Chapter 3

Programmable AutomationTechnologies

ContentsPage

summary .. .. ... .. .$....... . . . . . . . . . . . . . . . . . . . . . . .. ... . ....... 33

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Discrete Manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35The Manufacturing Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Functional Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Computer-Aided Design (CAD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Computer-Aided Manufacturing (CAM) Technologies . . . . . . . . . . . . . . . . . 48Programmable Automation and Manufacturing Management. . . . . . . . . . 69

Technical Trends and Barriers: Future Applications . . . . . . . . . . . . . . . . . . . 74Trends and Barriers in Five Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . 74Technical Trends and Barriers: Cross-Cutting Issues . . . . . . . . . . . . . . . . . 83

Future of the Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93Capabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93Future Levels of Use ofl?rogrammable Automation . . . . . . . . . . . . . . . . . 97

TablesTable No. Page

5. Principal Programmable Automation Techniques. . . . . . . . . . . . . . . . . . . 356.0perating Robot Installations, End of 1982 . . . . . . . . . . . . . . . . . . . . . . . 527. Examples occurrent Robot Applications . . . . . . . . . . . . . . . . . . . . . . . . . 548. Estimated Total Machine Tools in the United States . . . . . . . . . . . . . . . 599.Age of Machine Tools in Seven Industrial Nations . . . . . . . . . . . . . . . . . 61

10. Machine Tool Task Force Recommendations for Improving MachineTool Controls.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

11. CAD: Projections for Solution of Key Problems . . . . . . . . . . . . . . . . . . . . 9412. Robotics: Projections for Solution of Key Problems . . . . . . . . . . . . . . . . 9513. NC Machine Tools: Projections for Solution of Key Problems . . . . . . . . 9514. FMS: Projections for Solution of Key Problems . . . . . . . . . . . . . . . . . . . . 9615. CIM: Projections for Solution of Key Problems . . . . . . . . . . . . . . . . . . . . 96

FiguresFigure No. Page

l. Characteristics of Metal working Production, By Lot Size . . . . . . . . . . . 362. Organizational Diagram of a Manufacturing Firm . . . . . . . . . . . . . . . . . . 383. Steps in the Manufacturing Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394. Fundamental Operations in Metalworking . . . . . . . . . . . . . . . . . . . . . . . . 405. Sample Robot Grippers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516. Actual and Projected U.S. Annual Robot Sales and Installed Base

Through 1992 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537. Sample APT Computer Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588. Total Number of Numerically Controlled Machine Tools

in U.S. Metalworking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599. Programmable Automation Factory Hierarchy. . . . . . . . . . . . . . . . . . . . . 73

10. NBS Scheme for Distributed Factory Control . . . . . . . . . . . . . . . . . . . . . 8311. Sample Rule from the MYCIN Expert System . . . . . . . . . . . . . . . . . . . . 8512. Sample “Advice” From the MYCIN Expert System. . . . . . . . . . . . . . . . 8613. Machine Vision Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

Chapter 3

Programmable Automation Technologies

Summarv

This chapter is both a primer on program-mable automation (PA) tools and their poten-tial applications in manufacturing, and an as-sessment of the important problems and direc-tions for development of the technologies. Asdefined here, programmable automation in-cludes computer-aided design (CAD); comput-er-aided manufacturing tools—e.g., robotics,numerically controlled (NC) machine tools,flexible manufacturing systems (FMS), andautomated materials handling (AMH); andcomputer-aided techniques for management—e.g., management information systems (MIS)and computer-aided planning (CAP). Whensystems for design, manufacturing, and man-agement are used together in a coordinatedsystem, the result is computer-integratedmanufacturing (CIM).

The context for this analysis is primarilydiscrete manufacturing, as opposed to contin-uous-process industries such as chemicals orpaper. Discrete manufacturing includes a widerange of traditional metalworking industries(e.g., automobiles and farm equipment) as wellas other industries which are not primarilymetalworking (e.g., electronics). Of particularnote is that a great many of the products ofdiscrete manufacturing are made in batchesof perhaps a few dozen to a few hundred units.Because of this, it is often not economical touse single-purpose, automated machines(known as “fixed” or “hard” automation) tomanufacture the product. In such an environ-ment, programmable automation is potentiallyvery useful.

The essential difference between conven-tional factory machines and programmableautomation is the latter’s use of informationtechnology to provide machine control andcommunication. The use of computers andcommunications systems allows these ma-

U

chines to perform a greater variety of tasksthan fixed automation can perform, and toautomate some tasks which previously neces-sitated direct human control.

Programmable automation can respond tosome of the central problems of manufactur-ing. These include enhancing information flow,improving coordination, and increasing effi-ciency and flexibility (defined as both therange of products and volume of a specificproduct which a factory can economically pro-duce). By using programmable automation toaddress these problems, manufacturers hopeto increase their productivity and control overthe manufacturing process.

Though labor savings seem to be the mostobvious benefit of automation, savings throughmore efficient use of materials may be moresignificant in many manufacturing environ-ments. In particular, flexible manufacturingsystems can reduce waste, reduce levels offinished product inventory, and reduce themanufacturer’s substantial investment in theproducts that are in various stages of comple-tion, known as “work in process. ”

Some of the technical factors which holdback PA’s potential uses in manufacturinginclude relatively cumbersome programinglanguages, a general level of technical imma-turity in many areas of the technologies, long-established organizational barriers in industry(e.g., between manufacturing and design en-gineers), and the embryonic nature of effortsto maximize the effectiveness of man-machineinteractions.

Nevertheless, the technologies appear to bequite adequate technically for the vast majori-ty of near-term applications; there seems tobe a significant backlog ofwhich manufacturers haveexploit.

available toolsonly begun to

33

34 ● computerized Manufacturing Automation: Employment, Education, and the Workplace. —

The use of PA tools in integrated systems–e.g., FMS or CIM—is much more powerfulthan their use for a single task or process.Such integration not only magnifies the pro-ductivity and efficiency benefits of PA, butalso tends to induce changes in all parts of thefactory. Management strategies, product de-signs, and materials flow all change to bestmake use of such integrated systems.

Many industrialists have a vision of CIMthat includes maximum use of PA tools andcoordination between them, with few if anyhuman workers. Others downplay CIM as arevolutionary change and emphasize that fac-tories will adopt automation technologies asappropriate. It may not be appropriate (or economical) to remove all or most humans frommany factories. In any case, the widespreaduse of CIM and virtually unmanned factories

are unlikely to arise before the turn of thecentury.

Principal themes in the future developmentof PA technologies include increasing theirversatility and power, enhancing their capa-bility to operate without human intervention,and developing the ability of the tools to beintegrated. Researchers and industry spokes-men report progress in virtually all the fun-damental technical areas, although many ofthe currently identified problems in program-mable automation are complex enough to keepresearchers busy for many years to come. Ac-cording to many experts, the 1990’s may bringmany major technical advances which couldsignificantly expand the range of problems towhich programmable automation can beapplied.

