U.S. Leadership in Manufacturing: A Symposium at the Eighteenth Annual Meeting, November 4, 1982, Washington, D.C., National Academy of Engineering. (1983)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

INTRODUCTION Joseph Harrington, Jr. In the keynote address and the response to it James B. Quinn and Thomas J. Murrin have given us a challenging overview of the current state of manufacturing in the United States, particularly vis-a-vis manufacturing in other parts of the world, and they have outlined some impressive challenges to engineers, to managers of manu- facturing, and to government policymakers. If there was a doubt in anyone's mind, it was dispelled by that opening. Change is inevitable, and when dealing with change, there are three things to be done. You ask, where am I or where are we? Which way are we going? And what ought we to be doing about it? That, in essence is the structure of the next sessions of this meeting. This first session will address the problem of where we are, what is the state of the art? The steering committee was not able to consider state-of-the-art reports on all the important technologies of manufacturing thus, they selected four essential ones. The first is the manufacture of parts, parts production; the second is the testing and quality control problems that go with manufacturing the third is the assembly of parts; and the fourth the integrating factor. The latter is my theme too—the use of the data technology to integrate or to reintegrate all of these many components of our industry. Everything that we do in manufacturing, every act that transforms raw material into finished products, can be represented by data. We generate data. We transform it. We transmit. And we are at present blessed with the most powerful data processing equipment that has ever been known. The medium by which we will be reintegrating our industry is obvious. It is not uncommon to find people discussing the acts of manufacturing, the shaping of parts, inspection and assembly, and testing and the relationship to data flow, numerical control, and things of that nature. On the other hand, they speak as if on another plane about the management of those acts, the planning, the schedul- ing, the data collection and correction activities. And they are, indeed, discussing two different planes so far as the technologies are concerned; but the data that move in those two planes are the same data. The data move back and forth from one plane to another, and that is the important message here. This data flow is the medium by which we integrate and control our manufacturing technology. It is becoming a science. If we understand, we can measure. If we measure, we can control. And if we can control, we should be able to succeed. 59

NEW MANUFACTURING TECHNOLOGIES: PARTS PRODUCTION George H. Schaffer James B. Quinn has eloquently addressed some of the main challenges facing engineers, manufacturing managers, and policymakers. It seems to me that among all the challenges facing those concerned with parts production, there is one common thread, a common denominator—that parts production must operate in a climate of constant change. The challenge is to operate economically in the face of smaller lot sizes, shorter product cycles, greater model proliferation, socio- economic pressures, and political realities. Fortunately, the crescendo of change confronting today's manufacturing manager is accompanied by the rapid emergence of technological options that can provide the flexibility and fast response time needed to meet the challenges. The basic metalworking processes are not likely to change fundamentally, but their organization and control definitely will. I would like to explore some of these technologies and their effects on the tools of production. The American industrial genius has been to create mass production systems on an unprecedented scale and efficiency. With its specialized machinery, mass production depends to a large degree on the existence of stable markets and long production runs. But the days of the black automobile and the white refrigerator are long over. The requirement today is for product diversification and fast response to the changing demands of the marketplace. Mass production, as we know it, is not compatible with these requirements. In fact, according to some, the limits of expansion of mass production manufacturing have probably already been reached for all but the most mundane products, and batch production is on the rise. The choice of production mode— whether piece production, batch production, or mass production—is clearly influenced by the need for flexibility and the size of the production run. It stands to reason that the call for increased flexibility and the trend toward smaller production runs will result in a further concentration on batch manufacturing. We tend to think of modern mechanical manufacturing as a highly productive and efficient process. Nothing could be further from the truth. A classic and much-quoted study of batch manufacturing at what is now Cincinnati Milacron found that the average workpiece spends only 5 percent of its shop time on the machine took For 95 percent of the time, it is being moved around or is waiting for work to be done. And of the 5 percent spent on the machine, less than 30 percent is spent in actual metal removal. Machine positioning, loading, gaging, and idle time constitute 70 percent of the time on the machine. Although this study is old by now and other figures may vary, the dimensions of the problem of unproductive time are undoubtedly significant. 60

6l There is another problem. The seemingly conflicting demands for greater product diversification, higher quality, improved productivity, and lower prices cannot be met in the climate of organized chaos that is so characteristic of batch manufacturing. Manufacturing management frequently involves nothing more than solving one crisis after another by using the most expedient solution avail- able at the time. Such management by crisis is hardly conducive to achieving an optimum manufacturing system. These are structural problems that can be solved only by taking a new look at parts production. We need to consider the entire production process—from the design to the field support and service of product—as a continuing spectrum. As Joseph Harrington has put it so aptly, manufacturing is an indivisible continuous fabric extending from first conception of a product through design, production, and distribution to field maintenance. Of course, this continuum is composed of an incredibly complex, fine structure of many individual functions, each inex- tricably connected to and dependent on every other. It is the close interdependence, the symbiotic relationships within the fabric of discrete-parts manufacturing, that makes it so susceptible to chaos. But di- verse as the various parts of manufacturing may seem, there is a common element governing all manufacturing activities: What we call manufacturing is, in the ultimate analysis, a series of data processing operations or data transforma- tions. All of manufacturing involves creating, sorting, transmitting, analyzing, and modifying data. Therefore, everything done in manufacturing, whether in the physical act of material transformation or in planning and management, is part of a continuum of data processing. This data processing activity is the conceptual key to what is now referred to as computer-integrated manufacturing (see Figure l). Ultimately, FIGURE l Computer-integrated manufacturing (CIM) closed loop system.

