Idea Transcript
The International Journal of Flexible Manufacturing Systems, 12 (2000): 207–217 c 2000 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands. °
Manufacturing Cost Modeling for Product Design ANGELA LOCASCIO Supply Chain Operations Group, Motorola, Schaumburg, Illinois
Abstract. The process of product design is driven toward achieving design specifications while meeting cost targets. Designers typically have models and tools to aid in functional and performance analysis of the design but few tools and little quantitative information to aid in cost analysis. Estimates of the cost of manufacture often are made through a cost multiplier based on material cost. Manufacturing supplies guidelines to aid in design, but these guidelines often lack the detail needed to make sound design decisions. A need was identified for a quantitative way for modeling manufacturing costs at Motorola. After benchmarking cost modeling efforts around the company, an activity-based costing method was developed to model manufacturing cycle time and cost. Models for 12 key manufacturing steps were developed. The factory operating costs are broken down by time, and cost is allocated to each product according to the processing it requires. The process models were combined into a system-level model, capturing subtle yet realistic operational detail. The framework was implemented in a software program to aid designers in calculating manufacturing costs from limited design information. Since the information tool provides an estimate of manufacturing costs at the design prototype stage, the development engineer can identify and eliminate expensive components and reduce the need for costly manufacturing processing. Using this methodology to make quantitative trade-offs between material and manufacturing costs, significant savings in overall product costs are achieved.
1.
Introduction
Although the majority of a product’s cost, typically about 80%, is determined early in the design stage, many decisions about the design are made during this stage with little knowledge of the effect on downstream cost centers. Manufacturing costs, in particular, are difficult to estimate and depend on many factors. Design decisions that affect the cost to manufacture the final product often are based on rules of thumb or the urging of experienced manufacturing engineers. Several models attempt to quantify the “manufacturability” of a design. The popular Boothroyd-Dewhurst index, for example, builds an estimate of design manufacturability relative to factors such as assembly complexity and number of parts (Boothroyd, Dewhurst, and Knight, 1991; Boothroyd and Dewhurst, 1983). Other models attempt to quantify design for X metrics to guide design decision making (Thurston and Locascio, 1994) or model trade-offs between design goals (Otto and Antonsson, 1991). These methods provide an assessment of the worth of the overall design but, in their effort to remain generally applicable, do not necessarily capture the economic aspects of the design with the rigor needed for design decision making. More often, these methods ask designers not only to be experts in the technical aspects of design but to understand how the design may affect other aspects in the product’s life cycle. Sullivan (1991) noted that a paradigm shift is occurring in engineering economy as
208
ANGELA LOCASCIO
a result of the “engineer’s role in strategic and design-related decision processes.” Indeed, new approaches to addressing economic concerns in the design process are needed. Wilhelm and Parsaei (1992) suggest that the role of “nonquantifiables” in engineering economics needs more attention and note that promising approaches include the analytic hierarchy process and multiple-criteria decision models. Examples of successful applications of engineering economic and decision analysis include a design of a mechanical turnbuckle (Thurston and Locascio, 1994) and a case study of disk manufacture for airplane turbine engines (Park and Prueitt, 1990). Indeed, information technology approaches to integrating engineering economy into the design process are essential. This paper demonstrates a simple yet effective model for design decision making as it relates to manufacturing cost. Providing data to the designer as early as possible through information technology enables better product design. The motivation for such a framework is described in the next section. Detail on the specific manufacturing scenario, electronics assembly, is explained in section 3. An example design application is provided in section 4, followed by some observations about its implementation at Motorola. 2.