Introduction

The purpose of this chapter is to describethe technologies that together comprise “pro-g ammable automation, ” and to evaluate theirusefulness for manufacturing. In addition, thechapter examines how the technologies areevolving and what can be expected for the ca-pabilities and applications of these tools.

Programm able automation refers to a familyof technologies that lie at the intersection ofcomputer science and manufacturing engineer-ing. “Programmable’ means that they can beswitched from one task to another with rela-tive ease by changing the (usually) computer-ized instructions; “automation” implies thatthey perform a significant part of their func-tions without direct human intervention. Thecommon element in these tools that makesthem different from traditional manufacturingtools is their use of the computer to manipu-late and store data, and the use of relatedmicroelectronics technology to allow commu-

nication of data to other machines in thefactory.*

There are three general categories of func-tions which these tools perform-they are usedto help design products, to help manufacture(both fabricate and assemble) products on thefactory floor, and to assist in management ofmany factory operations. Table 5 outlines theprincipal technologies included in these cate-gories, each of which will be described in thenext section.

*Although “programmable automation” is less common thansome of the other terms used to describe automation technolo-gies, it is a relatively simple and unambiguous term for the toolsdiscussed here. “CAD/CAM” (computer-aided design/computer-aided manufacturing) is a catch-all term used in industry jour-nals and popular articles to refer to a set of technologies similarto the set defined here as programmable automation. However,CAD/CAM is also used to describe some specific computer-aideddesign systems, or to denote the integration of computer-aideddesign and manufacturing. Because of this ambiguity, the termwill not be used here. “Robotics” is another term that issometimes used in a broad sense to mean not only robots butthe whole family of automation tools.

Ch. 3—Programmab/e Automation Technologies ● 3 5

Table 5.—Principal ProgrammableAutomation Technologies

1. Computer-aided design (CAD)A. Computer-aided draftingB. Computer-aided engineering (CAE)

Il. Computer-aided manufacturing (CAM)A. RobotsB. Numerically controlled (NC) machine toolsC. Flexible manufacturing systems (FMS)D. Automated materials handling (AMH) and

automated storage and retrieval systems (AS/RS)

Ill. Tools and strategies for manufacturing managementA. Computer-integrated manufacturing (CIM)B. Management information systems (MIS)C. Computer-aided planning (CAP) and computer-

aided process planning (CAPP)NOTE: Bold type indicates technologies on which this report concentrates

SOURCE Off Ice of Technology Assessment

fact, the goal of much current research in auto-mation systems is to break down the barriersbetween them so that design and manufactur-ing systems are inextricably linked. However,these three categories are useful to frame thediscussion, particularly since they correspondto the organization of a typical manufactur-ing firm.

Further, this report does not attempt tocover each of the technologies in equal detail.It concentrates on those five which appear inbold type in table 5 because they are the coretechnologies of PA and their potential uses aremost extensive.

The three categories of automation technol-ogies-tools for design, manufacturing, andmanagement-are not mutually exclusive. In

Discrete Manufacturing

Some background about manufacturing isimportant to provide a context for assessingthe usefulness of these tools. Programmableautomation can affect many kinds of industry.This report focuses on PA applications for dis-crete manufacturing-the design, manufactureand assembly of products ranging from boltsto aircraft. The report does not systematicallycover nonmanufacturing applications such asarchitecture, or continuous-process manufac-turing-e.g., chemicals, paper, and steel. Otherrecent OTA reports have examined technolog-ical changes affecting process industries.1

Electronics manufacturing industries do notfit neatly into a discrete v. process classifica-tion. Some areas, particularly the fabricationof semiconductors, most resemble continuous-process manufacturing. Other portions suchas circuit board assembly are more discrete.

‘Cf. U.S. Industrial Competitiveness: A Comparison of Steel,Electronics, and Automobiles (Washington, D. C.: U.S. Con-gress, Office of Technology Assessment, OTA-ISC-135, July1982); Technology and Steel Industry Competitiveness(Washington, D. C.: U.S. Congress, Office of Technology Assess-ment, OTA-M-1 22, June 1980).

Because electronics industries have been lead-ers among metalworking firms in both produc-ing and using computerized factory automa-tion, they play a key role in this report.

To many industrialists, discrete manufac-turing means metalworking for mechanical ap-plications-shaping, forming, and finishingmetals into usable products such as engineblocks. However, an increasing proportion ofmechanical parts manufacturing involves plas-tics, fiber composites, or new, durable ceram-ics. These new materials both enable new pro-duction processes and are themselves affectedby automation technologies.

One way in which discrete manufacturingplants can be categorized that is especially rel-evant to automation applications is the voi-ume of a given part that they produce. As fig-ure 1 indicates, discrete manufacturing repre-sents a continuum from piece or custom pro-duction of a single part to mass production ofmany thousands. Although many people aremost familiar with mass-production factories,with their assembly and transfer lines, it isestimated that mass production accounts foronly 20 percent of metalworking parts pro-

36 Computerized Manufacturing Automation: Employment, Education, and the Workplace—

Figure 1 .—Characteristics of Metalworking Production, By Lot Size

Type ofproduction:

Piece Batch Mass1000/0

Highestimate

Lowestimate

10006

00,0 ● 1.

Largecomplex 1-1o 10-300part

Smallsimple 1-300 300-15,000part

Aircraft,Typical largeproducts turbines,

centrifuges

Typicalmachines

Manual machinetools, or stand-alone numericallycontrolled (NC)machine tools

Marine engines,large electricmotors,tractors

NC machine toolswith automated part-handlingmachining cellflexiblemanufacturingsystem

IOver 200

Over 10,000

Autos,fasteners,smallappliances

Transfer lines,dedicated orspecial machines

SOURCE Machine Tool Task Force Technology of Machine Took. October 1980

Ch. 3—Programmable Automation Technologjes ● 3 7—

duced in the United States, while 75 percentare made in a “batch” environment. z The def-inition of a “batch” varies according to thecomplexity of the part and the characteristicsof the industry. A common characteristic ofbatch manufacturing is that there is notenough volume to justify specialized machines(known as “hard automation”) to automatical-ly produce the part. The direct labor involvedin fabricating products in batches is relative-ly high (as shown in fig. 1), and constitutes alarge proportion of the cost of the item. Thesecharacteristics of batch manufacturing—itsprevalence, and its low level of automation andcorrespondingly high level of labor content—are important because they suggest a broadrange of uses for programmable automation.

