Manufacturing by Design

In many large companies, design has become a bureaucratic tangle, a process confounded by fragmentation, overspecialization, power struggles, and delays. An engineering manager responsible for designing a single part at an automobile company told me recently that the design process mandates 350 steps—not 350 engineering calculations or experiments but 350 workups requiring 350 signatures. No wonder, he said, it takes five years to design a car; that’s one signature every 3½ days. It’s not as if companies don’t know better. According to General Motors executives, 70 % of the cost of manufacturing truck transmissions is determined in the design stage. A study at Rolls-Royce reveals that design determines 80 % of the final production costs of 2,000 components.1 Obviously, establishing a product’s design calls for crucial choices—about materials made or bought, about how parts will be assembled. When senior managers put most of their efforts into analyzing current production rather than product design, they are monitoring what accounts for only about a third of total manufacturing costs—the window dressing, not the window. Moreover, better product design has shattered old expectations for improving cost through design or redesign. If managers used to think a 5 % improvement was good, they now face competition that is reducing drastically the number of components and subassemblies for products and achieving a 50 % or more reduction in direct cost of manufacture. And even greater reductions are coming, owing to new materials and materials-processing techniques. Direct labor, even lower cost labor, accounts for so little of the total picture that companies still focusing on this factor are misleading themselves not only about improving products but also about how foreign competitors have gained so much advantage. In short, design is a strategic activity, whether by intention or by default. It influences flexibility of sales strategies, speed of field repair, and efficiency of manufacturing. It may well be responsible for the company’s future viability. I want to focus not on the qualities of products but on development of the processes for making them. Converting a concept into a complex, high-technology product is an involved procedure consisting of many steps of refinement. The initial idea never quite works as intended or performs as well as desired. So designers make many modifications, including increasingly subtle choices of materials, fasteners, coatings, adhesives, and electronic adjustments. Expensive analyses and experiments may be necessary to verify design choices. In many cases, designers find that the options become more and more difficult; negotiations over technical issues, budgets, and schedules become intense. As the design evolves, the choices become interdependent, taking on the character of an interwoven, historical chain in which later decisions are conditioned by those made previously. Imagine, then, that a production or manufacturing engineer enters such detailed negotiations late in the game and asks for changes. If the product designers accede to the requests, a large part of the design may simply unravel. Many difficult and pivotal choices will have been made for nothing. Where close calls went one way, they may now go another; new materials analyses and production experiments may be necessary. Examples of failure abound. One research scientist I know, at a large chemical company, spent a year perfecting a new process—involving, among other things, gases—at laboratory scale. In the lab the process operated at atmospheric pressure. But when a production engineer was finally called in to scale up the process, he immediately asked for higher pressures. Atmospheric pressure is never used in production when gases are in play because maintaining it requires huge pipes, pumps, and tanks. Higher pressure reduces the volume of gases and permits the use of smaller equipment. Unfortunately, the researcher’s process failed at elevated pressures, and he had to start over. Or consider the manufacturer whose household appliance depended on close tolerances for proper operation. Edicts from the styling department prevented designs from achieving required tolerances; the designers wanted a particular shape and appearance and would not budge when they were apprised of the problems they caused to manufacturing. Nor was the machine designed in modules that could be tested before final assembly. The entire product was built from single parts on one long line. So each finished product had to be adjusted into operation—or taken apart after assembly to find out why it didn’t work. No one who understood the problem had enough authority to solve it, and no one with enough authority understood the problem until it was too late. This company is no longer in business. Finally, there was the weapon that depended for its function on an infrared detector, the first of many parts—lenses, mirrors, motors, power supplies, etc.