Estimating manufacturing costs for design
The design process usually places priority on meeting functional specifications and achieving performance goals associated with the technical aspects of the product design. During the early design stages, designers meet with representatives responsible for other aspects of the product realization process to discuss the design as it relates to other organizations, such as manufacturing, supplier support, and marketing. During these meetings, feedback usually is given the designers to guide the design in a way that minimizes the cost of bringing the product to market. Designers are quite willing to incorporate the feedback provided about the designs but need this information presented in a way that easily can be understood and compared with the other design metrics relating to specifications and materials. In the design review meetings, the manufacturing engineers review the proposed designs and suggest changes to aspects of the design that are thought to drive up manufacturing cost, quality, and cycle time. In these meetings, the manufacturing engineers usually point out the aspects of the design that are particularly troublesome, citing that these components will be too expensive to assemble. Often these components were selected by the designer because they meet functional specifications and have low material cost. Expensive processing, however, may offset the benefit of using inexpensive materials. Since product design and redesign are driven by cost reduction, knowing the total cost to produce the product is essential for making informed decisions at the design stage. Adding more components or another processing step most likely adds cost—but how much? The new product engineer and the development engineer need to quantify costs such as: • • • •
How much does adding a processing step increase manufacturing costs? What is the cost of building this new technology product versus the old? Is it more expensive to use one complicated assembly or several simple ones? What is the cost of adding another unique part to this assembly?
MANUFACTURING COST MODELING FOR PRODUCT DESIGN
209
• Will manufacturing cost increase by replacing one expensive component with four inexpensive ones? Although the factory engineers had some qualitative ideas about the answers to these types of questions, they usually could not provide definite quantitative answers. Providing quantitative data on how much additional processing will cost allows the designer to decide if a less expensive part truly is less expensive when integrated with the product. There exists an opportunity to bring some engineering economic principles to practice in providing quantitative answers to these design questions. In an effort to construct a suitable quantitative model, a benchmarking study was conducted of cost modeling tools used internally at Motorola. One conclusion was that the most complete and successful tools focused on the manufacturing processes in an activitybased fashion (Santina, 1996). We therefore decided on an activity-based costing (ABC) approach, where we focused specifically on the major factory processes, or activities, that affected the development and manufacturing questions posed. In a typical ABC analysis, the operations usually are decomposed into small steps, as in motion studies. In this application, we decompose the manufacturing process into only those activities that directly affect the design issues. The approach we take is to examine a particular printed wiring board and focus on the processing it requires. Costs are assigned according to the time and human resources consumed at each process step. As the board acquires additional processing, more costs are incurred. Based on the factory time consumed and the cost associated with factory operation over that time period, costs are allocated to the design. For example, in a typical surface mount manufacturing line that assembles components on the printed wiring boards used in most electronics, the factory processes may be arranged as in figure 1. The processes shown constitute the majority of the processing, resources, and cycle time associated with assembling the printed wiring board. Several processes, such as buffers and bar code readers, are not shown. The contribution of these processes to the overall manufacturing flow and cost is considered negligible in terms of their effect on the design issues. The processes considered for this analysis are summarized in table 1. Note that not all processes are required for every product. The designer’s selection of components determines the specific process routing. 3.
Application to electronics assembly
The 12 manufacturing processes detailed in table 1 represent the activities needed to be modeled individually and as a system working together to manufacture the product. For the individual processes, models of the factory time and human resources consumed are developed first. The process models must have enough detail to enable calculation of product cost to the individual component level. At the manual placement process, for example, a printed wiring board arrives along the conveyor at the manual placement station. The factory operator stops the board and places components, one at a time, designated manual placement, in the specified location
210
ANGELA LOCASCIO
Figure 1. Typical automated electronics assembly line: The solid line is a conveyor; the dashed line represents flow for batch processing.
on the board. The parameters describing the process are ts4 and t p4 , where ts4 is the setup time associated with arranging the station and acquiring components designated as manual place (process 4), and t p4 is the processing time required to place all components as designated. The processing time is the product of the time to place one component and the quantity of manually placed components. The human resource associated with this process is simply the number of associates required to staff the process, s4 . Some designs may have many manually placed components and therefore require several factory associates. Similar models are developed for each of the 12 individual processes to represent the cycle time (ts = {ts1 , ts2 , . . . , ts12 } and t p = {t p1 , t p2 , . . . , t p12 }) and staffing (s = {s1 , s2 , . . . , s12 }). The system level model is considered next. To represent the overall manufacturing system, several factors must be considered: • Conveyorized processing (“flowline”) for some processes.
