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.
DESIGN AND ANALYSIS OF INTEGRATED ELECTRONICS MANUFACTURING SYSTEMS LAURENCE C. SEIFERT ABSTRACT AT&T's reaffirmation of its commitment to manufacturing excellence has resulted in a number of productivity improvement programs. Processes are now designed to survive model changes, avoid suboptimal use of robotics and computing technology, and deal with culture factors. It was first necessary to focus resources and management attention on the commitment to manufacturing. Second, a methodology was evolved that mandates, in order of priority: (1) up-front systems analysis and eng~- neering, (2) application of "suprahuman" processing facilities along with automation of material handling facilities, (3) automation of information flows, and (4) reduction in labor through development of physical automation. Projects emphasize implementation management and dealing with the human tendency to resist change. A set of needs and opportunities for the U.S. technical community to achieve productivity improvements is described. It is AT&T's experience that the current capabilities of technology, the tools that support it, and the capabilities of the engineers who must implement it are more than sufficient to meet overall goals. Simply stated, present capability is not being widely applied across U.S. manufacturing operations. INTRODUCTION The past five years have been a difficult, yet exciting, period for the manufacturing segment of American industry. At AT&T, in addition to the more publicized divesti- ture of the local Bell Telephone Operating Companies, we have experienced a revival in our commitment to manufacturing ex- cellence. This renaissance is fueled by the reaffirmation of the strategic link between manufacturing and distribution productiv- ity and the achievement of our corporate goals. We firmly believe that manufactur- ing is a principal vehicle in bringing the benefits of new technology to market. This paper, however, is intended neither as a chronicle of the events that are reshaping manufacturing nor as a dissertation on the causes. It deals instead with what we have learned from our activities in manufactur- ing productivity improvement. It also de- 12 scribes the course we have set for ourselves, and it identifies problems and opportunities for furthering productivity gains. AT&T has conducted productivity im- provement programs at various manufac- turing facilities. This experience is the base from which we have established new direc- tions for further advances in manufactur- ing productivity. We have a vision for our manufacturing business that we continue to refine. Specific productivity improvement programs and related R&D activities are being driven to achieve this vision. Quality and simplicity are the corner- stones of our programs. The key factors for realizing the programs' full benefits are · the commitment of all involved per- sonnel and levels of management, and · designing and implementing total manufacturing and distribution systems and their linkages with other systems.
DESIGN AND ANALYSIS Manufacturing operations, which once were the Western Electric Company, Inc., are now deployer] as integral parts of the AT&T business groups. A close community of manufacturing management is main- tained to facilitate intergroup operations and to leverage R&D and other corporate activities. This is accomplished through a corporate oversight function. AT&T provides information movement and management (IM&M) services and the systems and products that support these services. This paper concentrates on the sys- tems and products that are sold to custom- ers and are the basis for the IM&M services. AT&T is a vertically integrated company in the sense that it employs a broac] range of internal functional expertise in the prod- uct realization processes. Figure 1 shows the grouping of AT&T's manufacturing and distribution operations and their underlying technologies. Table 1 describes the range of in-house R&D and production activities. Basic research is not included in the dia- gram for simplicity, although we maintain our commitment to this essential activity. PWBs Connectors Materials FIGURE 1 AT&T products business groups. 13 This structure is the focus of our pro- grams in manufacturing productivity im- provement. The following functions are contained under the heading of manufac- turing: · Process design and engineering · Translation of product design infor- mation into manufacturing information · Production planning, scheduling, and control · Incoming material control · Material ordering and stocking · Product fabrication, assembly, and testing · Product repair · Quality control · Product and process productivity im- provement · Manufacturing information manage- ment We consider manufacturing and distri- bution as a single business activity. This approach integrates the following functions with those listed above: Components \ j/ Products - - - Dlstrlbutlon \ Systems, Subsystems, Power, (Repair: /Stocking, Installation, Clrcult Packs \ / ~ Staging _ ~ Network Systems r Semiconductors ~ ~ Equipment etwork Sys ma Dlscretes I ~ ~ Media HlCs | | | Data Systems | _ Large Business | Systems | _ General Business _ Systems consumer Products T Underlying Technologles Manufacturing Technology -Systems Engin~rlng -CIM -Physical Sciences Network Systems r (OTC) _ C&ES (OEM) | . Federal Systems _ Consumer Products . Business Systems | Design Technologies -CAD/CAE \