The Manufacturing Process

Figure 2 illustrates the organization of ahypothetical metalworking manufacturingplant. Most of the elements in this diagramare present in some form in each plant, al-though factories are tremendously varied insize, nature and variety of products, and pro-duction technologies. One automobile factoryin New Jersey, for example, assembles 1,000cars per day in two models (sedan and wagon)with 4,000 employees; a small Connecticut ma-chine shop, by contrast, employs 10 people tomake hundreds of different metal parts foraircraft and medical equipment, typically inbatches of approximately 250.3

As illustrated in figure 3, the manufactur-ing process usually begins when managementdecides to make a new product based on in-formation from its marketing staff, or (in thecase of the many factories which produce partsof other companies’ products) management re-ceives a contract to produce a certain part.

‘M. E. Merchant, “The Inexorable Push for Automated Pro-duction, ” Production h’ngineering, January 1977, pp. 44-49.This 75 percent figure has become something of a legend in themetalworking industry largely through Merchant’s writings,though he notes that he has lost track of the original referencefor the statistic. While it is hard to substantiate given the di-versity of metalworking industry, Merchant and other industryexperts cite it as a good rough estimate. Personal communica-tion, M. E. Merchant, Nov. 7, 1983.

‘OTA work environment case studies.

Management sends the specifications for thesize, shape, function, and desired performanceof the product to its design engineering staff,who are responsible for developing the plansfor the product.* In most companies, designengineers make a rough drawing of the prod-uct, and then draftsmen and design detailersare responsible for working out the detailedshapes and specifications.

In some discrete manufacturing firms, de-sign may be undertaken at a distant location,or at a different firm. Automobiles, for exam-ple, are designed at central facilities, and thecomponent subassemblies-e. g., bodies, trans-missions, engines-are produced in plants allover the world.

The design of a product, especially a productof some complexity, involves an intricate setof tradeoffs between marketing considera-tions, materials and manufacturing costs, andthe capabilities and strengths of the company.The number of choices involved in design isimmense. Determining which of many alter-native designs is “best” involves makingchoices among perhaps 100,000 different ma-terials, each with different characteristics ofstrength, cost, and appearance; it also involveschoices between different shapes and arrange-ments of parts which will differ in ease offabrication and assembly (sometimes called‘‘manufacturability’ and in performance.

From the design, the production engineer-ing staff determines the “process plan’ ‘mac-hines, staff, and materials which will be usedto make the product. Production planning, likedesign, involves a set of complex choices. Ina mass production plant that manufacturesonly a few products, such as the auto plant de-scribed above, production engineering is a rel-atively well-structured problem. With highvolumes and fairly reliable expectations aboutthe products to be made, decisions about ap-propriate levels of automation, for example,

——.*In this description, as in the rest of the chapter, titles such

as ‘‘manager, “ “design engineer, ” or “draftsman” indicate theperson who performs these functions. In an actual company theroles may be less distinct, and boundaries between them fre-quently changing.

38 . Computerized Manufacturing Automation: Employment, Education, and the Workplace

Figure 2.—Organizational of a Manufacturing Firm

Personnel

Management Accounting

I - -1

Inventory Production Maintenanceengineering

Materialshandling Fabrication Assembly Finishing Quality control

and assurance

SOURCE Office of Technology Assessment

Ch. 3—Programmable Automation Technologies ● 3 9——— —. —

Figure 3.—Steps in the Manufacturing Process (Simplified)

Productideas and

Marketing/ needs Rough- Design detain BlueprintsDesign

Manufacturing/1management design drafting engineering

Design constraints IProcess plans

Materials Materials MaterialsInventory

Rough

handling Fabrication Finishingpart

Repetitive fabricationI

steps

Finished part

Quality Inspected Finished Inventory

control Assembly orpart product shipping

Subassemblies

SOURCE Office of Technology Assessment

are relatively straightforward. On the otherhand, for a small “batch” manufacturer suchas the Connecticut machine shop referred toabove, production engineering decisions canbe rather chaotic. Such an environment in-volves almost continuous change in the num-ber and types of parts being produced (size,shape, finish, material), the tools and levels ofskill needed to produce them, and unpredict-ables such as machine breakdowns and inven-tory control problems.

The steps in production are immensely var-ied, but most products typically require thefollowing:

1. Materials handling. —Materials arebrought from inventory to processing sta-tions, and from one station to another.Wheeled carts, forklift trucks, or convey-ors are typically used for this purpose.Early in the production process, largeparts are mounted on a pallet or fixture

to hold them in place and facilitate ma-terials handling.

2. Fabrication. —There is a tremendous vari-ety of fabrication processes. Plastic andceramic parts are extruded or molded; lay-ers of composite fiber material are treatedand “laid Up. ” The most common se-quence for three-dimensional (3-D) metalparts is casting or forging, followed bymachining.

Figure 4 illustrates the basic machin-ing processes which are the core of metal-working. The shape and size of the metalpart, as well as the desired finish and precision, determines the machine to be used.Some machine tools, such as lathes, aredesigned for cylindrical parts, e.g., driveshafts or rotors. Others, such as planers,are designed for prismatic parts with ba-sically flat outer surfaces, e.g., engineblocks. Abrasive cutting of metal pro-duces “chips,” the metal shavings re-

40 . Computerized Manufacturing Automation: Employment, Education, and the Workplace— . —

Figure 4.— Fundamental Operations in Metalworking

Turning

(

+Drilling

Work

Tool % 1 / w o r k

Shaping

r.— Arig. M Fig.

Depthof cut

Feed4

Original diameter

diameterB

[A] The four basic maching processes can, between them, theoretically [B] In any of the basic maching processes, speed, feed,produce any contour on a workplace Although the processes are old, they and depth of cut determine productivity The three variables

are the foundation of metalworking and are being made more produc- are shown here for turning on a lathetive and accurate

Sheet metal bending

+

Push

force

Wire drawing

Pull forceWork

Wire

Cold rollingDie

Extrusion (foward)

Final size

Ram Work

Embossing

w

Sheetmetal

Die

Fig. C

Size

Sheet metal

Punch

Fig. D

[C] Forming operations bend.squeeze, or stretch metal,imparting new sizes or shapes.or both

[D] Shearing deforms metalbeyond Its breaking point,thereby separating one por-tion of a metal sheet from an-other.

SOURCE M P Groover Fundamental Operations EEE Spectrum vol 20 No 5 May 1983 Reprinted with permission of /EEE Spectrum IEEE

Ch. 3—Programmable Automation Technologies ● 41——. . —

3.

4.

moved from the part, and these chipsmust be frequently removed from the ma-chine and recycled or disposed.

Simple parts may be machined in a fewminutes; large, complex ones such as shippropellers may take up to a few days. Thecomplexity of these parts is primarily afunction of their geometry–a propeller,for example, has continuously varyingand precise curves. Similarly, the com-plexity of a prismatic part depends on thenumber of edges and required tolerances—i.e., the amount a part or surface canvary from its specified dimensions. Com-plex parts usually require machining onmore than one machine tool. Including allmachining operations, the total time frommetal “blank” to finished part may varyfrom a few minutes to a few weeks. Thepartially completed “workpieces” await-ing further machining, finishing, assem-bly, or testing are known as work-in-proc-ess inventories, and often represent asubstantial investment for the manu-facturer.