—that were glued and soldered together into a compact unit. To save money, the purchasing department switched to a cheaper detector, which caused an increase in final test failures. Since the construction was glue and solder, bad units had to be scrapped. Someone then suggested a redesign of the unit with reversible fasteners to permit disassembly. But this time more reasonable voices prevailed. Reversible fasteners would have actually increased the weapon’s cost and served no purpose other than to facilitate factory rework. Disassembly would not have been advisable because the unit was too complex for field repair. It was a single-use weapon—with a shelf life of five years and a useful life of ten seconds. It simply had to work the first time. Manufacturers can avoid problems like this. Let’s look at a success story. One company I know wanted to be able to respond in 24 hours to worldwide orders for its electronic products line—a large variety of features in small-order batches. Engineers decided to redesign the products in modules, with different features in each module. All the modules are plug compatible, electrically and mechanically. All versions of each module are identical on the outside where assembly machines handle them. The company can now make up an order for any set of features by selecting the correct modules and assembling them, all of this without any human intervention, from electronic order receipt to the boxing of final assemblies. In another company, a high-pressure machine for supplying cutting oil to machine tools requires once-a-day cleaning. Designers recently reconfigured the machine so that normal cleanout and ordinary repairs can be accomplished without any tools, thus solving some bothersome union work-rule problems. There are no guarantees, of course, but the experiences of these companies illustrate how design decisions should be integrated, informed, and balanced, and how important it is to involve manufacturing engineers, repair engineers, purchasing agents, and other knowledgeable people early in the process. The product designer asks, “What good is it if it doesn’t work?” The salesperson asks, “What good is it if it doesn’t sell?” The finance person asks, “What good is it if it isn’t profitable?” The manufacturing engineer asks, “What good is it if I can’t make it?” The team’s success is measured by how well these questions are answered. The Design Team and Its Task Multifunctional teams are currently the most effective way known to cut through barriers to good design. Teams can be surprisingly small—as small as 4 members, though 20 members is typical in large projects—and they usually include every specialty in the company. Top executives should make their support and interest clear. Various names have been given to this team approach, like “simultaneous engineering” and “concurrent design.” Different companies emphasize different strengths within the team. In many Japanese companies, teams like this have been functioning for so long that most of the employees cannot remember another way to design a product. Establishing the team is only the beginning, of course. Teams need a step-by-step procedure that disciplines the discussion and takes members through the decisions that crop up in virtually every design. In traditional design procedures, assembly is one of the last things considered. My experience suggests that assembly should be considered much earlier. Assembly is inherently integrative. Weaving it into the design process is a powerful way to raise the level of integration in all aspects of product design. A design team’s charter should be broad. Its chief functions include: 1. Determining the character of the product, to see what it is and thus what design and production methods are appropriate. 2. Subjecting the product to a product function analysis, so that all design decisions can be made with full knowledge of how the item is supposed to work and all team members understand it well enough to contribute optimally. 3. Carrying out a design-for-producibility-and-usability study to determine if these factors can be improved without impairing functioning. 4. Designing an assembly process appropriate to the product’s particular character. This involves creating a suitable assembly sequence, identifying subassemblies, integrating quality control, and designing each part so that its quality is compatible with the assembly method. 5. Designing a factory system that fully involves workers in the production strategy, operates on minimal inventory, and is integrated with vendors’ methods and capabilities. The Product’s Character Clearly it is beyond the scope of this article to establish by what criteria one judges, develops, or revamps the features of products. Recently in these pages, David A. Garvin has analyzed eight fundamental dimensions of product quality; and John R. Hauser and Don Clausing have explored ways to communicate to design engineers the dimensions consumers want—in the engineers’ own language.2 Character defines the criteria by which designers judge, develop, or revamp product features. I would only reiterate that manufacturing engineers and others should have something to say about how to ensure that the product is field repairable, how skilled users must be to employ it successfully, and whether marketability will be based on model variety or availability of future add-ons. An essential by-product of involving manufacturing, marketing, purchasing, and other constituencies in product conception, moreover, is that diverse team members become familiar enough with the product early in order to be able to incorporate the designers’ goals and constraints in their own approaches. As designers talk with manufacturing or field-service reps, for example, they can make knowledgeable corrections. (“Why not make that part out of plastic? I know a low-cost source.” “Because the temperature there is 1,000º; plastic will vaporize.” “Oh.”) Product Function Analysis This used to be the exclusive province of product designers. But now it is understood that to improve a product’s robustness, to “design quality in” in Genichi Taguchi’s good phrase, means thoroughly understanding a product’s function in relation to production methods. Product designers and manufacturing engineers used to try to understand these relations by experience and intuition. Now they have software packages for modeling and designing