62 computer-integrated manufacturing attempts to achieve a closed-loop feedback system whose prime outputs are finished products. It comprises a combination of software and hardware that includes product design, production planning, pro- duction control, production equipment, and production processes. According to some of the best authorities, computer-integrated manufac- turing has already demonstrated greater potential for improving manufacturing capability than has been shown by all other known types of advanced manufac- turing technology combined. Unfortunately, computer-integrated manufacturing is not a shelf item readily available for application, nor can it be achieved by management fiat. It is continuing evolution, a goal that can be achieved only by planning from the top down and implementing from the bottom up. Unquestion- ably, the computer is emerging as a dominant—perhaps the most dominant— manufacturing tool. That dominance started with isolated applications that have evolved into islands of computer-based systems. Although the applications vary in scope and complexity, they feature a common characteristic Invariably, computers are used to provide more accurate and more timely information than is possible with cur- rent manual systems. The basic information-handling tasks required during the life cycle of a product are evolving into a series of computer-based systems that form the basis for computer-integrated manufacturing. First in this interdependent chain of information systems is the means for capturing the information generated during design. Much of that information deals with geometric data, which are readily transformed into a geometric model, a representation of shape and size in computer memory, through the use of computer graphics systems. Such computer-aided design (CAD) is clearly taking over the design of products. The geometric model can be used to generate fully dimensioned engineering drawings but is also the key to a host of related design/engineering/manufacturing functions, many of which can be performed concurrently, greatly compressing the product development cycle. Some of these activities are now being called computer-aided engineering (CAE) and are intended to automate the entire mech- anical product development process. Starting with the geometric model of a prototype, CAE uses the computer early in the design process to simulate performance of the proposed product. More directly related to manufacturing, numerically controlled (NC) parts programs are readily generated from geometric models with most of today's computer-graphics systems. The computer alone enhances the programmer's capability in a number of ways: • It provides calculation capability beyond that of the machine's NC system and removes from the programmer the burden of manual calculations and geo- metric constructions. • Program reliability is enhanced because the programmer has fewer oppor- tunities to make errors. • The programmer's need for intimate knowledge of the idiosyncrasies of each NC machine and its specific coding requirements is greatly reduced because the computer typically uses a shop-oriented language. • The computer is not restricted to generating NC codes. It can also provide management information for estimating and planning, including tool management.

63 Such computer-generated NC programs can then be verified by three-dimensional graphic simulations. One of the most significant emerging information systems, sometimes re- ferred to as the "glue" of computer-integrated manufacturing, is group technology (GT). It is a manufacturing philosophy, an organizational principle with far-reaching implications. The underlying principle is relatively simple and not particularly new: Identify and bring together related or similar components and processes to take advantage of their similarities in design and/or manufacturing. GT uses well-structured classification and coding schemes and associated computer programs to exploit the sameness or similarity of parts, processes, and equipment. On the one hand, this reduces duplication of engineering effort; on the other hand, it affords an opportunity to group similar parts and processes, thereby achieving economies of scale otherwise not possible in batch manufacturing. The grouping principle can have a profound effect on virtually every aspect of the manufacturing cycle (see Figure 2). This effect is particularly true in batch-manufacturing operations, which typically involve seemingly endless variations of parts and processes. By helping to identify select similarities, GT can provide considerable benefits for most of the functional areas in a manu- facturing organization: product engineering, manufacturing engineering, production control, and procurement. In product engineering, GT can reduce part proliferation, encourage design standardization, provide manufacturing feedback, and facilitate cost estimating. GT can help manufacturing engineering with process selection, tooling selection and grouping, machine procurement, facilities planning, materials flow, and materials handling. It can also help to bring newly available technology to the attention of planners by automatically including recently acquired applicable equipment or capabilities as processing alternatives. OTHER AREAS PROCUREMENT PRODUCTION PRODUCT ENGINEERING MANUFACTURING ENGINEERING PRODUCTION CONTROL FIGURE 2 Functional areas affected by group technology (GT).

In production, GT can reduce lead times, production delays, and setup times. It can also help with asset utilization, materials-handling decisions, and equipment selection to achieve appropriate quality levels. Production control can use GT for scheduling, stock accountability, expedit- ing, and reducing work-in-process inventory. Buy-or-make decisions and the establishment of economic order quantities can also be handled through GT. Ultimately, GT can affect customer support by improving the handling of dealer inventory and by shortening delivery times. All of these benefits are achieved by identifying and assessing an array of information, then retrieving and grouping designs, parts, or processes on the basis of select attributes. Although GT implies the establishment of manufacturing cells to handle families of parts, those cells need not necessarily involve the physical rearrange- ment of a facility. Most of the benefits of GT can be realized through admini- strative means, without such physical rearrangement. Closely related to GT and another key factor in effecting computer- integrated manufacturing is computer-aided process planning. A planner must manage and retrieve a great deal of data and many documents, including established standards, machinability data, machine specifications, tooling inventories, stock availability, and existing process plans. This is primarily an information-handling job, for which the computer is an ideal tool. There is another advantage to using computers to help with process planning. Because the task involves many interrelated activities, determining the optimum plan requires many iterations. Because computers can readily perform vast num- bers of comparisons, many more alternative plans can be explored than would be possible manually. A third advantage in the use of computer-aided process planning is uniformity. It has been said that if you ask ten planners to develop a process plan for the same part, you would probably end up with ten different plans. Obviously, they cannot all be the best plan. This also means that essentially the same job planned at dif- ferent times will be done differently. Which plan will govern facilities planning? Which will be used for estimating future work? Which plan will be used for scheduling and shop loading? There are basically two approaches to computer-aided process planning: var- iant and generative. In the variant approach, a set of standard process plans is established for all the parts families identified through GT. The standard plans are stored in computer memory and retrieved for new parts according to their family identification. Again, GT helps to place the new part in the appropriate family. The standard plan is then edited to suit the specific requirements of a particular job (see Figure 3). A generic variant approach is illustrated by the computer-aided process planning system developed under the auspices of Computer Aided Manufacturing-International Inc. (CAM-I). In the generative approach, an attempt is made to synthesize each individual plan using appropriate algorithms that define the various technological decisions that must be made in the course of manufacturing. In a truly generative process planning system, the sequence of operations, as well as all the manufacturing process parameters, would be established automatically, without reference to prior plans. No such system exists, however. So-called generative process-planning systems are still specialized systems developed for a specific operation or a