Finally, there are several kinds of metalparts which are not machined. These in-clude sheet metal parts, which arestamped and/or bent in sheet-metalpresses, and parts made by “powder met-allurgy, ” a technology for forming metalparts in near-final shape by applying ex-treme pressure and heat to metal powder.Finishing. -Many fabrication processesleave “burrs” on the part which must beremoved by subsequent operations. Insome cases, parts are also washed,painted, polished, or coated.Assembly.–The finished parts are puttogether to produce a final product or, al-ternatively, to produce “subassemblies”which are portions of the final product.In most factories assembly is still primari-ly a manual activity, although this phaseof manufacturing is receiving increasedattention, ranging from design strategiesthat minimize and simplify assemblytasks to automation of the tasks them-selves.

5. Quality assurance and control.–There aremany quality strategies. They can be di-vided roughly into those that take placebefore or during fabrication and assembly(quality assurance or QA) and those thattake place after a product or subassemblyis complete (quality control or QC). Quali-ty has been receiving increasing attentionin industrial literature and discussions, al-though the extent to which companies areactually paying more attention to quali-ty on the factory floor is uncertain. Thereappears to be a movement toward QA asopposed to QC in order to enhance quali-ty and prevent the production of faultyproducts, as opposed to detecting flawsafter production. Strategies for QA rangefrom “quality circles, ” in which a team ofemployees helps address production is-sues which affect quality, to in-processmeasurement of products as they aremanufactured. In the latter, developingproblems in production equipment cansometimes be detected and corrected be-fore the machine makes a bad part. Mostcomplex products are produced with somecombination of QA and QC.

Strategies for attaining the more tradi-tional quality control vary widely accord-ing to the nature and complexity of thepart. The dimensions of mechanical prod-ucts can be measured, either with manualinstruments or with a Coordinate Measur-ing Machine or laser measurement device;or the product can be compared to one ofknown quality or to a master gage. Elec-tronic products can be tested with otherelectronic devices or probes.

This brief outline of the manufacturing proc-ess suggests some of the key problems in man-ufacturing. Underlying each of these problemsare the central concerns for any business,those of minimizing cost and risk. The prob-lems include:

● Information flow. — In any company,small or large, the amount of informationthat must flow between and among de-sign, manufacturing, and management

42 ● computerized Manufacturing Automation: Employment, Education, and the Workplace

staff is staggering. For example, in adesign process involving several teams ofpeople, how does one make sure that alldesign and manufacturing personnel areworking from the most up-to-date set ofplans? How can staff get up-to-date infor-mation on the status of a particular batchof parts, or the performance of a particu-lar machine tool or manufacturing depart-ment? How can the company keep trackof work-in-process and other inventory?Coordination.– Beyond merely obtaininginformation in a timely fashion, the com-pany must use that information to deter-mine how to effectively coordinate itsoperations. One set of such issues in-volves coordination of design and produc-tion efforts. How can one design productswhich can be manufactured most effi-ciently with a given set of tools? How canone minimize the number of parts in orderto facilitate assembly? Another set ofcoordination issues arises on the factoryfloor itself. What is the most efficient wayto allocate machines and personnel? Howdoes one adapt the schedule when condi-tions inevitably change (raw materialsdon’t arrive, production is slower than ex-pected, etc.)?Efficiency. —Given a large set of choicesregarding tools, personnel, and factory or-ganization, a company generally seeks tomake the most products using the fewestresources. This involves concerns such as:How can the company minimize expen-sive work-in-process inventories? Howcan manufacturers maximize the percent-age of time spent making parts, as op-posed to moving them, repairing or set-ting up machines, and planning? How Cm

the use of expensive capital equipment bemaximized? Finally, quality issues with-in the production process can have a largeimpact on efficiency. How can manufac-turers maximize the number of productsmade right the first time, and hence min-imize scrap, rework to correct manufac-turing errors, and testing?

● Flexibility. –Increasingly, issues of flex-ibility and responsiveness in the manufac-turing enterprise are prominent for man-ufacturers, especially for traditional“mass production” plants. Flexibility isdefined here as the range of products andthe range of volumes of a specific productwhich a plant cm economically produce.Increased levels of competition, shorterproduct cycles, and increased demands forcustomized products are some of the rea-sons for an emphasis on flexibility. Thisconcern raises such questions as: How canthe turnaround time for design and man-ufacture of a product be reduced? Howcan the “setup” time for producing a newproduct be reduced? What is the optimumlevel of technology for both economy ofproduction and maximum flexibility?

Programmable automation offers improve-ments in each of these four key areas of man-ufacturing by applying computerized tech-niques to control tools of production, to gatherand manipulate information about the manu-facturing process, and to design and plan thatprocess. Further, the use of PA promises formany manufacturers an increase in their degree of control over the enterprise. Many in-dustrialists argue that the more closely man-ufacturing processes are tied to one another,and the more information is readily availableabout those processes, the less chance thereis for human error or discretion to introduceunknown elements into the operation. Suchcontrol is much harder to realize than it ap-pears in theory. The issue of control will be arecurrent theme in this and subsequent chap-ters of this report.

In summary, programmable automation canhelp make factories “leaner” and more respon-sive, hence reducing both costs and risks inmanufacturing. It is not, however, a panaceafor problems in manufacturing. Each factoryhas different appropriate levels of automation,and there are technical and organizational bar-riers to implementing programmable automa-tion most effectively. PA’s capabilities and

Ch. 3—Programmab/e Automation Technologies ● 4 3. — —

characteristics from a technical standpoint technologies themselves. The organizationalwill be elaborated in the rest of this chapter, and social concerns will be addressed at lengthbeginning with functional descriptions of the in following chapters of the report.

Functional Descriptions

This section briefly describes the operation lyze a design and maximize a product’s per-of each programmable automation technology formance using the computerized representa-and its applications in manufacturing. tion of the product.

Computer-Aided Design (CAD)

In its simpler forms, CAD is an electronicdrawing board for design engineers and drafts-men. Instead of drawing a detailed design withpencil and paper, these individuals work at acomputer terminal, instructing the computerto combine various lines and curves to producea drawing of a part and its specifications. Inits more complex forms, CAD can be used tocommunicate to manufacturing equipment thespecifications and process for making a prod-uct. Finally, CAD is also the core of computer-aided engineering, in which engineers can ana-

Photo credit Cincinnati Milacron Corp

A designer works on a two-dimensional part drawingat a CAD terminal. The “light pen, ” held in his righthand, can be used to point to parts of the drawing and

give commands to the computer

The roots of computer-aided design technol-ogy are primarily in computer science. CADevolved from research carried out in the late1950’s and early 1960’s on interactive comput-er graphics-simply, the use of computerscreens to display and manipulate lines andshapes instead of numbers and text. As S. H.Chasen of Lockheed-Georgia describes the ra-tionale behind this research: “The ability ofthe computer to spill out reams of geometricdata had outpaced our ability to cope with it.’”SKETCHPAD, funded by the Department ofDefense (DOD) and demonstrated at Massa-chusetts Institute of Technology in 1963, wasa milestone in CAD development. Users coulddraw pictures on a screen and manipulatethem with a “light pen ”-a pen-shaped objectwired to the computer which locates points onthe screen. Such early systems were expensiveprototypes and required most of the comput-ing power of the then-largest computers. Asa consequence, most of the early users of CADwere aerospace, automobile, and electronicsmanufacturers.