14 rials · Management of orders · Staging of customer system and mate- · Stocking of finished goods · Installation and customer service This mode} also employs product-focused operations, as compared with functional operations, which are minimized. Terms such as "focused factories" are used to de- scribe these operations. They will eventu- ally result in a merger of manufacturing and distribution operations into product family operations that are singularly man- aged. The balance of this paper deals with our experiences with productivity improvement programs and the directions we have set for design of manufacturing processes and for related R&D programs. A LEARNING EXPERIENCE Our successes have exceeded our failures, but understanding both is essential. Failure mode analysis is just as valuable in acquir- ing knowledge about integrated manufac- turing systems as it is in achieving yield improvements or in refining chemical reac- tions. LA URENCE C. SEIFER T TABLE 1 R&D and Manufacturing Activities AT&T Production R&D AT&T Contract Purchase Category Common Technologies Materials X X X Product CAD X X Manufacturing technologies Systems engineering X Systems integration X X CIM software X X X Production facilities X X X X Components Integrated circuits X X X Discrete devices X X X Interconnection technologies X X X Systems and Products Media (light guide, cable, etc.) X X X Electronic/photonic systems X X X The following generalizations, culled from our learning experiences, are pre- sented as major examples of what not to do again or as one answer to the question many managers ask, "Why does computer- integrated manufacturing fail?" Do Not Accept Process Performance As It Is Yield improvement, normally considered an element of quality improvement, almost always offers the most significant opportu- nity for improving process productivity. Design for manufacturability (DFM) and its companion, failure mode analysis (FMA3, have not always been the primary targets of productivity improvement proj- ects. Yield-quality-FMA programs that is, a total quality control (TQC) program, must be implemented before new facilities or new automation capabilities can provide the expected performance benefits. When TQC has been addressed, performance usu- ally has exceeded expectations, and addi- tional production capability has been real- ized. If capacity planning is not done on the basis of productivity goals, it is likely to deploy more capacity than is needed. This
16 fectively, with high-quality performance, assemble a large number of components on a repetitive basis. The third method is the use of a robot to perform the task. The cost of assembly by robots not only is lower but also yields higher quality by lessening the "human error" factor. The fourth method uses high-speed machines configured specif- ically for the task. Normally, odd-form components can be installed only by the first three methods, with the last method being limited to com- ponents conforming to certain physical pa- rameters. However, we have found few cases in which a circuit cannot be designed with components that are compatible with the last method of assembly. Examples of ex- ceptions to this generalization are analog and high-voltage components. The use of high-speed dedicated machines results in much lower assembly costs than can be achieved by the first two methods, a differ- ence that is greater than the differential in labor rates between the United States and most Asian countries. Preferred component use offers not only improvements in assem- bly costs but also additional benefits real- ized from material management, parts stocking, volume purchasing, test program reuse, and fewer quality and reliability en- gineering activities. A strong DFM effort between product designers and manufacturing engineers, es- pecially for component engineering in the manufacture of electronic circuit packs, has greater leverage in reducing overall costs than does automation aimed at simply re- ducing assembly labor cost. Surviving a Model Change Implementing major automation proj- ects only to have the manufacturing pro- cess not survive a model change is the most costly expenditure of all. This is why flexi- ble manufacturing is everyone's objective. It is essential, however, that one clearly un- derstands the circumstances under which LA URENCE C. SEIFER T hard or flexible capital facilities are desir- able and when more manual alternatives are appropriate. Avoid Suboptimal Use of Computing Technology Applying computing technology to vari- ous operations mandates a careful coordi- nation of all aspects of information flow. The suboptimal use of computing technol- ogy often leads to confusion, requiring en- gineering and material management per- sonnel to adjudicate differences. An example illustrates this situation. In one instance, product information came to the shop floor by three separate paths, with different functional groups independently "adding value" to the information. Fur- ther, there were 14 manual information translation points between the various in- formation systems that served the three paths. It was obvious why the information on the shop floor was inaccurate. The Culture Phenomenon Adequately dealing with people and their natural reluctance to accept change is crit- ical. How many of us have heard comments like the following: "There's nothing wrong with our manu- facturing processes; the problem is the product designers." "We can't afford to spend money on pro- ductivity programs." "We already have too many engineers." "We simply need to get labor rates down." "If only the product forecasts were more accurate." Many believe that the most difficult problem to overcome is the propensity for operating personnel to act on their own au- thority, for engineers not to listen to prob- lems, and for management to fear and resist change. As W. Edwards Deming remarked in a recent seminar, "If we know what to