MANUFACTURING COST MODELING FOR PRODUCT DESIGN Table 1.
211
Electronics assembly processes.
Process
Function
Screen print
Apply solder paste on the printed wiring board
Small component placement
Automation to place small components on the printed wiring board; may be more than one station
Large component placement
Automation for placing large, high-precision parts on the printed wiring board; may be more than one station
Manual placement
Placement of components of odd shape or delicate nature on the printed wiring board; may be more than one operator at this station
Inspection
Automated inspection station to verify placement of components
Reflow
High-temperature oven to complete the connection of the components to the solder
Manual insertion
Manual insertion of components on printed wiring board
Wave solder
High-temperature connectivity for manually inserted components
Hand solder
Manual soldering of components that cannot withstand temperatures of reflow or wave solder processing
Test 1
Verify performance and functionality of product
Test 2
Verify performance and functionality of product
Final manual placement
Final assembly of other components not required at test
• Batch processing for other processes. • Process routings unique to each design (i.e., one product may require manual placement of some components, while another product may have all automated placement of components). • Setup and processing times may not be balanced across all processes, so the gating or “bottleneck” process must be considered. • Yield at each process step, inspection point, or test station. To calculate the fixed and variable costs associated with building a product, the facility and labor resources required to assemble the product are required. The variable costs represent the cost of the factory operators and technicians supporting the factory during the time that the product is manufactured. The fixed costs represent the portion of the committed manufacturing costs, overhead, and support staff (indirect labor) consumed during product assembly. 3.1.
Factory changeover and setup
Each automated component placement machine must be configured with the correct software program and stocked with the correct materials to assemble the printed wiring board. Changing over the machine requires downloading software, arranging part feeders, and verifying the feeder setup. Manual stations require the operator to locate the parts, obtain the correct assembly instructions, and arrange the assembly station. Similar operations are required to prepare the other processes. All processes are set up simultaneously and the production assembly is not initiated until every process is ready. The total time that the
212
ANGELA LOCASCIO
factory is not in production due to the factory changeover is the setup time of the process that takes the longest to prepare, or ts = max{tsi } for i = 1, 2, . . . , 12. The fixed and variable costs associated with setup, C f s and Cvs , respectively, are then calculated as C f s = Fts Cvs = Lts where L is the loaded hourly labor rate and F is the hourly adjusted fixed operating costs (includes committed costs and indirect labor, primarily). The loaded labor rate refers to a typical hourly wage for a factory operator (often called direct labor) and includes the common expenses that a company incurs for that employee, such as insurance and retirement benefits. The term hourly adjusted fixed operating expenses is calculated from the operating expenses over a given time period (typically tracked monthly) divided by the production hours available over that time period. 3.2.
Factory processing
The factory processing time is the actual time that the product spends on the assembly line in production. We assume that production begins immediately after factory setup and concludes when the desired quantity is completed. As with factory setup, we can calculate the fixed and variable costs from C f p = F Tp
12 X
si
i=1
Cvp = L T p where T p is the total processing time of a single unit, calculated by considering the total time consumed in building the product both in the flow-line portion of the assembly and in the batch processing. The total time that the product spends in the flow-line area is gated by the process with the greatest processing time. The processing time in the flow-line portion therefore is given by tflow = Q max ti where Q is the quantity of units in a standard build lot size and i = 1, 2, . . . , 6 (i.e., a process only in the flow line). The product then proceeds through the batch processes and the processing time through this area is the sum of the processing times of each station, as given by tbatch = Q
12 X i=1
t pi
MANUFACTURING COST MODELING FOR PRODUCT DESIGN
213
The total processing time then is found from T p = tflow + tbatch 3.3.
Process cost
The total cost of each unit can be thought of as a compilation of costs incurred at each process. A product design that requires more processing steps, time, and labor resources will cost more than a design requiring less. We need a way to decompose the product cost in terms of the cost incurred at each process. This level of detail will allow us to answer some of the questions posed by the design and manufacturing engineers. The percentage of resources consumed at each processes is given by the ratio T p /Tsystem , where Tsystem =
12 X
t pi
i=1
is the total time that one unit spends actually being acted on in the system. The process cost then is calculated from Cprocess = (T p /Tsystem )[(Cvs + C f s ) + (Cvp + C f p )]/Q where the estimated cost per unit is [(Cvs + C f s ) + (Cvp + C f p )]/Q 3.4.