Several key developments in the 1960’s and1970’s facilitated the maturation of CAD tech-nology. They included the continuing decreasein cost of computing power, especially with thedevelopment of powerful mini- and microcom-puters, which were primarily a result of elec-tronics manufacturers learning to squeezemore and more circuitry into an integrated cir-cuit chip. Another important technological ad-vance was the development of cheaper, more

4S. H. Chasen, “Historical Highlights of Interactive ComputerGraphics, ” Mechanical Engineering, November 1981, pp. 22-41.

44 ● Computerized Manufacturing Automat/on: Employment, Education, and the Workplace

efficient display screens. In addition, comput-er scientists began to develop very powerfulprograming techniques for manipulating com-puterized images.

How CAD Works. —There are variousschemes for input of a design to the computersystem, each with its advantages and disad-vantages. Every CAD system is equippedwith a keyboard, although other devices areoften more useful for entering and manipulat-ing shapes. The operator can point to areas ofthe screen with a light pen or use a graphicstablet, which is an electronically touch-sensi-tive drawing board; a device called a “mouse”can be traced on an adjacent surface to movea pointer around on the screen. If there is al-ready a rough design or model for the prod-uct, the operator can use a “digitizer” to readthe contours of the model into computer mem-ory, and then manipulate a drawing of themodel on the screen. Finally, if the part issimilar to one that has already been designedusing the CAD system, the operator can recallthe old design from computer memory and editthe drawing on the screen.

CAD systems typically have a library ofstored shapes and commands to facilitate theinput of designs. There are four basic functionsperformed by a CAD system which can en-hance the productivity of a designer or drafts-man. First, CAD allows “replication,” the abil-ity to take part of the image and use it in sev-eral other areas of the design when a producthas repetitive features. Second, the systemscan “translate” parts of the image from onelocation on the screen to another. Third is‘‘scaling, “ in which CAD can “zoom in’ on asmall part, or change the size or proportionsof one part of the image in relation to theothers. Finally, “rotation” allows the operatorto see the design from different angles or per-spectives. Using such commands, operatorscan perform sophisticated manipulations ofthe drawing, some of which are difficult or im-possible to achieve with pencil and paper. Re-petitive designs, or designs in which one partof the image is a small modification of a pre-vious drawing, can be done much more quick-ly through CAD. On the other hand, CAD can

be cumbersome, especially for inexperiencedusers. Drawing an unusual shape maybe fairlystraightforward with a pencil, but quite com-plex to accomplish using the basic lines andcurves in the system’s library.

The simplest CAD systems are two-dimen-sional (2-D), like pencil-and-paper drawings.And like sets of those drawings, they can beused to model 3-D objects if several 2-D draw-ings from various perspectives are combined.For some applications, such as electronic cir-cuit design, 2-D drawings are sufficient. Moresophisticated CAD systems have been devel-oped in the past few years which allow theoperator to construct a 3-D image on thescreen, * a capability which is particularly use-ful for complex mechanical products.

Most CAD systems include a few CAD ter-minals connected to a central mainframe orminicomputer, although some recently devel-oped systems use stand-alone microcomput-ers. As the operator produces a drawing, it isstored in computer memory, typically on amagnetic disk. The collection of digitizeddrawings in computer storage becomes a de-sign data base, and this data base is thenreadily accessible to other designers, manag-ers, or manufacturing staff.

CAD operators have several options for out-put of their design. All systems have a plot-ter, which is capable of producing precise andoften multicolor paper copies of the drawing.Some systems can generate copies of the de-sign on microfilm or microfiche for compactstorage. Others are capable of generating pho-tographic output. In most cases, however, thepaper output from CAD is much less impor-tant than it is in a manual design process.More important is the fact that the design isstored on a computer disk; it is this versionwhich is most up-to-date and accessible, and

*In a practical sense, any image on a computer screen is twodimensional, The difference between a “3-D” image as discussedhere and any other 2-D drawing that appears three-dimensional(e.g., a painting, a photograph or any drawing with perspec-tive) is that this image, unlike a paper drawing, can bemanipulated as if it were a real 3-D object. For example, theoperator can instruct the CAD system to rotate the object, andhe/she then sees another face of the object,

— —- Ch. 3—Programmable Automation Technologies ● 4 5— — — —

\

. 4 ”

,

Illustrafion credit Computervision Corp.

The designer has removed a section of this three-dimensional CAD image in order to better visualizepart relationships and assembly information

which will be modified as design changesoccur.

The CAD systems described above are es-sentially draftsmen’s versions of word proc-essors, allowing operators to create and easi-ly modify an electronic version of a drawing.However, more sophisticated CAD systemscan go beyond computer-aided drafting in twoimportant ways.

First, such systems increasingly allow thephysical dimensions of the product, and thesteps necessary to produce it, to be developedvia the computer and communicated electron-ically to computer-aided manufacturing equip-ment. Some of these systems present a graphicsimulation of the machining process on the

screen, and guide the operator step-by-step inplanning the machining process. The CADsystem can then produce a tape which is fedinto the machine tool controller and used toguide the machine tool path. Such connectionsfrom computer-aided design equipment tocomputer-aided manufacturing equipmentshortcut several steps in the conventionalmanufacturing process. They cut down thetime necessary for a manufacturing engineerto interpret design drawings and establishmachining plans; they facilitate process plan-ning by providing a visualization of the ma-chining process; and they reduce the time nec-essary for machinists to interpret processplans and guide the machine tool through theprocess.

46 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace

Second, these more sophisticated CAD sys-tems serve as the core technology for manyforms of computer-aided engineering (CAE).Beyond using computer graphics merely tofacilitate drafting and design changes, CAEtools permit interactive design and analysis.Engineers can, for example, use computergraphic techniques for simulation and anima-tion of products, to visualize the operation ofa product or to obtain an estimate of its per-formance. Other CAE programs can help en-gineers perform finite element analysis-es-sentially, breaking down complex mechanicalobjects into a network of hundreds of simplerelements to determine stresses and deforma-tions. Computerization in general made finiteelement analysis feasible for the engineer’suse, while CAD systems make it significant-ly less cumbersome by assisting the engineerin breaking down the object into “elements.