DESIGN AND ANALYSIS do, why don't we do it?" We can design and implement the best manufacturing processes and facilities, but if its users are not "buying in," we have wasted time and resources. Overall, the most important element in achieving an improvement is fully compre- hending the problem. After all, if there were no problems, there would be no op- portunities for improvements. But a super- ficial understanding of a problem will lead to solutions that generate still more prob- lems. DIRECTIONS A Corporate Focus Our current approach to achieving on- going manufacturing excellence has a num- ber of elements. Foremost among these el- ements is a set of activities that focuses resources and the attention of personnel on manufacturing, thereby demonstrating a commitment to doing what is necessary to bring continuous improvement to the man- ufacturing operations. Executive attention and support, the allocation of necessary re- sources, including manufacturing engineer- ing and central R&D, and the establish- ment of performance goals are key factors for achieving success. AT&T has put in place the following activities to accomplish this. A Manufacturing Technology Board Each business group has a manufactur- ing technology board that is made up of the chief engineers of each of our manufactur- ing facilities, representative product devel- opment directors, and a manufacturing R&D director. The objectives of these boards are to develop plans for improving the productivity of the manufacturing op- erations, for overseeing the programs, and for ensuring commonality among opera- tions. Each of the directors typically over- 17 sees each specific program for example, information automation or surface mount technology implementation. A Corporate Manufacturing Officer A corporate manufacturing officer, re- porting directly to the president of the cor- poration, has been designated to facilitate manufacturing planning. This individual is charged with establishing and implement- ing a corporate manufacturing plan, which must be approved annually by the highest levels of the corporation. Organizations supporting this office consist of functional planning groups for various aspects of man- ufacturing operations, the manufacturing R&D operations, and the environmental control staff. A Manufacturing and Distribution Council A manufacturing and distribution coun- cil, consisting of all corporate officers who have direct manufacturing or distribution responsibilities and chaired by the corpo- rate manufacturing executive, has been es- tablished. The goal of the council is to fa- cilitate and oversee the development and implementation of important plans for all manufacturing and distribution functions. The council's objective is to ensure that AT&T's critical manufacturing and distri- bution resources are optimized for the ben- efit of AT&T as a whole. Internal Manufacturing R&D Capabilities The internal manufacturing R&D capa- bility was started 30 years ago. It has re- cently been supplemented with the Manu- facturing Development Center, a capability for production facilities systems integration and replication, and with a significant commitment of resources by AT&T Bell Laboratories for computer-integrated man- ufacturing (CIM) systems development.
18 | Manufacturing Systems Eng (MSE) | MSE - Process Eng -1 Characterizatlon, design *, and engineering of manufacturing processes. Design for Manutacturabillty (DFM) Informatlon Mechanizatlon Mechanizatlon = Integration ** ~ Automatlon Physical Automatlon - For: | Suprahuman Processing t/ Technology/machinery that do what humans cannot do well. | Flow Control =1/ Matorlal Handling, Queue Control Labor Productivity FIGURE 3 Manufacturing productivity realization model. High-V=bility Projects Specific high-visibility projects have been established in each business group. These projects, targeted on the group's business priorities, are supported by central R&D personnel. The purpose of these initiatives is to accelerate productivity improvements in certain operations. The projects also ef- fect a technology transfer, to the operations engineering staffs, of a number of technol- ogies, with recently increased emphasis on manufacturing systems engineering disci- plines. Projects that demonstrate significant im- provement serve as an example an exis- tence proof- of what is possible. They stimulate ongoing work and other projects. Care has to be taken, however, to ensure LA URENCE C. SEIFER T Prloritl" | MSE Process Eng ~ Design for Manufacture J First ~~ Suprahuman Proce~lng | Then / ~ Flow Control 1 Then /1 Informatlon Mechanizatlon I Then Labor Productivity La t Automatlon s * Emphasis: Customer expectatlons/simpilcity/TQC, with JIT, In-llning, flexibility concepts ** Linkages: product CAD, distrlbutlon, financial, and other Informatlon systems that this is not just a Hawthorne effect.) It is essential that the disciplines necessary for ongoing improvement, as well as the desire for ongoing improvement, be built into the fabric of the operations. Methodology The second element of the program has been the establishment of a methodology for productivity improvement programs. Previous experiences have led us to follow a sequence of activities, as shown in Figure 3. The methodological Hawthorne effect is usually defined as the problem arising in field experiments when subjects' knowledge that they are in an experi- ment alters their behavior from what it would have been without that knowledge. Documentation of this effect came in the classic studies at the Hawthorne Works of Western Electric Co. (Mayo, 1945~.