components to guide them through process choices—software that would have been thought fantastic a generation ago. (See the insert, “Designing for Predictability: New MCAE Tools.”) Designing for Predictability: New MCAE Tools by: Philippe Villers A well-designed product is a predictable product. Managers particularly need to predict reliability, manufacturing costs, and manufacturability. In the past, engineers have dealt with these three issues only after engineering has completed the drawings, the near-final stage in the development cycle. But by then it may be too late. Moreover, when a product has hundreds, or even thousands of components, it’s just not feasible to design, prototype, and redesign slightly different versions of each. Two new, integrated, mechanical computer-aided engineering (MCAE) systems permit engineering teams to test before they build, so they can design for total quality with reliability, performance, and manufacturing costs in mind from the start. The first of these new systems model products not just physically but analytically: a designer can not only draw a component on the screen but prompt the computer to model and test the feasibility of its features, in the manner of a business planner using an electronic spreadsheet to play “what if” games. Engineers can vary assumptions about materials, speeds, loads, size, and other operating conditions. In this way, developers can both see the effects of hypothetical stresses and estimate product costs while making design decisions. A company making internal-combustion engines, for example, may use an integrated MCAE system to design reliability and smoothness of operation into the counterbalance for the crankshaft. The optimal counterbalance keeps forces from the piston’s up-and-down motion and centrifugal forces from the unbalanced crankshaft both low and constant—to reduce vibration. The system works as follows: a desktop workstation paints an image of a cylinder with all its operating parts on the screen. The engineer then selects values for counterbalance features like angle, thickness, diameter, etc., from menus. As choices are made, the system automatically computes the merits of the design, based on about 100 engineering equations, including compression ratio and stroke. So design variations are tried, evaluated, and discarded with near instantaneous response. This puts robustness and performance optimization into the very first counter-balance designs. The second new system is an expert system which projects probable production costs for various part or assembly configurations and provides guidance as to their manufacturability. Another parts maker might use this system to project the manufacturability of and costs for, say, its stamped carburetor parts. Typically, the PC screen has several menus listing aspects of a part’s design, like materials used and manufacturing process involved. Each menu holds progressive layers of possible choices. If the engineer selects metal from the list of materials, the system offers a choice of ferrous or non-ferrous. Under ferrous metals, one can pick from carbon steel, stainless, cast iron, and so on. There are automatic default values that the engineer might not normally specify, such as surface finish and carbon content. The system also draws on its own data base for manufacturing information like material density and base unit cost. Once the designer completes the menu sequence, the system produces an approximate part cost that includes materials, processing, and tooling expenditures. MCAE systems don’t eliminate prototyping entirely, but they can drastically reduce the number of trials and help engineers address reliability, performance, and cost problems early—and thus design for total quality from the very beginning. Recently I worked on a product containing delicate spinning parts that had to be dynamically balanced to high tolerances. In the original design, partial disassembly of the rotating elements after balancing was necessary before the assembly could be finished, so the final product was rarely well balanced and required a lengthy adjustment procedure. Since total redesign was not feasible, the team analyzed the reassembly procedure solely as it pertained to balance and concluded that designers needed only to tighten various tolerances and reshape mating surfaces. Simple adjustments were then sufficient to restore balance in the finished product. Another important goal of product function analysis is to reduce the number of parts in a product. The benefits extend to purchasing (fewer vendors and transactions), manufacturing (fewer operations, material handlings, and handlers), and field service (fewer repair parts). When a company first brings discipline to its design process, reductions in parts count are usually easy to make because the old designs are so inefficient. After catching up, though, hard, creative work is necessary to cut the parts count further. One company I know saved several million dollars a year by eliminating just one subassembly part. The product had three operating states: low, medium, and high. Analysis showed that the actions of one part in the original design always followed or imitated the actions of two others. Designers eliminated the redundant part by slightly altering the shapes of the other two parts. This change could never have been conceived, much less executed, if the designers hadn’t had deep knowledge of the product and hadn’t paid attention to the actions underlying its engineering. Design for Producibility Recently, a company bragged to a business newsweekly about saving a mere $ 250,000 by designing its bottles for a new line of cosmetics to fit existing machines for