Patt- • - . matrn 1*9 Standatd •equence Part- famrfy search •equ*oce Pan- classiftcahon code PfOC»M plan FIGURE 3 Computer-assisted process-planning system of Computer Aided Manufacturing-International Inc. (CAM-I). particular type of manufacturing process, and probably will be for the foreseeable future. The logic is based on a combination of past practice and basic technology. Another fundamental requirement for computer-integrated manufacturing Is an effective manufacturing control and manufacturing planning system. These systems are typically modular and address such functions as materials require- ments planning (MRP), inventory control, capacity planning, scheduling, fore- casting, and cost control. To be effective, the modules must be linked to an overall management information system. What effect will the advent of computer-integrated manufacturing have on the tools of production? Clearly, there is a move to flexible manufacturing systems (FMS)--programmable production systems consisting of two or more machine tools linked by materials-handling elements, including robots, and supervised by a computer-based scheduling and control system. The recent International Machine Tool Show demonstrated that manufacturer* who want to make this move will find the machine-tool industry ready with the necessary cells, systems, and peripheral equipment. A major emphasis at the show was on fitting each machine or accessory into flexible, electronically controlled, automated combinations with other units. In fact, implementing FMS was the recurrent theme at the show. Virtually every new NC lathe or machining center or punch press offered the ability to b« readily incorporated into a multimachine cell or a fully integrated manufacturing system. Robot loading was a common element (see Figure <4).

66 FIGURE 4 Robot loading machine, part of a flexible manufacturing system. FIGURE 5 Flexible manufacturing system at Mazak machine tool factory near Cincinnati.

FIGURE 6 Front panel of a computer- numerically-controlled machine tool. The most impressive FMS demonstrated was a multimillion-dollar, five- machine system that Yamazaki is now installing at its Mazak machine tool factory outside Cincinnati (see Figure 5). Unlike most previous FMSs, which have been designed to handle either rotational or prismatic parts, the Mazak system carries both types of parts around on lazy-Susan pallets, which are slid off wire-guided carts at turning stations and machining centers to be plucked by loader robots. Tool storage capacity of l20 tools is divided among four interchangeable carousels. Control is the key to any flexible manufacturing system and is achieved, particularly at the machine level, with computer numerical control (CMC). The use of CNC is such systems is increasing, but a parallel and more visible devel- opment, as amply demonstrated at the l982 International Machine Tool Show, relates to its increasing versatility in stand-alone job-shop machines. Virtually all of the CNCs exhibited featured direct-programming capability at their own user-friendly front panels (see Figure 6). The user that the control builders have in mind is the person operating the machine, and the friendly features include full-color-graphics/CRT displays, interactive menu-driven program development, soft-button function assignments and, in some instances, aspects of computer-automated process planning the new MPC II grinder control from Landis Tool even included a voice synthesizer to enunciate and confirm keyboard entries as they are made. Another necessity for FMSs is the emergence of untended machines, which

68 carries its own imperatives: tool management, for instance—ensuring that the right tool is at the right place at the right time and that dull tools are auto- matically replaced. For example, the emergence of untended lathes has shifted the emphasis from swapping cutter styles, which have always been available with tool turrets, to renewing dulled edges. The Sandvik Block Tool System was incorporated by Cincinnati Milacron into an 84-tool chain-type magazine, which itself can be brought to the lathe by wire-guided cart to be swapped automatically for a magazine of used tools. As machine tools become untended, the means for checking cutting tools and for monitoring the manufacturing processes in a timely fashion also becomes essential. Virtually all of the manufacturing systems at the International Machine Tool Show included some form of in-process or postprocess inspection. Much of the former, particularly for tool verification, was accomplished with on-the- machine probing systems. Tucked away in the magazine of many tool changers was a probe that could be brought into play just like any other tool. Most of these were touch-trigger proves used to automatically adjust tool offsets, correct for home position errors, and detect and compensate for material variations, such as those encountered in castings. Not all the probes were mounted in the tool changer. The J&L FMS lathe uses two retractable touch-trigger probes to verify tool locations after automatic tool changing. And there is an analog probe, a tool-changer-mounted electronic plug gage from Federal Products that provides dimensional data for automatic adjustment of a DeVlieg Microbore boring cartridge while the cartridge is in the tool magazine. In many instances, robots serving manufacturing cells or full-blown manu- facturing systems alternated between feeding blanks to machines and presenting the machined part to a postprocess gaging station, whose dimensional data were used to make corrections in the machining program. There is a need for diagnostic devices that reliably predict a failure—of a bearing or tool, for examples-just before it occurs, instead of identifying the component after it has failed. Untended machines will also work harder, racking up more continuous oper- ating time in a given period than conventional machines. That means earlier replacement. And, of course, technological obsolescence also has a tendency to reduce the useful life of machine tools. In summary, the flexibility and fast response needed to meet the challenge of constant change facing manufacturing must be addressed on two levels. First, organizational computer-based technologies, such as group technology, and auto- mated process planning, must transform the organized chaos of manufac- turing into continuous-process-like, fast-response systems. Second, the physical tools of manufacturing must be provided with flexibility and control capabilities to tie into a systems-oriented production environment.