DESIGN AND ANALYSIS The first principle of the methodology is that redesign and reengineering of the pro- cesses must take place before an attempt is made to apply either information automa- tion or physical automation. We call this discipline manufacturing systems engineer- ing (MSE). Most benefits come from simpli- fying systems, improving yields, and reduc- ing the cost of quality. The techniques of MSE and the tools that support them are based in traditional in- dustrial engineering practice. The current tools, however, are more powerful because of today's computing technology. Central R&D resources have been valuable in de- veloping these new tools, establishing tech- nology transfer procedures, and supporting specific projects. We have embarked on a major program aimed at fostering the use of systems analy- sis disciplines and process design and engi- neering tools. Figure 4 shows current results 100 80 a, ID - 60 = - o ~ 40 cat 20 o 19 by category and number of tools that are supported centrally by R&D. The percent- age of facilities using these tools might be somewhat misleading, as it was calculated by the number of locations using the tools at least once. Use has been accelerating, and the results are encouraging. An annual technical symposium is held for AT&T manufacturing systems engineers, papers are presented, and much "networking" goes on. Attendance has grown from 35 in 1985, to 85 in 1986, and more than 200 in 1987. Tools developed or supported by the R&D staff are listed in the appendix to this paper (Eckel, 1986~. The second principle of the methodology is that manufacturing is information-inten- sive. Many believe that mechanization of information transfer through computer technology will result in significant im- provements in productivity. Some have suc- ceeded and others have failed to achieve Legend C~ Engineerlng - Studles ~ Operatlons - Ongolng Use . ~ Capacity Simulation Design Data Anal. Quality Reliability Sched. 2 2 1 2 2 2 FIGURE 4 Implementation of process engineering software tools at AT&T manufacturing facilities. The number centrally supported is shown beneath each column.
20 this promise of CIM. CIM can become the vehicle for embedding procedural disci- pline in operations by reducing the need for manual intervention. Since mechanization equals automation plus integration, the major benefits of information mechaniza- tion come from the integration component of that relation. Providing a customized information sys- tem for each operation can be expensive, both for initial implementation and on- going enhancements. Therefore, AT&T has developed a set of corporately supported systems using modularized open systems software architectures. Systems meeting the requirements of various production pro- cesses can be individually configured. They also provide users some flexibility to better adapt the systems to their needs. Difficulty arises in interfacing the array of various suppliers' production facilities with the system. Most new manufacturing equipment, which once had only program logic controllers (PLCs), now requires in- creasing amounts of computing power. There are few information or data-base Production management, enabilng, planning, support Production execution 4 Individual shops 3 Cells 2 Workstations 1 ~ 1 Equipment FIGURE 5 CIM systems response time requirements. LA URENCE C. SEIFER T standards or interface protocols in most manufacturing operations. Many efforts are uncler way now in this country to standard- ize the interfaces to obviate this problem, the most notable of which is the manufac- turing automation protocol (MAP), cham- pioned by General Motors Jones, 1988~. Figure 5 shows the levels in the computer-integrated manufacturing archi- tecture (CIMA) that AT&T has adopted for internal use. Also shown are the system de- sign requirements for response times by level. These requirements result from traf- fic studies of information flows in many advanced operations. Figure 6 is a systems diagram of the fam- ily of factory floor systems. The systems family that has been chosen to be supported with a corporate information system, called productivity improvement systems for manufacturing (PRISM), has been initially structured for product assembly and testing operations. The integrates] circuit fabrica- tion factories use another, internally devel- oped, family of systems. Figure 6 uses the following acronyms anti abbreviations: 1 /l 7 Corporate ,l 6B Factory 6A Factory within factory 5 Floor(s) Wide spectrum of appilcatlons and response times t Predictable response 1 ~ ( - 0.2 see) ITlght coupilng (10 ms-0.2 see) Application dependent Quick and accurate response (Down to 10 ms)