(1)
Design issues—cost per component
The required process steps and processing time depend directly on the designer’s selection of components. The component-type selection dictates the machines and stations capable of assembling the product. The cost, ci j , attributed to a single component j at a single process i is estimated as the process cost at process i divided by the number of components affected at i. The total manufacturing cost associated with an individual component j is simply the sum of the ci j over all processes that affect component j. The unique processing defined for each design is completely represented by this model. This level of detail allows us to quantify the cost of manufacturing a particular design and feed back useful information early in the design process. 4.
Example
A designer developed a preliminary design layout and is considering making several changes to reduce overall product cost. The designer would like to replace several of the small components (resistors and capacitors) with an integrated circuit that may have a higher
214
ANGELA LOCASCIO Table 2.
Bill of materials for Design A.
Component Type Small chips: Resistors, capacitors, and the like Large parts: Integrated circuits, shields, and the like
Quantity 324 42
Hand-soldered parts
1
Final manual assembly
3
material cost than the equivalent small components but will provide space savings on the board and possibly save manufacturing cost. In addition, there is a need for one “odd part,” which is inexpensive but requires hand soldering during manufacture. The component supplier offers a functionally equivalent version of the same part that can be assembled by the automatic placement equipment; this version of the part is much more expensive, however, than the hand-soldered version. The original printed wiring board design, Design A, has a bill of materials (BOM) that may be summarized as shown in table 2. The estimated manufacturing cost per unit from equation (1) for this design is $85.50. As detailed in section 3, this estimate constitutes the effect of both setup and processing costs associated with each process affected in the factory for building this design. The cost of each process is a function of labor and overhead rates. The processes affected are determined by the component selections of the designer. This estimate represents the manufacturing cost only and does not include any materials associated with the product. Making the proposed design changes results in a change in the process steps, process time, and required labor. Using equation (1), the cost per unit for this proposed design becomes $77.86, a savings of $7.64 or about 9%. Noting the estimated savings in manufacturing cost, the designer can make the appropriate trade-off between the material and manufacturing costs to find the overall lowest-cost design. If the designer would like to make other changes, such as adding functionality by adding more component circuitry to the printed wiring board, the effect on manufacturing cost depends on several factors, including the process bottleneck and the design’s distribution of components over the processes. For example, since the process bottleneck is the gating factor in the product build consumption of production resources, increasing process time at that step (by adding components) increases overall assembly time. Adding processing at a nonbottleneck process, on the other hand, may not affect the overall assembly time unless that additional processing shifts the bottleneck. To help the manufacturing engineer understand how this analysis could benefit manufacturing operations, consider figure 2. This chart shows the distribution of costs over the processes required for Design A. The process cost analysis aids the manufacturing engineer in identifying the key cost drivers in the factory and in anticipating additional resources that will be required when this design advances from prototype design to full production. From this chart the factory engineer, for example, can observe that the test time associated with this design is the single largest component of the overall processing time. Depending on the volume of the product, the manufacturer may want to add one or more test stations
MANUFACTURING COST MODELING FOR PRODUCT DESIGN
Figure 2.
215
Cost per process.
to improve the throughput. Note also from this chart that, although the hand soldering process represents only a small portion of the total cost, it is created by only one component. Elimination of this part eliminates the need for an entire process step and its staffing. The manufacturing engineer also can use this framework to analyze modifications to the design that are thought to reduce manufacturing costs. Typically, the factory tooling and conveyor system is configured to accept a standard-width printed wiring board. For small products such as consumer electronics, several units may be arranged on a single printed wiring board during fabrication, then separated at the end of the line into individual units. The manufacturing engineer can analyze various arrangements and quantities of units of the printed wiring board, allowing quantitative trade-offs between manufacturing cost savings and tooling expense required to support the changes. Consider the design example, Design A. The manufacturing engineer thinks that arranging this design with two units on a single printed wiring board will save setup time and speed production. Using the preceding method, the cost per unit for the “two-up” design is $89.50, an increase of $4.00 over the original design. Although it was thought that building multiple units on the same printed wiring board would save time and expense, for this design, the result actually adds time and cost per unit.