filling, labeling, and capping. This plan seems so obvious, and the savings were so small as compared with what is possible, that the celebration seemed misplaced. But it’s a better outcome than I remember from my first job with a drug company. It spent a fortune to have a famous industrial designer create new bottles and caps for its line. They were triangular in cross section and teardrop shaped, and they would not fit either existing machines or any new ones we tried to design. The company eventually abandoned the bottles, along with the associated marketing campaign. Obviously, nothing is more important to manufacturing strategy than designing for the production process. In the past, this has meant designing for manufacturing and assembly, and value engineering, which both strive to reduce costs. But now we have to go beyond these goals. To take the last point first, value engineering aims chiefly to reduce manufacturing costs through astute choice of materials and methods for making parts. Does the design call for metal when a ceramic part will do? If metal, should we punch it or drill it? Value engineering usually comes into play after the design is finished, but the thoroughness we seek in design can be achieved only when decisions are made early. Moreover, design for producibility differs from design for assembly, which typically considers parts one by one, simplifies them, combines some to reduce the parts count, or adds features like bevels around the rims of holes to make assembly easier. Valuable as it is, this process cannot achieve the most fundamental improvements because it considers the product as a collection of parts instead of something to satisfy larger goals, such as reducing costs over the product’s entire life cycle. Nippondenso’s approach vividly illustrates how an overriding strategy can determine a product’s parts and the production process. The Delco of Japan, Nippondenso builds such car products as generators, alternators, voltage regulators, radiators, and antiskid brake systems. Toyota is its chief customer. Nippondenso has learned to live with daily orders for thousands of items in arbitrary model mixes and quantities. The company’s response to this challenge has several components: The combinatorial method of meeting model-mix production requirements. In-house development of manufacturing technology. Wherever possible, manufacturing methods that don’t need jigs and fixtures. The combinatorial method, carried out by marketing and engineering team members, divides a product into generic parts or subassemblies and identifies the necessary variations of each. The product is then designed to permit any combination of variations of these basic parts to go together physically and functionally. (If there are 6 basic parts and 3 varieties of each, for example, the company can build 36=729 different models.) The in-house manufacturing team cooperates in designing the parts, so the manufacturing system can easily handle and make each variety of each part and product. Jigless production is an important goal at this point, for obvious reasons. Materials handling, fabrication, and assembly processes usually employ jigs, fixtures, and tools to hold parts during processing and transport; the jigs and fixtures are usually designed specifically to fit each kind of part, to hold them securely. When production shifts to a different batch or model, old jigs and tools are removed and new ones installed. In mass-production environments, this changeover occurs about once a year. In dynamic markets, however, or with just-in-time, batches are small, and shifts in production may occur hourly—even continually. It may be impossible to achieve a timely and economical batch-size-of-one production process if separate jigs are necessary for each model. Nippondenso’s in-house manufacturing team responds to this problem by showing how to design the parts with common jigging features, so that one jig can hold all varieties, or by working with designers to make the product snap or otherwise hold itself together so that no clamping jigs are needed. By cultivating an in-house team, Nippondenso also solves three difficult institutional problems. First, the company eliminates proprietary secrecy problems. Its own people are the only ones working on the design or with strategically crucial components. Second, equipment can be delivered without payment of a vendor’s markup, thus reducing costs and making financial justification easier. Third, over the years the team has learned to accommodate itself intuitively to the company’s design philosophy, and individual team members have learned how to contribute to it. Designers get to know each other too, creating many informal communication networks that greatly shorten the design process. Shorter design periods mean less lead time, a clear competitive advantage. (It is worth noting that many Japanese companies follow this practice of designing much of their automation in-house, while buying many product components from outside vendors. American companies usually take the opposite tack: they make many components and buy automation from vendors.) Nippondenso uses combinatorial design and jigless manufacturing for making radiators (see the diagram). Tubes, fins, headers, and side plates comprise the core of the radiator. These four snap together, which obviates the need for jigs, and the complete core is oven soldered. The plastic tanks are crimped on. The crimp die can be adjusted to take any tank size while the next radiator is being put in the crimper, so radiators can be processed in any model order and in any quantity. When asked how much the factory cost, the project’s chief engineer replied, “Strictly