DESIGN AND ANALYSIS ARX/PPS Accounting . 21 IMPAC/AMAPS | St pollers ~ Purchasing ~ MRP-: ·: UNICAD Desl' an Factory Operations _ MOVES Recelving i/1 PRISM future candlciates Materlals flow - ~ Informatlons flow FIGURE 6 PRISM area of focus. SFC SPECS MOVES MPCS-CP ARX PPS IMPAC 1 s ~ ~i -'es 1 Shop Floor Engineerlng Control Operatlons ~ 1 ' MO - Store room Kltting Mfg. mgmt. MPCS-CP | ~///S///~ I MPCS-Fo I . Clrcult Pack Assy. & T - t i//// Clrcult pack store room ~/////////~ i//////// ~////////~ | Malntenan~ | Quallty 1 L Warranty 1 Equipment I Assy. & Test | Shipping | Customer ~ Dlstrlbutlon Shop Floor Control Synchronized Production Engineering Control System Materials Operations Velocity System Manufacturing Process Control System Circuit Packs MPCS-EQ Manufacturing Process Control System Equipment MPCS-LOT Manufacturing Process Control System Lot Processing Accounting Receiving Executive System Planning Procurement System Integrated Manufactur- ing Planning and Control AMAPS MRP-II UNICAD Advanced Manufacturing Accounting Production System Manufacturing Resource Planning Unified Computer-Aided Design For those functions shown without an in- formation system title, local systems are currently in use. Rather than take the reader through a long discourse on the functions performed by each system, suffice it to say that we have accomplished and implemented sev- eral paperless information streams. For ex- ample, we are able to generate assembly aids for shop personnel electronically through direct linkage with the product de- signers' CAD system. The current status of deployment of these
22 30 25 20 ID - a' _ 15 o At 10 s o Legend Under consideration O Committed ~ OperaUng 1/87 MPCS MPCS CP LOT , . MPCS ED ///// _ . MPCS SPECS SFC LA URENCE C. SEIFER T i. MOVES FUGUE 7 PRISM deployment. systems is shown in Figure 7. Of course, AT&T 3B computers are used, and all soft- ware runs under the UNIX operating sys- tem. What has been described is a success- ful CIM program. It is an emulation of the approach AT&T has used for similar sys- tems that we have been providing for tele- phone companies for more than IS years. Figure 8 shows the structure of a typical system. Highlights of the approach are these: · Users are involved in writing system requirements. · Users are involved in managing devel- opment programs and priorities. · A structured programming develop- ment environment is maintained. · Full commercial documentation, train- ing, and maintenance support, including 24-hour on-call backup, are provided. · Software is structured to (a) allow unique application configurations; (b) ac- commodate site-specific programs and in- terfaces; and (c) provide users with some design space. We are now in the process of developing real-time versions of several of the engineer- ing and quality analysis tools discussed. When these tools are embedded in PRISM systems, a whole new set of capabilities will be available to local engineering and oper- ating personnel to achieve unprecedented · or... . Operatmg etnclencles. Advances in microcomputing are now making it possible to achieve true real-time process control with no human interven- tion. Much of our R&D effort is aimed at exploiting this opportunity. We have al- ready built and installed workstations that include all elements of the process control model shown in Figure 9 and are connect- ing these workstations to PRISM systems. In one example of this procedure, a modi- fication developed for photolithography
DESIGN AND ANALYSIS .~ : :;:;;; ;;; ;.;; ,:. Loading . : ' :ALGS ~. rMRP-II l | IMPAC/AMAPS | TRANSACTION PROCESSING Communications, WIP, Dropouts, Completions, Priority control, Serlalizatlon MVINC a)// CB ~ ~ Routing ~ MPCS:flexiln,: system a: FIGURE 8 PRISM open architecture SFC example. steppers provides an analysis of processed silicon wafers, knowledge-based analysis, and direct feedback control to a motor- driven platen to adjust for microscopic de- viations. 3. Cell 23 ~- MPCS ~ CP 1 ~ I l --- I · BEE Core Common ~ Site specific The third principle of the methodology is that physical automation is introduced to replace labor or to facilitate product flow only after completion of the aforemen- tioned product and information system de- Controller 2. Workstation - ~1 ~ ~ (Process model) 1. Equipment t CIMA levels FIGURE 9 Process control model. Effectors Sensors Process _
24 signs. Physical automation should be ap- plied first where there is no choice but to use machinery that is "suprahuman pro- cessing." For example, as more surface- mount technology is deployed for the as- sembly of components to circuit boards, it has been found that high-pin-out IC pack- ages with a pitch of 25 mils or less between leads require assembly by vision-assisted machines for accurate placement. Supra- human processing, obviously, is the domi- nant mode in semiconductor wafer fabri- cation. But much IC packaging continues to be done in a more manual mode. Process discipline and control can be facilitated, as with information flows, by automation of product and material handling facilities. The second choice for automation is for better control over the operation of the process. After all of the other priorities are satis- fied, the final choice for automation is the use of machines for the replacement of manual labor. We all have visions of totally automated processes, the so-called lights- out factory. Great care must be exercised, however, not to generate a system that re- quires more ongoing engineering labor for product changes and enhancements than has been reduced in the assembly and test operations through automation. PROGRAM IMPLEMENTATION