation, could be quite powerful Ultimately, of course, one would like to extend this type of model to explore &dquo;profit uncertainty &dquo; By parameterizing a revenue flow such as described by Haffner and Graves [10] (e g slope, width, and height of the revenue curve dependent on time-tomarket, product quality, and features, etc.), and combining it with cash flow for full-life cycle costs, a &dquo;profit uncertainty distribution&dquo; could be characterized. Other possible applications might include using the model’s activity representations to help assign activity precedence constraints in a steward diagram [9]; cost breakdowns for components and subassemblies for value engineering analyses; and investigation of the indirect cost effects of number of parts Of chief interest is the model’s potential for expediting feasibility studies for both new products and new development processes, particularly in mature manufacturing industries This research was driven by observed difficulties in US manufacturing organizations to predict and measure the cost, quality, and time-to-market impact of product and organizational innovations that cross parochial department &dquo;walls&dquo; [5]. These walls-or at least their magnitude-seem to be in some respects peculiarly American, and the consequent enstrangement of managers and engineers-and design-related engineering and management research in academia-is perhaps near the heart of many US competitive problems

References [1]

[2] [3] [4]

[5] [6] [7]

[8]

[9] [10]

H B Bebb, "Quality Design Engineering The Missing Link in U S Competitiveness", keynote address to the 1989 National Science Foundation Engineering Design Conference Amherst, Massachusetts, 11-14 June 1989 K Clark and T Fujimoto, "Overlapping Problem-Solving in Product Development", Managing International Manufacturing, Elsevier, New York, 1989 D Clausing and S Pugh, "Enhanced Quality Function Deployment", Proceedings, Design Productivity International Conference, Honolulu, Hawaii, 6-9 February 1991 J R Dixon, M R Duffey, RI Irani, K L Meunier, and M O Orelup, "A Proposed Taxonomy of Mechanical Design Problems", Proceedings, ASME Computers in Engineering Conference, San Francisco, California, 31 July-3 August 1988 J R Dixon and M R Duffey, "The Neglect of Engineering Design", California Management Review, Berkeley, California, Winter 1990 J R Dixon and D W Rosen, "Languages for Feature-Based Design and Manufacturability", International Journal of Systems Automation Research and Applications, 1992 M R Duffey and Q Sun, "Knowledge-Based Design of Progressive Stamping Dies", Journal of Material Processing, Vol. 28, No 1-2, September 1991 S E Elmaghraby, Activity Networks Project Planning and Control by Network Models, John Wiley & Sons, New York, 1977 S D Eppinger, D E Whitney, R P Smith, and D A Gebala, "Organizing the Tasks in Complex Design Projects", ASME Design Automation Conference, Chicago, September 1990 E W Haffner and R J Graves, "Managing New Product Timeto-Market Using Time-Cost Trade-Off Methods", International Journal of Management Science, Vol 16, No 2, 1988

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010

59

[11]

[12]

P. Mahajan, "Design for Stamping Estimation of Relative Die Cost for Stamped Parts", Master’s Thesis, Department of Mechanical Engineering, University of Massachusetts, Amherst, Massachusetts, 1991. R B Stewart and R M Wyskida, Cost Estimator’s Reference Manual, John Wiley & Sons, New York, 1987

Duffey is an Assistant Professor of Engineering Management in the School of Engineering and Applied Science at the George Washington University in Washington, DC. He received his Ph D in Mechanical Engineering at the University of Massachusetts at Amherst His current research interests include design methodologies and computer tools for product development Michael R

[13]

K T Ulrich, D Sartorius, S Pearson, and M Jakiela, "A Framework for Including the Value of Time in Design-for-Manufacturing Decision Making", MIT Working Paper No. 3243-91-MSA, February 1991

Professor Dixon is a Registered Professional Engineer in Massachusetts, and a Fellow of the American Society of Mechanical Engineers (ASME) He has consulted on design and innovation for IBM, Scott Paper, and the US Department of Transportation. He served as Program Director for Engineering Design at the National Science Foundation (NSF) in 1988-1989 and also on the National Research Council Committee on Engineering Design during 1989-1990 Since 1983, Dr Dixon has been Director of the Mechanical Design Automation Lab at the University of Massachusetts at Amherst where engineering design research has been funded by NSF and a consortium of industrial firms

Downloaded from http://cer.sagepub.com at University of Groningen on July 1, 2010