speaking, you have to include the cost of designing the product.” A factory isn’t just a factory, he implied. It is a carefully crafted fusion of a strategically designed product and the methods for making it. How a Radiator Is Made—the Combinatorial, Jigless Method Except for final testing, this radiator is fabricated entirely without manual labor. Radiators differ by core length, width, and depth. They are available in various sizes and offer many heat-transfer capacities. Without changing part sizes, a designer can program different shapes for fins and diameters for tubes, thus allowing the same production system to achieve new heat-transfer capacities. Without a guiding strategy, there is no way to tell what suggestions for improvement really support long-range goals. Some product-design techniques depend too much on rules, including rule-based systems stemming from expert systems. These are no substitute for experienced people. Volkswagen, for example, recently violated conventional ease-of-assembly rules to capture advantages the company would not otherwise have had. In the company’s remarkable Hall 54 facility in Wolfsburg, Germany, where Golfs and Jettas go through final assembly, robots or special machines perform about 25 % of the final-stage steps. (Before Hall 54 began functioning, Volkswagen never did better than 5 % )3 To get this level of automation, VW production management asked to examine every part. It won from the board of directors a year-long delay in introducing the new models. Several significant departures from conventional automotive design practices resulted, the first involving front-end configuration. Usually, designers try to reduce the number of parts. But VW engineers determined that at a cost of one extra frame part the front of the car could be temporarily left open for installation of the engine by hydraulic arms in one straight, upward push. Installing the engine used to take a minute or longer and involved several workers. VW now does it unmanned in 26 seconds. Another important decision concerned the lowly screw. Purchasing agents usually accept the rule that low-cost fasteners are a competitive edge. VW engineers convinced the purchasing department to pay an additional 18 % for screws with cone-shaped tips that go more easily into holes, even if the sheet metal or plastic parts were misaligned. Machine and robot insertion of screws thus became practical. Just two years later, so many German companies had adopted cone-pointed screws that their price had dropped to that of ordinary flat-tip screws. For once, everyone from manufacturing to purchasing was happy. Assembly Processes Usually assembly sequence is looked at late in the design process when industrial engineers are trying to balance the assembly line. But the choice of assembly sequence and the identification of potential subassemblies can affect or be affected by—among other factors—product-testing options, market responsiveness, and factory-floor layout. Indeed, assembly-related activities with strategic implications include: subassemblies, assembly sequence, assembly method for each step, and integration of quality control. Imagine a product with six parts. We can build it many ways, such as bottom up, top down, or from three subassemblies of two parts each. What determines the best way? A balance of many considerations: construction needs, like access to fasteners or lubrication points; ease of assembly (some sequences may include difficult part matings that risk damage to parts); quality control matters, like the operator’s ability to make crucial tests or easily replace a faulty part; process reasons, like ability to hold pieces accurately for machine assembly; and, finally, production strategy advantages, like making subassemblies to stock that will be common to many models, or that permit assembly from commonly available parts. Again, software now exists to help the designer with the formidable problem of listing all the possible assembly sequences—and there can be a lot, as many as 500 for an item as simple as an automobile rear axle. It would be impossible for a team to attack so complex a series of choices without a computer design aid to help, according to a preestablished hierarchy of goals like that just discussed—access to lubrication points, etc. Another virtue of this software is that it forces the team to specify choices systematically and reproducibly, for team members’ own edification but also in a way that helps justify design and manufacturing choices to top management. Consider then, automatic transmissions, complex devices made up of gears, pistons, clutches, hydraulic valves, and electronic controls. Large transmission parts can scrape metal off smaller parts during assembly, and shavings can get into the control valves, causing the transmission to fail the final test or, worse, fail in the customer’s car. Either failure is unacceptable and terribly expensive. It is essential to design assembly methods and test sequences to preempt them. With respect to assembly machines and tooling, manufacturers should consider the following questions: Can the product be made by adding parts from one direction, or must it be turned over one or more times? Turnovers are wasted motion and costly in fixtures. As parts are added in a stack, will the location for each subsequent part drift unpredictably? If so, automatic assembly machines will need expensive sensors to find the parts, or assembly will randomly fail, or parts will scrape on each other too hard. Is there space for tools and grippers? If not, automatic assembly or testing aren’t options. If a manufacturing strategy based on subassemblies seems warranted, are the subassemblies designed