The best designs and most accurately fo- cused programs are of no value if they are not implemented or if they are not proper- ly used after implementation. Careful "de- sign for implementation," implementation planning and management, and user "buy- in" are required for successful programs to improve manufacturing productivity. The most successful approach has been the one in which the final user "project manages" the program. This approach leads to more complete consideration of how new oper- ating disciplines affect the users. People are more likely to follow design intentions and LA URENCE C. SEIFER T deal with change if they are a party to gen- erating the change. There has been a great deal of debate over whether it is better to rearrange a manufacturing process in an existing oper- ation, while production continues, or to build an all-new process in a new location, the "greenfield approach." The advantage of the greenfield approach is that produc- tion is not disrupted. Verification proce- dures can be accomplished with limited ex- ternal variables. The disadvantages are that the new process is not tested in real produc- tion situations. Furthermore, there are ad- ditional costs for product samples and du- plicate facilities. The greenfield approach is best used for production lines for new prod- ucts. The advantages of rearranging exist- ing processes include reuse of existing facil- ities and, probably most importantly, the availability of an analytical characteriza- tion of process performance. Rearrange- ment while in production requires a much closer working relationship with the oper- ating personnel. In this situation, the han- doff of the new process designs and tech- nology is evolutionary, and the technology transfer to the operating personnel is con- tinuous. SUCCESSES Projects that followed AT&T's manufac- turing productivity realization model (Fig- ure 3) have yielded remarkable results. Two such examples are the power unit shop at the AT&T Denver Works and the entire operation of the AT&T Oklahoma City Works. Both factories are product assembly and test operations and can be character- ized by the model shown in Figure 10. These are "focused factories" in that they perform all manufacturing functions de- scribed earlier, with the exception of com- ponent fabrication. These are large factories and, for certain product families, are separately managed as "factories within a factory." This ap-
DESIGN AND ANALYSIS Logical process: Receiving Incoming component certification b. ~ do JO _ Kltting ~ _ Storeroom . _| Cables I assembly I FIGURE TO Assembly and test operations model. proach has been likened to shopping malls, where certain common services (e.g., secu- rity, heat, and light) are shared on a bill- able basis, but each store unit is responsible for its own business performance. The Denver power unit shop employed a few dozen people and represented only a portion of Denver's operations (see block "g" of Figure 10~. Since it had always pre- sented problems, it was identified as the pilot for the Denver Works improvement program. The operations were analyzed and characterized, and it was decided to 25 f. ClrcuK pack assembly and test ~1 Cabinet and backplane _ assembly h. Shipping Systems test reengineer the process to an in-line, just- in-time, manual kanhan operation. The en- gineers concerned set up desks on the shop floor and involved shop personnel in design and implementation phases of the project. They stayed with the shop until everyone was comfortable that the performance of the new process met objectives. No physical or information automation was involved. The results are shown in Figure 11. Oklahoma City Works employs approxi- mately 6,500 people. The manufacturing operation for the digital switching system Power Unit Assembly Shop Application of Just-ln-Time (JIT) and Total Quality Control (TQC) Techniques Reduction (%) 100 OutpuVEmployee Product velocity Process time In-shop time 64% 1 61%= Increase (%) 200 400 600 800 1 1 1 1 233% Work In process Inventory Process down time Process analyals and enginoorlng time: 1 Month Initial Implementation Interval: 1 Month FIGURE 1 1 Denver Works productivity improvement initiative. 1 1 1 7s0%
26 100 80 3 0 60 - o ~ 40 CL 20 o Manuhcturlng Cost (Labor, Load, Materlale) 67% 49% 45% Material Labor Load . _ _ Manufacturer Labor 10 I ~ / ~ . 1 ~ . ~ . ~ . ~ . U) {D mmoom A__ US {D {D to ~ _ _ Cal 8 Q CL 6 o ~ 4 lo 2 o LA URENCE C. SEIFER T Shipments Mlillons Un" ~ U. tD _ _ _ FIGURE 12 Oklahoma City Works productivity improvements, SESS systems. product family is the largest factory-within- a-factory at Oklahoma City. The program followed the mode! shown in Figure 3. In- formation automation in the form of the PRISM systems was deployed in Oklahoma 30 20 City beginning in 1984. The results are shown in Figures 12 through 14 for costs and capacity, cost of quality, and intervals. It should be noted that cost improvements were realized at about the same rate across 10 ; L n \ - _ _ _ _ - 1983 1984 1985 1986 1987 Year FIGURE 13 Oklahoma City Works quality/productivity (cost of quality includes prevention, appraisal, and failure).