Many of these analytical functions are de-pendent on 3-D CAD systems which can notonly draw the design but also perform “solidmodeling”—i.e., the machine can calculate anddisplay such solid characteristics as the vol-ume and density of the object. Solid-modelingcapabilities are among the most complex fea-tures of CAD technology, and will be dis-cussed in more detail in later sections of thischapter.

Applications. -At the end of 1983 therewere an estimated 32,000 CAD workstationsin the United States.s

Aerospace and electronics uses of CAD havealways led the state of the art. For example,the Boeing Commercial Airplane Co., whichbegan using CAD in the late 1950’s, employedthe technology extensively in the design of itsnew-generation 757 and 767 aircraft. Boeinguses CAD to design families of similar partssuch as wing ribs and floor beams. CAD allowsdesigners to make full use of similarities be-tween parts so that redesign and redraftingare minimized. Moreover, CAD has greatlysimplified the task of designing airplane in-

Source: Dataquest.

teriors and cargo compartments, which areoften different for each plane. Moving seats,galleys, and lavatories is relatively simple withCAD, and the system is then used to generateinstructions for the machines which later drilland assemble floor panels according to thelayout. Finally, Boeing uses CAD and relatedinteractive computer graphics systems as thebasis for computer-aided engineering applica-tions such as checking mechanism clearancesand simulating flight performance of variousparts and systems.e

Computer-aided engineering has also be-come important in the automobile and aero-space industries, where weight can be a criticalfactor in the design of products. These indus-tries have developed CAE programs which canoptimize a design for minimum material usedwhile maintaining strength.

Applications for the design of integrated cir-cuits are similarly advanced. Very large-scaleintegrated (VLSI) circuits, for example, havebecome so complicated that it is virtually im-

‘W. D. Beeby, (former) Director of Engineering ComputingSystems, Boeing Commercial Airplane Co., “Applications ofComputer-Aided Design on the 767” (Seattle: Boeing, 1983).

Steo 0

Illustration credit: General Motors Corp.

A computer-aided engineering system developed atGeneral Motors Research Laboratories can helpdesigners develop parts which are of minimum mass,yet are capable of performing under the structuralloads. The CAE system tries to make the part thinnerand lighter with each step; shading changes whichappear on the computer screen show simulated stress

levels within the design limits for this part

Ch. 3—Programmable Automation Technologies ● 4 7— — —— —.———— ——

possible for a person to manually keep trackof the circuit paths and make sure the patternsare correct. There is less need here for geomet-rically sophisticated CAD systems (integratedcircuit designs are essentially a few layers oftwo-dimensional lines), and more need for com-puter-aided engineering systems to help thedesigner cope with the intricate arrangementof the circuit pattern. Such CAE programs areused to simulate the performance of a circuitand check it for “faults,” as well as to optimizethe use of space on the chip.7

CAD is also beginning to be used for non-aerospace mechanical design, and in smallerfirms; these developments are being spurredon by the marketing of relatively low-priced‘‘turnkey” systems—complete packages ofsoftware and hardware which, theoretically,are ready to use as soon as they are deliveredand installed. While a standard and reasonablypowerful system based on a minicomputer is

7S. B. Newell, A. J. de Geus, and R. A. Rohrer, “Design Automation for Integrated Circuits, ” Science, Apr. 29, 1983, pp.465-471.

Photo credit Computervision Corp

CAD systems are used frequently in electronicsindustries to design and analyze

complex circuit patterns

typically in the $500,000 range, many smallermicrocomputer-based systems have been in-troduced in the past year for under $100,000,in some cases for as low as $10,000 to $20,000.Very low-cost systems which run on commonmicrocomputers have been introduced, andthese have potential uses in a wide variety offirms which otherwise might not considerCAD (see ch. 7). The cost of custom-developed,specialized systems such as those describedabove for aerospace and electronics applica-tions is harder to gauge but runs well into themillions of dollars.

The potential advantage of CAD for largeas well as small mechanical manufacturingfirms is that it addresses several of the prob-lems in manufacturing referred to at the begin-ning of this chapter. It facilitates use of pre-vious designs, and allows design changes tobe processed more quickly. Because CAD re-duces the time necessary for many designtasks, it can also improve design by allowingdesigners to “try out” a dozen or a hundreddifferent variations, where previously theymight have been limited to building perhapsthree or four prototype models. It also allowsmany drawings to be constructed more quick-ly, especially with an experienced CAD oper-ator. Comparisons of design time with CADrange from 0.5 to 100 times as fast as manualsystems, with 2 to 6 times as fast being typi-cal. * For instance, Prototype and Plastic MoldCorp. in Middletown, Corm., is a small firmthat uses CAD to design short-lived metalmolds for plastic parts. The firm’s presidentreported that designs could be produced withCAD roughly twice as fast as previously. Forexample, they received specifications for aplastic part mold by air express one Saturdaymorning, and planned to return the designdrawings by air express that evening-a featwhich, they reported, would have been impos-sible without CAD.**

Many of these represent comparisons of the timenecessary for a very narrowly defined task, and exclude timenecessary for related tasks on a CAD system such as settingup the machine, manipulating files, or recovering from a ma-chine failure.

**OTA site visit, protype and Plastic Mold Corp., Mid-dletown, Corm., June 3, 1983. One scientist has pointed out thatthe time savings would be even more striking if Prototype and

48 . Computerized Manufacturing Automation: Employment, Education, and the Workplace.

Other applications of CAD, though notdirectly connected to manufacturing, includemapping, architectural drawing and design,graphics for technical publishing, and anima-tion and special effects in cinematography.

Computer-Aided Manufacturing (CAM)Technologies

Computer-aided manufacturing (CAM) is awidely used term in industrial literature, andit has various meanings. Here it is defined sim-ply as those types of programmable automa-tion which are used primarily on the factoryfloor to help produce products. The followingsections provide functional descriptions of fourCAM tools: robots, numerically controlled ma-chine tools, flexible manufacturing systems,and automated materials handling systems.

Robots

Robots are manipulators which can be pro-gramed to move workplaces or tools alongvarious paths. Most dictionary definitions de-scribe robots as “human-like,” but industrialrobots bear little resemblance to a human. *

There is some controversy over the defini-tion of a robot. The Japan Industrial RobotAssociation, for example, construes almostany machine that manipulates objects to bea robot (essentially including the “hard auto-mation” mentioned earlier), while the oft-quoted Robotic Industries Association (RIA)definition** emphasizes that the robot must

Plastic Mold’s staff could have transmitted the design infor-mation by telephone computer links; such activities have begunto be feasible within the last few years.