so they do not fall apart during reorientation, handling, or transport? There are clear advantages to combining consideration of these assembly procedures and/or quality control strategy with design. Designers who anticipate the assembly method can avoid pitfalls that would otherwise require redesign or create problems on the factory floor. They can also design better subassemblies to meet functional specifications—specifications that will be invaluable when the time comes to decide whether to take bids from outside vendors or make the part on the company’s own lines, specifications that will determine how to test the subassembly before adding it to the final product. Designers concerned about assembly must ask: What is the best economic combination of machines and people to assemble a certain model-mix of parts for a product line (given each machine’s or person’s cost and time to do each operation, plus production-rate and economic-return targets)? How much time, money, production machinery, or in-process inventory can be saved if extra effort is put into design of the product, its fabrication and assembly processes, so that there are fewer quality control failures and product repairs? A process that yields only 80 % successful assemblies on the first try may need 20 % extra capacity and inventory—not to mention high-cost repair personnel—to meet the original production goals. Where in the assembly process should testing take place? Considerations include how costly and definitive the test is, whether later stages would hide flaws detectable earlier, and how much repaired or discarded assemblies would cost. These are generic problems; they are hard to answer, and they too are stimulating the development of new software packages. This new software enhances the ability of manufacturing people to press their points in (often heated) debates about design. Hitherto, product designers, more accustomed to using computer modeling, have had somewhat of an upper hand. Factory System Design Many features of good product design presuppose that machines will do the assembly. But automation is not necessary to reap the benefits of strategic design. Indeed, sometimes good design makes automatic assembly unnecessary or uneconomic by making manual assembly so easy and reliable. Regardless of the level of automation, some people will still be involved in production processes, and their role is important to the success of manufacturing. Kosuke Ikebuchi, general manager of the General Motors-Toyota joint venture, New United Motors Manufacturing Inc. (NUMMI), believes that success came to his plant only after careful analysis of the failures of the GM operation that had preceded it: low-quality parts from suppliers, an attitude that repair and rework were to be expected, high absenteeism resulting in poor workmanship, and damage to parts and vehicles caused by transport mechanisms.4 The assembly line suffered from low efficiency because work methods were not standardized, people could not repair their own equipment, and equipment was underutilized. Excess inventory, caused by ineffective controls, was another problem. Work areas were crowded. Employees took too much time to respond to problems. NUMMI’s solutions focused on the Jidoka principle—quality comes first. According to NUMMI’s factory system today, workers can stop the line if they spot a problem; the machinery itself can sense and warn of problems. Two well-known just-in-time methods of eliminating waste—the kanban system of production control and reductions in jig and fixture change times—are important to NUMMI’s manufacturing operation. But lots of other things also contribute to this plant’s effectiveness: simplified job classifications, displays and signs showing just how to do each job and what to avoid, self-monitoring machines. NUMMI has obtained high-spirited involvement of the employees, first by choosing new hires for their willingness to cooperate, then by training them thoroughly and involving them in decisions about how to improve the operations. Design Means Business The five tasks of design bring us back to the original point. Strategic product design is a total approach to doing business. It can mean changes in the pace of design, the identity of the participants, and the sequence of decisions. It forces managers, designers, and engineers to cross old organizational boundaries, and it reverses some old power relationships. It creates difficulties because it teases out incipient conflict, but it is rewarding precisely because disagreements surface early, when they can be resolved constructively and with mutual understanding of the outcome’s rationale. Strategic design is a continual process, so it makes sense to keep design teams in place until well after product launching when the same team can then tackle a new project. Design—it must be obvious by now—is a companywide activity. Top management involvement and commitment are essential. The effort has its costs, but the costs of not making the effort are greater. 1. J. Corbett, “Design for Economic Manufacture,” Annals of C.I.R.P., vol. 35, no. 1, 1986, p. 93. 2. David A. Garvin, “Competing on the Eight Dimensions of Quality,” HBR November–December 1987, p. 101; and John R. Hauser and Don Clausing, “The House of Quality,” HBR May–June 1988, p. 63. 3. E.H. Hartwich, “Possibilities and Trends for the Application of Automated Handling and Assembly Systems in the Automotive Industry,” International Congress for Metalworking and Automation, Hannover, West Germany, 1985, p. 126. 4. Kosuke Ikebuchi, unpublished remarks at the Future Role of Automated Manufacturing Conference, New York University, 1986.

Manufacturing by Design

Partner Center