on 0 - o' 0 o' ~ jot A of a, a) · . - ._ ,0 ~ Y on O Q 0 by ~ off' a, O o $ 0 O ~ In ' 1 o a, 1 ... =~ _, - 1, ~ us {D of on _ _ - - 1 <a \ \\ 0 \ _ \ _ _ _ 27 \ \ ~ a, 7\ \ ~ = ''\\L \\ \~ - - ~P on As. - - o on - Cd - c5 . be . Cal Cal ~4 Cd Ct Cal Cal Cal I` ~ Cal 8 Q c~ - o o~ m ~ u' (D .= bO .= c~ - ~ ct - c~ ~ - ~ ~ o I-' 3 ~o .~= C) ~ Ct Ct .~: ° o o4 o o C) C~ .=
28 4: In 3 as lo 3 o t ID ID c, ~ 1 LA URENCE C. SEIFER T By" OL \ l 1983 \ \ ~ \\ ~ . "art 1984 1985 1986 Year ---em Digital IC --I Translators * Memories - O Capacitors ~ Linear IC A Mlcroproc FIGURE 15 Component incoming quality at Oklahoma City Works. labor, load, and material. Labor costs were halved, although no effort was made to di- rectly replace human labor with robotics or any other physical automation. The labor reductions came from yield, quality, ant! throughput improvements. Material cost reductions were achieved by cooperative ef- forts of product designers and manufactur- ing engineers, including a major effort with all parts suppliers, both in-house and out- side, to improve incoming quality levels. Figure 15 shows actual incoming compo- nent quality levels. This is exceedingly im- portant, since 80 percent of the defects found during testing of the circuit packs result from problems in the quality of the components. The knowledge derived from component engineering, process analysis, and FMA ac- tivities is now being used in the most pro- ductive of all activities, DFM. Figure 16 shows the value of DFM in yields of first production runs for new circuit packs. We consider these results for complex circuit packs, using state-of-the-art devices, to be remarkable. A study was conducted in 1986 to try to evaluate the elements needed for world- cIass manufacturing productivity. Table 2 ranks those elements based on available quantifications of savings and also by a sur- vey of involved individuals. The latter in- cluded management, engineers, factory workers, and product designers. The fact that "teamwork" ranked highest in impor- tance emphasizes the value of treating the human element in the process of improving manufacturing productivity. The productivity improvement tech- niques described have also been applied to the manufacture of integrated circuits. As technological advances drive more and more circuit functions onto ICs and IC packages, the performance of the semicon- ductor fabrication processes will dominate manufacturing viability. These processes are more complex, yields are usually low, facilities are capital-intensive, and use of
DESIGN AND ANALYSIS 100 80 60 40 20 JFMAMJJASONDJFMAMJJ ASONDJ FMAMJ JASON D 1984 1985 1986 Start-up Month FIGURE 16 First production runs, new circuit packs. state-of-the-art technology is normal. The manufacturing productivity realization model (Figure 3) is still valid, however. Figure 17 shows the results achieved in a silicon wafer fabrication process using this technique. CONCLUSION These successes are an "existence proof' of the value of first applying systems engi- neering disciplines to manufacturing proj- ects, followed by information mechaniza- tion (automation plus integration), and 29 lastly the deployment of optional physical automation. There is significant room for improvement in manufacturing in the Unitecl States, and it is achievable within the capabilities of existing engineering dis- ciplines and tools. NEEDS AND OPPORTUNITIES Many opportunities exist to further pro- ductivity gains in U.S. manufacturing. The following is a list of the most important areas for improvement: TABLE 2 Elements of World-Class Manufacturing, 1986 Study of Oklahoma City Works Ranking by Quantifiable Savings Ranking by Consensus of Participants 1. Design for manufacture 2. Productivity improvement program 3. Quality management 4. Application of new process technology 5. Manufacturing research planning 6. Manufacturing systems engineering Not measurable: Customer satisfaction Teamwork Training 1. Teamwork 2. Quality management 3. Customer satisfaction 4. Manufacturing systems engineering 5. Design for manufacture 6. Training 7. Productivity improvement program 8. Manufacturing resource planning 9. Application of new process technology
30 Crystal growth Wafer preparation l Epita~cial growth l Davico fabrication / / Cleanino l _ Deposition | | Utho' araphy ~3 1 nsoact measure Etching _ Doping Device packaging \ \ FIGURE 17 Performance improvement in semiconductor manufacturing processes. · The systems engineering analytical dis- cipline and its software tools (simulation, queueing, etc.), although readily available and rich in capability, are not as widely used as they could be. Probable causes for this are inadequate training of engineers and the lack of awareness of the benefits by corporate and manufacturing manage- ment. Efforts must be expanded to make the manufacturing community aware of these capabilities and to encourage the use of these systems. · Design for manufacturability currently uses rather basic techniques. More ad- vanced analytical techniques and tools for product designers are needed for linking product performance to process capabili- ties. · Lack of computer interface standards LA URENCE C. SEIFER T MOS InitIatIve-Results 7/86-12/86 · Interval decreased 40% · Capacity Increased 10% · Yield Increased 27% · Inventory decreased 40% among suppliers of production equipment impedes project implementation. Too much specialized "translation" software is re- quired for CIM connection, thus reducing information processing effectiveness and adding to engineering costs. Semiconductor Equipment Communications Standard (SECS) and GM-S MAP standards activities are examples of attempts to rectify the situ- ation. · Opportunities exist for advancing the capabilities of production processes by ap- plying new real-time computing technology capabilities to the following: process con- trol; embedding software tools into CIM platforms, using system performance anal- ysis, with "knowledge-based" and artificial intelligence techniques, and including ad- vanced routing, scheduling, and comput-