*In this sense the technical usage of the term “robot” dif-fers from its dictionary definition (and from its roots in litera-ture, in particular Karel Capek’s 1923 novel, R. U.R. (Rossum'sUniversal Robots) A Fantastic Melodrama. (Garden City, N. Y.:Doubleday, Page & Co., 1923). A robot which resembled ahuman would be an “android,” in robotics parlance. Such amachine has not been designed, and there does not appear tobe substantial movement toward human-like robots (except,perhaps for motion pictures and other entertainment purposes).Later sections of this chapter will discuss adding certain an-thropomorphic characteristics and skills to robots.

● *RIA (a trade association of robot manufacturers, consta-ants, and users, formerly the Robot Institute of America)defines a robot as a “reprogrammable multifunctional manip-ulator designed to move material, parts, tools, or specializeddevices, through variable programed motions for the perform-ance of a variety of tasks. ”

be flexible,one task to

Photo credit Cincinnati Milacron Corp

A robot used for welding

or relatively easily changed fromanother. The RIA definition thus

excludes preset part-transfer machines usedfor decades as a part of largebatch and mass-production systems, whose path can be changedonly by mechanically reworking or rearrang-ing the device. Also excluded are “manualmanipulators” or “teleoperators’ ’-devicesdirectly controlled by a human such as thosefor remote handling of radioactive material.

As OTA observed in an earlier report onthis subject,8 industrial robots have a dualtechnological ancestry, emerging from: “l) in-dustrial engineering automation technology,a discipline that stretches historically over acentury; and 2) computer science and artificialintelligence* technology that is only a fewdecades old.” Indeed, there is still a dichotomy

‘Exploratory Workshop on the Social Impacts of Robotics:Summary and Issues (Washington, D. C.: U.S. Congress, Officeof Technology Assessment, OTA-BP-CIT-11, February 1982).

*Artificial intelligence research seeks to develop computersystems that can perform tasks which are ordinarily thoughtto require human intelligence.

Ch. 3—Programmable Automation Technologies ● 4 9———. — —.

Photo credit Cincinnati Milacron Corp

Robot used for loading and unloading a machine tool

among experts regarding the applications and While it is uncertain to what extent artificialresearch directions for robotics. Some empha- intelligence researchers will succeed in devel-size the need for anthropomorphic capabilities oping intelligent machines in the next few dec-in robots such as “intelligence,” vision, and ades, it is certain that robots currently avail-mobility, while others view robots as simply able, and those likely to be available in thea more versatile extension of other manufac- next decade, neither look like humans nor haveturing tools. more than a fraction of the dexterity, flexibili-

50 .Computerized Manufacturing Automation: Employment, Education, and the Workplace

ty, or intelligence of humans. A more accurateterm for these machines might be “program-mable manipulator. ” Nevertheless it is clearthat much of the great popular interest in ro-botics is rooted in the prevailing vision (ornightmare) of intelligent robots with human-like characteristics. Artificial intelligence willbe discussed in more detail in the “TechnicalTrends and Barriers” section of this chapter.

How Robots Work.—There are three mainparts of a typical industrial robot: the control-ler, manipulator, and end-effecter. The con-troller consists of the hardware and software—usually involving a microcomputer or micro-electronic components—which guides themotions of the robot and through which theoperator programs the machine. The manipu-lator consists of a base, usually bolted to thefloor, an actuation mechanism-the electric,hydraulic, or pneumatic apparatus whichmoves the arm-and the arm itself, which canbe configured in various ways to movethrough particular patterns. In the arm, “de-grees of freedom” –basically, the number ofdifferent joints-determine the robot’s dexter-ity, as well as its complexity and cost. Final-ly, the end-effecter, usually not sold as partof the robot, is the gripper, weld gun, or othertool which the robot uses to perform its task.

The structure, size, and complexity of theunit varies depending on the application andthe industrial environment. Robots designedto carry lighter loads tend to be smaller, andoperated electrically; many heavier units movetheir manipulator hydraulically. Some of thesimpler units are pneumatic. Some of theheaviest material-handling robots and thenewer light-assembly robots are arrangedgantry-style, that is, with the manipulatorhanging from an overhead support. A few ro-bots are mobile to a limited degree, e.g., theycan roll along fixed tracks in the floor or intheir gantry supports.

Similarly, there is a great variety of end-effectors, particularly grippers, most of whichare customized for particular applications.Grippers are available to lift several objects

at once, or to grasp a fragile object withoutdamaging it (see fig. 5).

Programing.-There are essentially twomethods of programing a robot. The mostcommonly used method is “teaching by guid-ing. ” The worker either physically guides therobot through its path, or uses switches on acontrol panel to move the arm. The controllerrecords that path as it is “taught.” Just begin-ning to emerge is “offline programing,” wherean operator writes a program in computer lan-guage at a computer terminal, and directs therobot to follow the written instructions.

Each method of programing has advantagesthat depend on the application. Teaching byguiding is the simplest and is actually superiorfor certain operations: spray painting is an ex-ample where it is useful to have the operatorguide the robot arm through its path, becauseof the continuous, curved motions usually nec-essary for even paint coverage. However,teaching by guiding offers minimal ability to“edit” a path—i.e., to modify a portion of thepath without re-recording the entire path. Off-line programing is useful for several reasons:1) production need not be stopped while therobot is being programed; 2) the factory floormay be an inhospitable environment for pro-graming, whereas offline programing can bedone at a computer terminal in an office; 3) ascomputer-aided manufacturing technologiesbecome more advanced and integrated, theywill increasingly be able to automatically gen-erate robot programs from design and manu-facturing data bases; and 4) an offline, writ-ten program can better accommodate morecomplex tasks, especially those in which“branching” is involved (e.g., “if the part isnot present, then wait for the next cycle”).These branching decisions require some kindof mechanism by which the robot can senseits external environment. However, the vastmajority of robotic devices are unable to sensetheir environment, although they may have in-ternal sensors to provide feedback to their con-troller on the position of the arm joints.

— —. . .

Ch. 3—Programmable Automation Technologies ● 5 1. —

Figure 5.—Sample Robot Grippers

uInternal, 3 fingersFor small diameters Fitted to the diameter Fitted to the length

For large objectsInternal For cast parts Vacuum, double

u

Vacuum, curved surface Vacuum, several parts Vacuum pad, several parts Vacuum, record player

‘5.’”

Balloon lifter, bottlesVacuum corrugated surface

SOURCE Tech Tran Corp Industrial

Magnet Iifter Magnet lifter

Robots A Summary and Forecast 1983

52 Computerized Manufacturing Automation: Employment, Education, and the Workplace— — — . — —

Sensors.–Devices for sensing the externalenvironment, while often used in conjunctionwith robots, are a growing technology in them-selves. The simplest sensors answer the ques-tion, “Is something there or not?” For examp-le, a light detector mounted beside a convey-or belt can signal when a part has arrived be-cause the part breaks a light beam. Somewhatmore complex are proximity sensors which, bybouncing sound off objects, can estimate howfar away they are. The technology for thesedevices is fairly well-established. But the mostpowerful sensors are those which can interpretvisual or tactile information; these have justbegun to become practical.