ASSEMBLY Joseph F. Engleberger Our keynote speakers both referred to robotics. I will consider only one segment of robotics—assembly. In the l936 movie "Modern Times," Charlie Chaplin blew the whistle on the abuse that we handed our working force in the modern assembly plant. At the time of that film, of course, labor was cheap, it was plentiful and, goodness knows, it was intimidated. None of those situations exists today, but even that far back we did have some technology to get assembly done automatically. Rotary turn- tables, for instance, represent the various automatic machines that would put pieces together (see Figure l). Professor Boothroyd, University of Massachusetts, outlined characteristics that one would ordinarily expect to have in part of an assembly so that classical, or what we often call "hard automation," could be used: • Volume of at least one million per year • Steady volume of production • Market life of at least 3 years • Size on the order of 0.5 to 20 inches with individual parts to be automatically assembled generally between 0.05 and 5 inches in their maximum dimensions • Consisting of parts that do not deform significantly under their own weight or will not break when dropped from a height of about 3 inches onto a hard surface If one considers these things, he will see many restrictions. A steady high volume is needed. The product needs to be around for quite awhile without undergoing change. Certain limitations on size are needed because these parts are sorted out by tumbling them. Moreover, the parts should not be squishy, such as cloth; they cannot change their charateristics as they tumble. Thus, there are serious limitations to the kinds of things that can be done with hard automation. Let me take one example from an industry that has espoused robotics more than any other, and that would be in the assembly of a speedometer. In an automotive speedometer there is the odometer. Every company in the world makes odometers on hard automation machines that spit them out one every second or so. But every company in the world then assembles the speedometer with people sitting on a line and putting them together, including those odom- eters. Why? Because the designer intervenes. He wants the speedometer long 69

70 FIGURE l Rotary index machine. ROTARY INDEX MACHME or he wants it round, he wants gas gauges, he wants "idiot lights"; so every model auto needs a special speedometer assembly. We simply cannot use hard automation. Not only are the parts too variable, but they are becoming obsolete in every model year. Robots come into the act in the programmable automation area by taking over some of the activities now done by humans and some of the activities now done by hard automation. Consider a robot that can simply replace the human on the line and give the robot the same sphere of influence, the same speed, the same accuracy, and let it stand on the line to replace a human. If such a creature were available to manufacturers, they might hark back to that hard automation machine and say, "Well, instead of just having these feeders, I am going to have programmable automation standing around the rotary table." Thus a range of products can be assembled on one flexible system. Group technology, which has been mentioned more than once, certainly can permeate the assembly process as well as the parts manufacturing process (see Figure 2). Perhaps, however, there is something in this assembly that requires judgment, so that a human must be in the loop. Therefore, we use a conventional assembly with which people and foremen are very comfortable. It is a line on which pallets with tooling index along in one-second jumps. Each pallet stays about eight sec- onds in station. At the moment, this sort of a line has human operators stationed along it. Some stations, however, can be operated by robots, perhaps two at the station if there is hand-to-hand coordination necessary (see Figure 3). There may be a process in which we must have human judgment, and the human is in that station; or perhaps the human is in the station only for a period of time until we can iron out some of the processing for that station.

7l FIGURE 2 Group technology manufacturing system. What are the economic issues? Figure 4, for instance, is from the robot group at Draper Laboratories, which is examining costs in assembly in an attempt to determine where programmable automation can best be used. One can assume that once the learning period is over, manual labor is going to be constant, no matter how many assemblies are made. One can also assume that if hard auto- mation is used, the costs will decrease as the quantity increases. Once hard automation is created, it spits assemblies out at high rates. There is an area, though, in which programmable automation comes into play. If only one of something is being manufactured, no kind of automation works ef- ficiently, but programmable automation may offer the best economics. Inexorably the two straight lines are going higher in this plot, and the curved line is going lower because programmable automation, robotics, group technology, and CAD/CAM all have economy-of-scale benefits. Thus, there is ever-growing opportunity for the use of robotics. Where is the technology going that will be significant for the use of robotics in assembly? Essentially there are two critical areas. One is vision and another is tactile sensing. In terms of vision, a camera may be mounted looking down on the scene. It understands that scene in world coordinates, the X, Y, Z, and 6 of a part or parts, and it communicates these coordinates to a robot arm, telling it where to put its hand to find the part. However, an eye in a robot does not have to be in the ceiling or in a head; it can actually be in the palm. The eye may project its own beam of laser light, and it looks at a scene with a vidicon camera and analyzes