5.
Conclusion
The manufacturing cost estimation framework described here emphasizes two key points in application to design cost modeling: 1. The level of detail was analogous to that required to address the questions posed by the designer.
216
ANGELA LOCASCIO
2. The activity models were applied only to the portions of the design process that required it, not to every process in the factory nor to the level of detail of typical ABC studies. For this reason, some manufacturing considerations, including quality and reliability, were not considered explicitly. It was decided that the detail required for these models would outweigh the incremental benefit to design improvement. The result is that the model is quite specific in addressing the design questions but does not try to model the manufacturing process to the level of detail of typical activity-based costing models. For design cost modeling, the typical ABC detail level may add too much complexity to a problem that needs it only in certain aspects. And the designer can identify what those aspects are. Although the framework was developed here for electronics assembly, the key steps can generalize to other product design and manufacturing scenarios as well. We already successfully have generalized this approach to other manufacturing operations at Motorola. The methodology described in this paper has been implemented in software to facilitate easy calculation of manufacturing cost estimates. Design and manufacturing engineers were part of the team that developed this tool and their contributions were integral to its successful implementation in the design and manufacturing organizations at Motorola. The research, design, and implementation of this tool comprised several iterations over a period of about two years. A team of about 10 engineers were directly involved in the development and several dozen design engineers participated in evaluation and testing. Because design engineers participated on the development team, the tool was very well received by the design community and adopted immediately into the standard design process. The use of this tool at each stage in the design has become a new product introduction requirement, with the estimates of manufacturing cost reported as a key metric for the design prototypes. The equations, embedded in the software, are transparent to the design and manufacturing engineers using them. No detailed understanding of engineering cost analysis is required to benefit from the results this tool provides. This tool generates the quantitative proof for the intuition that design and manufacturing engineers have for cost improvements. The software uses a bill of materials as input and performs the calculations shown here. This methodology allows design and manufacturing engineers to quantify the impact of design decisions on manufacturing. The impact of this design economic tool is estimated to save Motorola several million dollars in overall product costs annually. Acknowledgments I gratefully acknowledge the contributions of Tom Babin, Brad Bakka, Dan Flondro, Jim Hermann, and Anil Singh to the development of the design cost model and implementation at Motorola. References Boothroyd, G. and Dewhurst, P., “Design for Assembly: Manual Assembly,” Machine Design, pp. 140–145 (December 1983). Boothroyd, G., Dewhurst, P., and Knight, W. A., “Selection of Materials and Processes for Component Parts,” Proceedings of the 1992 NSF Design and Manufacturing Systems Conference, pp. 255–263 (1991).
MANUFACTURING COST MODELING FOR PRODUCT DESIGN
217
Otto, K. N. and Antonsson, E. K., “Trade-off Strategies in Engineering Design,” Research in Engineering Design, Vol. 3, pp. 87–103 (1991). Park, C. S. and Prueitt, G. C., “Evaluating a New Technology Alternative: Case Study,” The Engineering Economist, Vol. 36, No. 1, pp. 31–54 (Fall 1990). Santina, P., “DFM Meets ABC,” Circuits Assembly (September 1996). Sullivan, W. G., “A New Paradigm for Engineering Economy,” The Engineering Economist, Vol. 36, No. 2, pp. 187–200 (Spring 1991). Thurston, D. L. and Locascio, A., “Decision Theory for Design Economics,” The Engineering Economist, Vol. 40, No. 1, pp. 41–72 (Fall 1994). Wilhelm, M. R. and Parsaei, H. R., “‘Irreducible’ Analysis by Use of Fuzzy Linguistic Variables,” First Industrial Engineering Research Conference Proceedings, Chicago, pp. 37–39 (May 20–21, 1992).