DESIGN AND ANALYSIS erized kanhan capabilities in CIM plat- forms to further the production flexibility. Application of anthropomorphic robots for flexible assembly may well serve the assem- bly of large items. Flexible approaches for suprahuman machinery capabilities, how- ever, are needed for electronics and pho- tonics equipment and integrated circuit fabrication. 31 REFERENCES Eckel, E. J. 1986. Quality in AT&T Network Systems. AT&T Technology Journal 65~2) :30-38. Morris- town, N.J.: AT&T Network Systems Group. Jones, V. C. 1988. MAP/TOP Networking: A Foun- dation for Computer-Integrated Manufacturing. New York: McGraw-Hill. Mayo, E. 1945. The Social Problems of an Industrial Civilization. Cambridge, Mass.: Harvard Univer- sity Press. APPENDIX AT&T MANUFACTURING SYSTEMS ENGINEERING TOOLS ENGINEERING Capacity SESAME SCHEDULING EVALUATION SYSTEM FOR AUTOMATED MANUFACTURING ENVIRONMENTS Determines the manufacturing interval, maximum utilization rate, number of shifts required, and number of machines required to maintain production rate for insertion machines that produce high-runner codes. DCM DETERMINISTIC CAPACITY MODEL LB/L QNA Simulation Models GSM Estimates the production capacity and machine utilization in a clean room as a function of the product mix. LINE BALANCE/LAYOUT Determines the minimum number of workstations and balances the workload among the stations to meet throughput require- ments in assembly lines. QUEUEING NETWORK ANALYZER Provides estimate of work in progress, interval, and utilization in general manufacturing systems using analytic queueing tech- niques. GENERALIZED SIMULATION MODEL Provides a generalized simulation model for complex manufactur- ing systems and estimates work in progress, interval, and level of utilization. This is a tool for addressing complex issues. PERFORMANCE ANALYSIS WORKSTATION Builds graphics-based simulation models of queueing networks. This tool is used to answer design and engineering questions quickly.
32 LA URENCE C. SEIFER T Design QCAP QUALITY AND COST ANALYSIS PLAN Analyzes the design and manufacturing process of an assembled electronic product. This tool is used to evaluate design, improve process, enhance quality, and decrease manufacturing cost. OPERATIONS Data Analysis and Monitoring BCR RTAS Quality PCAP BAR CODE READER Provides a systematic mechanism for tracking and mon- itoring portable, rechargeable bar code readers. REAL-TIME ANALYSIS SYSTEM Uses statistical techniques to monitor and analyze a manufactur- ing process in real time and improve efficiency and output by identifying problems as they occur. AGDAT ADVANCED GRAPHIC DATA ANALYSIS TOOL Displays and analyzes defect data using control charts. PROCESS CHARACTERIZATION ANALYSIS PACKAGE Analyzes and summarizes large quantities of process and test data used by engineering. ILIAD I LEARN. I AID DIAGNOSIS Employs an improved technique to "learn" effective troubleshoot- ing methods from operators. Newly acquired repair knowledge is automatically added to ILIAD's knowledge base on a daily basis ensuring its responsiveness to ongoing changes in the manu- facturing process. In addition, ILIAD produces a number of anal- ysis reports that help engineers and operators identify the root causes of many problems in the product as well as in the process. Reliability STAR STATISTICAL ANALYSIS RELIABILITY TOOL Identifies reliability problems and assists engineers with solving them by using statistical analysis techniques. SUPER SYSTEM USED FOR PREDICTION AND EVALUATION OF RELIABILITY Predicts system reliability at any phase of product design using a variety of reliability modeling techniques.
DESIGN AND ANALYSIS Scheduling DMIS FAS LBS 33 DYNAMIC MANAGEMENT INFORMATION SYSTEM Estimates anticipated flow of product in clean rooms as an aid for scheduling of machines. SCHEDULING TOOL BASED ON GROUP TECHNOLOGY Used to schedule lines with medium- to low-volume codes by characterizing product families. LIGHTWAVE INTEGRATED TECHNIQUE FOR ENGINEER- ING SPANS Blends fibers with differing performance specifications to meet customer specifications. CABLE PRODUCTION SCHEDULER Sequences jobs through a production line choosing appropriately from candidate machines and produces machine assignments and job sequences for each machine. FINAL ASSEMBLY SCHEDULER Sequences final assembly in an attempt to smooth the demand on feeder shops. LOADING, BUFFER SIZE, SEQUENCING Quickly estimates production rate and machine use on circuit pack assembly lines and estimates effect of loading, lot sizes, and buffers on those rates. NCMS NUMERICAL CONTROL MACHINE SEQUENCER Used to reduce processing times by making drilling, component insertion, and metal-punching operations more efficient.