72 that scene. This eye, then, is in the palm, and the palm essentially roams over the workpiece to examine what is happening in that scene, what is different about this set of parts. If, for good economic reasons, we do not make parts so that they are very precise, we need humans with eyesight to weld those parts. However, the robot with some eyesight can examine a scene and then do very much as the human would in performing the welding job. Tactile sensing is also a lovely development. It has come out of academia, and there is an interesting story behind it. In Draper Labs a student getting his Ph.D. degree in computer science was telling his colleagues (with equations) how he was building a wrist-force sensing system that, through zeroing all torques and forces, would enable an assembly to be made by a robot. One of the mechanical engineering Ph.D. students looked at all this and said, "Hell, I could build some- thing like that mechanically," and he did. This is strictly a passive device that essentially makes parts float together without the help of a computer. One more feature of the robots is their mobility. Some robots can traverse a floor for 40 feet, arrive within 2 inches of destination, go through a docking procedure, and lock into the dock. The arm performs a job there, and the robot moves to another station as needed. Now, we have all these attributes, all of the programmability that I spoke of before, and you say, "Gee, it must be easy, isn't it, to do assembly with robots?" Just so that you can see how difficult it still remains, I want you all to be able to do what I call "play the robot assembly game." You can do this at home; it is a low-budget game. First, rub petroleum jelly on your glasses, and then tie one hand behind your back. If this particular assembly job requires two hands, get a friend to rub FIGURE 3 Assembly line with human and robot operators.

73 Hard Automation Manual $9/Hr. Programmable (Robotic) 0.2 0.5 1.0 2.0 5.C Annual Production Volume, Millions FIGURE 4 Comparison of costs for manual, hard automation, and programmable assembly. petroleum jelly on his glasses and tie a hand behind his back. Next, put a mitten on that one hand and then pick up chop sticks. You now have every attribute of an assembly robot today, and all you do is to assemble something according to detailed instructions. So, we still have some work ahead of us to beat the assembly game. On a serious point, I would say one other thing. The robot is actually pretty good at assembly. The trouble we have in programmable assembly is with the peripheral activity of presentation of the parts, particularly if they are small parts. If human intervention is necessary, the human might as well put the parts together. Therefore, you need either a vast depth of black art with feeders, or you need a robot with the ability to do something it cannot do yet, which is to look into a tub and pick out randomly-oriented parts. That is called the "bin picking problem" or, among the technical types, the "occulsion problem." It has not been solved. Joseph Harrington and I were talking earlier about a grave manufacturing deficiency. Almost everything in the world was once oriented in a factory. Someplace somebody took each part out of a machine or a station and threw it in a

74 TABLE l U.S. Export Competitiveness Hourly Compensation Country ($ U.S.) United States l2.6l Sweden ll.l7 Netherlands l0.9l Belgium l0.ll West Germany l0.06 Canada 9.36 Denmark 8.l0 France 7.57 Britain 7.35 Italy 6.97 Japan 5.72 box. If you had a scrap ticket for every time you lost orientation in a factory, people would pay more attention to the blessing of orientation and preserve it. About 65 years ago the catch phrase of Detroit automation became "never drop a part." I can tell you that is obeyed mostly in the breech. So, if we can rationalize the workplace and we can use CAD/CAM and group technology and say that we are not going to drop the parts, we will be able to do a lot more with programmable assembly. Table l is extracted from a recent Wall Street Journal. It tells something about the problem in this assembly arena. U.S. labor is most expensive (yen were about 265 to the dollar when this was created; they are now about 278 to the dollar). We have more than a two-to-one range between our toughest competitor and us. Let me continue on the Japanese because Japan was emphasized by our key- note speaker. I have gone to Japan every year for the last l5 years, and about 5 years ago I asked, "How about robots for assembly?" The Japanese said, "Oh, no way. We are not going to do that. The Japanese are perfect people for assembly. We are extremely conscientious, we are quality-conscious, we are fast, we are small, and we can put little parts together very quickly. We are not interested in robots for assembly." Last year on my trip the projections that the Japan Industrial Robot Association had for all the robotic activities they saw through l990 forecast the single largest class of robot activity as being assembly. I said, "What happened?" "Well, we made some demographic studies; we concluded that we will never have enough people; we are a monolithic society; we have a fixed population; we want people to retire earlier. We will never have enough labor. We plan to make this country a country of only knowledge workers." Then they said something that was very important to me. They said, "Of course, when we use the robots we will not do assembly the same way anymore." We talk to people in the United States about using assembly robots, and they

75 say, "Oh, we do assembly on one shift. There is no economic justification for robots. What do you need? You need a roof; you need some benches and some little tubs—virtually no capital investment. You just sit a girl at the bench and have her assemble." The Japanese immediately saw, as implementers, that if you start to do assembly with programmable automation you have a capital-intensive activity, and when you have a capital-intensive activity you are going to run it around the clock. That is going to be a hard sell unless attitudes change in this country. I am going to close with a sociological observation that tickled me. Most of you, probably in college, read Orwell's l984. I read it again a few years after college, and I found out by reading Futurist Magazine that a tremendous number of his predictions have come true. In fact, the magazine listed l37 predictions, of which l00 have already come to pass: Rapid access to and retrieval of information, data banks containing detailed personal information, think tanks where experts plan future wars, poisons capable of destroying vegetation (like Agent Orange), disease germs that are immunized against antibodies, lack of heating fuel and electricity, merging of the genders. Now, one wonders, is Orwell going to be right completely? Is his accuracy ever going to end? What will still come to pass in the next year-and-a-half ? Eric Fromme, the philospher/psychiatrist, observed, "George Orwell's l984 is the expression of a mood, and it is a warning. The mood it expresses is that of near-despair about the future of man, and the warning is that unless the course of history changes, men all over the world will lose their human qualities, become soulless automatons and will not even be aware of it." Orwell wrote l984 in l948. All he did was transpose the last two numbers. By l948, as a well-read person in scientific literature and science fiction, he had to know about RUR, Rossum's Universal Robots, a successful play in l922. He had to know about Issac Asimov's early stories, his I, Robot stories. He had to know about Russian science fiction. And yet, I can assure you, never once in this pre- dictive book did he mention the word "robot." He did not because his worst nightmare was that people would be automatons, that people would become robots. There is not any way that we in the robot business could possibly compete with a human who has been robbed of his personality as a human being. A robotized human is the cheapest possible labor there could be. So, I put it to you that there is hope. There is only a year-and-a-half left, and if everyone gives the robot industry sufficient support, people are not going to become automatons, and we will have exorcized that particular Orwellian nightmare.

TESTING AND QUALITY Susan Foss The introduction of global products within our marketplace has taught us a valu- able lesson: that we cannot sell a product based on price alone. Quality in the product gives the competitive edge. Currently, the manufacturing philosophy of this country utilizes an appraisal- oriented assessment for product quality. I call this assessment "after the fact." Over the past 20 years the use of this type of open-loop system has escalated. From present indicators, however, this method is not working, or as is often heard, "you cannot inspect quality into a part." The preceeding suggests that a new philosophy needs to be implemented, one using a prevention-oriented or before-the-fact approach. It should be a total closed-loop system that starts with the product conceptual stage and continues 76

77 through postwarranty. Emphasis at the onset should be on quality of the product in concept/design, processing, warranty, and postwarranty stages. There should be an information-flow system that allows pertinent data to be available wherever needed throughout the product flow. The closed-loop approach would change the present inspectors, as we know them today, so that they would be more like auditors, not necessarily auditing the part but auditing the machine tool capability. This would change the present in-line measuring machine function from distinguishing between good parts and bad parts to one of feeding information back to the machine tool for needed corrections. The machine operators would again be asked to serve partly as machine managers, since they know best the functional capabilities of each of their machines. The expansion of this closed-loop approach would automatically lead to the life-cycle dimensioning concept. Needed information or data is shared from product concept to product mortality. Emphasis is placed on using what we know and on learning from it. In moving from an appraisal approach to the prevention method, several tools are necessary. Computer-aided design/computer-aided manufacturing (CAD/CAM) provides a consistent base for part definition. As known today, it operates in the design, manufacturing, and some concept and assembly stages. Closed-loop inspection provides dimensional history of machine tools and/or parts. It needs to operate from concept through postwarranty stages, with total feedback at the pertinent functions. Statistical analysis techniques should be an integral part of this system. Concept/design quality simulation provides compatibility of design and processes and provides quality information from

78 concept through assembly functions using simulation techniques. We have the technology and the ability to implement this approach. Today I will discuss two tools already mentioned: closed-loop inspection and concept/design quality simulation. Deere & Company has a computer-aided inspection and reporting system (CAIR), which presently uses a man-closed loop. It performs the normal inspector operation, monitors process capabilities, and determines machine trends by part dimensions. It provides Deere with a rapid and accurate means of obtaining a data base for parts and machine tool dimensional data and also provides analysis routines to assist in interpreting this data. CAIR operates in l6 North American and 3 European facilities. Presently it contains 26 multitasking minicomputers that communicate with 87 digital devices. A typical CAIR system has a multitude of digital input devices. These include: • Manual and computer-controlled measurement machines • On-line digital gaging

79 • Engine test cells • Theodolites, which are surveyors' electronic transits used in place of measurement ma- chines for measuring large frames and fixtures • Booms on excavators or frames for large motor graders Typical output devices include: • Printers and cathode ray tubes (CRTs) for analysis and/or inspection reports

80 • And within one year, direct feedback to the machine tool CAIR offers, in addition to an inspection report, a series of analysis programs or routines to assist with data interpretation. These routines can on-line, real-time analyze the "just captured" data from the operating process and instantly display the results on the CRT or printer. For instance: • A histogram displays the frequency distributions as a bar chart and prints pertinent statistical results of the parameter that is under investigation. A normal curve is used because experience has indicated that this closely approximates manufacturing processes. • Trend analysis displays the actual measured dimensions of the part versus the number of pieces measured. A best-fit line is drawn through the data points with a 99.7 percent confidence band constructed around this line. Real-time process variations can now be readily seen, and needed adjustments can be made.

81 • Feature analysis analyzes bidirectional data—in particular, the X and Y locations of bores. This routine displays the relative locations of actual measured data, indicated by the x's, with respect to the bore print specifica- tion indicated by the target circles. All the x's exceeding the circles are outside the print specifications and indicate bore location shifts. Previously, extensive time had been consumed in trying to sort through tabulated data to establish this shift. • With this routine the features are visualized and required machine tool adjustments can be readily seen and made (again by the x's, now inside the target circles). Extensive use of this analysis technique has been made when setting up new flexible machining lines at several of our facilities. It alone has provided substantial cost avoidance by significantly reducing delays encountered with machine tool deliveries. Daily use of this routine is made to monitor existing manufacturing processes and provide process control information. In addition to the analysis routines, CAIR can provide the product engineer with • A historical dimensional data base for experimental parts • A way to select tolerance bands • Increased knowledge for modifying part dimensions • Tolerance degradation effects on part performance • Vendor process capability For the manufacturing engineer CAIR provides • A dimensional data base for fixtures and tooling • Process capability information

82 • Assistance in modifying tooling and fixtures • Tolerance degradation of these items • Machine auditing capabilities The benefits of CAIR are being demonstrated daily. A time savings of 7:l is realized in our inspection operations, and process capability studies have been increased by a factor of l5. A second tool needed to achieve a prevention system was developed by Chevrolet Division of General Motors. It is called variation simulation modeling (VSM), and uses a closed-loop approach. This system provides General Motors with a probablistic approach for early detection of component variation problems as new product designs are developed. This includes the process as well as the concept/design areas. VSM presently resides on a large main frame computer system and is available throughout General Motors Corporation through interaction with the Chevrolet Division VSM team. A typical VSM system consists of • Cathode ray tubes for input • Mainframe computer • Printers and CRTs for analyzing results Variation Simulation Modeling operates in the following manner: WATER PUMP SEAL PRESSURE SIMULATION VSM C> • OU»F(0 81J4 LOGIC • A mathematical model is constructed that establishes the relationship between the components of the assembly and imitates the assembly operation involved in putting the components together. This model describes the relationship between the parts of the system and how the system operates on the parts.

83 • Each component of the assembly is defined by its specific nominal and associated tolerances. For the concept/design area this would be blueprint values, and for the process area this would be measured data. In addition, a probability distribution function is associated with each dimension. Available functions include Normal, Uniform, Constant, and Random. • The use of these density functions allows the random selection of component dimensions by having a random number generator pick a number between 0 and l (such as 0.5843) and generate an associated component dimension based on its nominal values and tolerances. The cumulative distribution function (CDF) portrays the area under the probability density function, and this area also varies from 0 to l. This CDF is used to simulate the process operation of an operator randomly picking a part from a bin. VSM DATA ANALYSIS FRIDAY NOV 14, 1980 2:30 PM PRESSURE EXERTED ON ROTOR FACE VARIABLE: ROTPR W.OMb W.H J.O30 «U»6E % OUT OF tKC 1 • Once the assembly is modeled and appropriate distributions chosen for each of the component dimensions, a simulation can be run for whatever sample size is desired and an analysis conducted. This will indicate the statistics for

the sample run and the percentage out of spec as predicted by the model if the assembly as such is designed and/or processed. If the simulation analysis shows that our system is not meeting design intent, a determination must be made as to which variable(s) at the component level is significantly contributing to the overall variation. This is done by means of a tolerance sensitivity analysis. Once this variable or variables are determined, they are held at their nominal values by eliminating the random selection process, and the model is rerun, the object being to make percentage out of spec equal to zero. For the design engineer, VSM • Provides a substantial cost avoidance during concept/design and process stages Ensures correct tolerance interactions Aids with assigning tolerances Makes the probablistic approach practical Minimizes potential quality problems Reduces prototype builds For the manufacturing engineer, VSM • Provides substantial cost avoidance during concept/design and process stages • Minimizes potential quality problems • Aids process change decisions • Aids machine tool decision • Ensures a smooth flow from design through process stages The benefits of VSM are many. A few of these are: • Reduction in concept/design time by a factor of ten • Cost avoidance by reducing prototype builds and rebuilds • Cost avoidance by identifying potential quality problems prior to prototype builds r tt I—*- In summary, computer-aided inspection and reporting and variation simulation modeling, when merged with CAD/CAM, will provide the needed tools for a prevention approach. CAIR provides the dimensional history of machine tools

and/or parts, VSM provides compatibility of design and processes, and CAD/CAM provides a consistent base for part definition. The competitive edge against global products within our marketplace begins with quality, be it in the agricultural, the automotive, or even the aircraft industries. By using the important tools currently available, quality will start within the conceptual stage of design.

MANUFACTURING INFORMATION FLOW William D. Beeby About seven years ago, we at Boeing had done a lot of experimenting and testing of various applications of the computer for doing individual jobs in design and manufacturing. Our top management began to realize the amount of money that we were spending and requested that we come up with a total concept of how we were going to put this all together and make it pay off. The basic presentation I am giving today was actually used seven years ago. It will relate some of the successes we have had in integrating engineering and manufacturing and also some of our failures. I must comment now that most of our failures have not been technical fail- ures. They have been psychological failures because of our inability to convince people that there is a new way of doing business. One has to think differently in this new world of computers and automation. We started with a concept, a master concept that the manufacturing business was data intensive. Every function we performed was done because of data that were passed from one organization to another, whether we were doing very pre- liminary design or actually going into production. The same data supported assembly or detailed design, release and control of that design, and manufac- turing. Each of these areas needed the same data. The data that started in preliminary design were merely enriched, modified, improved, and used by various organizations in doing their jobs. We developed a system that would use a common data base; and all organizations, from prelim- inary design all the way to customer support, would have that data available to accomplish their tasks. We started describing this process by saying that when we received a go- ahead for a preliminary design, the engineer would actually start to load three different data bases. We broke it up into three data bases seven years ago because at that time no one could conceive of a single data base that could accomplish all the functions we felt had to be accomplished. So we set up a data base that was to be used for business systems—for production control, parts lists, inventory control, etc. Then we developed a geometric data base for handling the master models and the geometry of the product. For the business systems we used an off-the-shelf data manager, but for geometry we were unable to find a commercial data base manager, and it became necessary for us to develop and build a geometric data base that would handle all of the coordinate systems, the centerline data, preliminary geometry, and our master models. We also created some cases in which, from a surface program, we could generate a mathematical surface definition that could then be automatically fed to a machine tool that would in turn create our wind tunnel models. It would also give us results of testing instantaneously. We could then make corrections or improvements to the product and try it again. Because we passed this data to 86

87 other designers, we also needed a data base that could store our design analysis data. This data base now holds our specifications and the results of analysis tests so that the data will be available for other people to check against. When the go-ahead for a new product was given, we would extract the data that preliminary design had stored and start from that baseline to develop a more production-oriented model. We gave the engineer local storage space and his own stress analysis and fatigue analysis programs. This allowed him to analyze his work as he completed the assembly and installation design. We also made the data available for the technical staff analysis programs so we could verify that the design we were creating matched the specifications. As the design progressed, the detail designer was able to use the data already in the geometric data base and start putting in the effectivity, the parts list data and the used on notes in the business system. Here was a case for which we have had a substantial success in that the detail designers completely accepted the use of computer data for doing their detail design. Over ko percent of all the designs released on the 768 and 757airplane programs were released using this data base system. They actually came out of the computer and were described mathematically rather than on paper drawings. We still have paper drawings, however, because many of our subcontractors cannot use computer data. We committed ourselves to developing a number of special programs that would allow the design engineer to create flat patterns automatically and to develop hole patterns from design criteria, giving dimensions, finishes, and material specifications. Putting such data in the computer allowed the engineer to specify the material he was using and the conditions it had to withstand; the computer then selected the optimum finish per the Boeing design manual. Special programs were also developed to help the engineer in weight and stress analyses. Previous papers have discussed group technology. I want to add that at Boe- ing we have forced this group technology approach back up into engineering so that we can use it in selecting standard parts, to avoid new designs when old ones are already available. We have developed an automated system to help the engineer find these designs in the computer. One of the most critical areas of development was the release and control of design information. It is very hard to convince an engineer that what he sees on a piece of paper is actually stored on a disk and that nobody has changed it. We had a very difficult problem in establishing an engineering release system that would permit very tight control of computer data. It has been done, and manufacturing now considers the data in the computer as having the same authority as the paper drawing. Manufacturing planning is able to extract from these two data bases all of the material requirements, the number of parts to be built, and the types of assem- blies. They are also able to call up a copy of the geometry and modify it for their needs. In other words, they are able to add more material if they need it to grip a part for stretching or if they need certain holes left blank for an assembly operation later on. This allows them to release a manufacturing drawing without the need to copy any pertinent engineering information. The data are then stored in the geometric data base for use in fabrication and assembly work. I wish to stress again that there is only one data base common to both engineering and manufacturing. Quality control then begins to look at the part data and to decide what con- trols are necessary and how they are going to monitor the part. They will add

88 the necessary information to the manufacturing plan and create a quality control drawing for fixtures necessary to control quality. All the data are then passed to the tool designer. He can now start to design the tooling—the dyes, jigs, holding fixtures, etc.—around the engineering geometry without having to reinterpret the engineering drawing. Tool design was one of the areas that turned out to be the most surprising and satisfying in the design and building of the 767 and 757. Tool design was quite skeptical of the ability of engineering to store geometry data and control it and then to allow them to use it. But there was a big drive to use computer-assisted design/computer-assisted manufacturing (CAD/CAM) in the tooling department, and it resulted in a large reduction in the number of designers required to design tools for the new product. Of course, at the same time, the numerical control (NC) programmer made big strides when the data was properly defined by the design engineer. When the design engineer did not have a thorough understanding of the way in which the NC programmer needed the data structured, however, the data became almost use- less. The communication gap between the manufacturing engineer and the product design engineer is probably the most difficult thing that has to be dealt with in total integration of manufacturing systems. In the fabrication of detail parts, the common data base has proved to be extremely valuable. In some cases the data have actually been transferred from engineering through a manufacturing postprocessor directly to NC machine tools. The same data were used by quality control to fabricate inspection devices. Initially we did not consider subcontracting a very important link in this total integration process. It was not until we started releasing design information on the 767 that we began to realize that over 60 percent of the fabrication was ac- complished by subcontractors. On the 757, which followed the 767 by about eight months, we did start an extensive program with four major subcontractors to furnish them not only drawings but also tapes that would give a complete mathematical description of the parts and assemblies they were to build. Integrating the subcontractor will probably be the biggest effort that we have to make in the future. The success of the communication with the subcontractor is dependent upon standards. There is a strong move within this country to accept IGES as a standard communication of geometry. Although it is not wholeheart- edly supported by all companies at this time, in the near future there will be enough demonstration of this standard that all industry will move to accept it. The next important area after subcontractors is subassembly, in which we are doing some automation for our wire bundles as well as controlling cube storage for the component parts. We are doing all of the extractions for quality control processing from the computer data base. We have begun to use a few robots experimentally. We are doing more and more with automatic machines in the major assembly areas. They are flexible manufacturing machines, primarily for drilling and riv- eting, which is the major activity in the assembly of aircraft. The data to drive these machines are being supplied directly by the engineering data base. One of the largest of these machines drills and rivets all the skins and stringers on the wing panels. On the first 767 this machine accepted all of the geometry data directly from the engineering data base, and manufacturing merely added the sequence for drilling and riveting. This installation required over 33,000 holes to be drilled and rivets to be driven. The operation was completed without a single error. This is the kind of performance that can be expected when employing extremely accurate data that can be used without the need for human interpretation.

SESSION 2 INTEGRATION OF THE MANUFACTURING SYSTEM

Session 2 participants. Left to right (standing) D. C. Burnham, Frank Daley, Session Chairman Gordon H. Millar, Robert P. Clagett, (seated) M. Eugene Merchant, Joel D. Goldhar, Arnold M. Kriegler, and James E. Ashton.