Ideally, vision sensors could allow a robotsystem to respond to changes in its environ-ment, and inspect products, as well as or bet-ter than a human could. However, using com-puters to process images from a video camerahas proven to be an extraordinarily difficultprograming task. Routine variations in light-ing, the complexity of the everyday environ-ment, common variations in shape or texture,and the difference between a 2-D camera im-age and a 3-D world all complicate the taskof computer processing of a video image.

It is also important that robots be viewedas part of the overaIl changes taking placein manufacturing concepts with the increas-ing diffusion of automated manufacturingequipment, including computer-aided manu-facturing and computer-aided design sys-tems. The impact of new production con-cepts, equipment and systems on productioncontrol and machine utilisation, inventorycontrol and management efficiency will to-gether have a much greater productivity im-pact than the industrial robot alone.

As noted earlier in this chapter, interna-tional comparisons of robot “populations” arealso plagued by inconsistencies in the def-inition of a robot, particularly between theUnited States and Japan. Regardless of thedefinition of robot used, Japan leads the worldin number of robots in use. The reasons for Ja-pan’s emphasis on robot technology includea historical shortage of labor, and a tendencyto devote more engineering expertise to man-ufacturing processes than does the UnitedStates. In addition, the United States faced

Table 6.—Operating Robot Installations, End of 1982

Country Number Percent of total

Other kinds of sensing devices, from prox-imity sensors to touch and force sensors, havereceived much less attention than machine vi-sion, but they also could play an importantrole in the factory environment, particularlyfor assembly applications. Sensors will be dis-cussed in greater detail later in the chapter.

Applications. –Table 6 displays some of themost recent robot use estimates. Figure 6 es-timates the robot sales and total use in theUnited States for the next decade. Such sta-tistics should be interpreted with caution,however. In particular, the number of robotsin use is a highly imperfect measure of the lev-el of automation and modernization in an in-dustry or country. Process changes in manu-facturing which increase productivity may ormay not include robots. As one report on in-ternational use of robots observes:9

9‘OECD, “Robots: The Users and the Makers, ” The OECDObserver, July 1983.

Japan . . . . . . . . . . . . . . .......31,900 66United States. . . . . . . . . . . . . . 6,301 13West Germany. ... , ... , ... , 4,300 9Sweden . . . . . . . . . . . . . . . . . . . . 1,450 3Italy . . . . . . . . . . . . . . . . . . . . . . . 1,100 2France . . . . . . . . . . . . . . . . . . . 993 2United Kingdom . . . . . . . . . . . . 977 2Belgium . . . . . . . . . . . . . . . . . . . 305 tPoland . . . . . . . . . . . . . . . . . . . . 285 tCanada . . . . . . . . . . . . . . . . . . . . 273 tCzechoslovakia . . . . . . . . . . . . . 154 tFinland , . . . . . . . . . . . . . . . . . . 98 tSwitzerland . . . . . . . . . . . . . . . . 73 tNetherlands . . . . . . . . . . . . . . . . 71 tDenmark . . . . . . . . . . . . . . . . . . . 63 tAustria . . . . . . . . . . . . . . . . . . . 50 tSingapore . . . . . . . . . . . . . . . . . . 25 tKorea . . . . . . . . . . . . . . . . . . . . . 10 t

Total . . . . . . . . . . . . . . . . . . . . 48,428Less than 1 percent.

Note: This table does not include 9,000 “variable sequenced manipulators”which are included i n the RIA’s estimate for France. Statistics on robots differbecause of differing definitions of a robot, because of different methodologiesfor collecting data, and because “operating robot installations” (as used in thistable) may differ from “robot population, ” which includes some robots in labora-tories and others not yet in use A December 1983 study by the U S. InternationalTrade Commission, for example (“Competitive Position of U S. Producers ofRobotics (in Domestic and World Markets”) gives slightly different figures for robotpopulation in the United States and West Germany (7 232 and 3,500, respectively)

SOURCE Robot Instutitute of America, Worldwide Robotics Survey and D/rectory,1983

—

Figure 6.—ActuaI15(

12[

10(

7:

NOTE”

5(

2!

3 .100

Ch. 3—Programmable Automation Technologies ● 5 3— — — .

and Projected U.S. Annual Robot Sales and Installed Base Through 1992

133,800

Installedbase

96.100

69,200

50,000

36,300

26,500

19,500

14,400 13,70010,800

8200 9,8006200

4500

l , 4 0 0 ~

19,200

1990

26,900

1991

A n n u a l

u n i t

s a l e s

37,700

19921981 1982 1983 1984 1985 1986 1987 1988 1989Year

The projections above are highly speculative. Robot sales have not grown nearly as fast as most Industry observers expected, and one industry analystsuggests that the above figures may be as much as 30 to 50 percent too high (E Lustgarten, Vice President, Paine, Webber, Mitchell, Hutchins, Incpersonal communication, Feb 7, 1984) On the other hand, robot vendors and the Robotic Industries Association still believe that a tremendous upsurgeI n robot sales Is forthcoming, and the projections above may even be too low (L Lachowicz, Robotic Industries Association, personal communicationFeb 7, 1984) See ch 7 for further discussion of the robot industry and its prospects

S O U R C E Tech Traing A Summary and forecast 1983

54 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace

labor surpluses throughout the 1970’s, whichtended to induce manufacturers to use laborinstead of equipment in production. Chapter9 will discuss international comparisons inmore detail.

Sophistication in reprogramiability, as wellas size and degrees of freedom, are some of thekey cost factors for an industrial robot. A sim-ple “pick-and-place” machine with 2 or 3 de-grees of freedom costs roughly $5,000 to$30,000, while more complex programmablemodels, often equipped with microcomputers,cost approximately $25,000 to $90,000 anduplo

Table 7 lists some of the potential applica-tions for industrial robots. Many of the firstapplications of robots have been for particular-ly unpleasant or dangerous tasks. One of theearliest uses, for example, was for loading andunloading die-casting machines, a hazardousand unpleasant task because of the extremeheat. The best-known uses, however, havebeen in spray painting and spot welding in theauto and related industries. In these applica-tions, robots have proven to be useful for per-forming particularly hazardous and monoto-nous jobs while offering enough flexibility tobe easily adapted to changes in car models orbody styles.

There are a number of motivations behindthe use of robots on such unpleasant jobs. Im-provement of job conditions (and, consequent-ly, worker morale) is one of them, though itmay not be the primary one. Such jobs oftenhave high worker turnover and inconsistentproduct quality because of their unpleasant-ness. Also, compliance with the occupationalsafety and health regulations that protect peo-ple performing these tasks adds to productioncosts. In addition, tasks like spray paintingand spot welding are often relatively easy toautomate because the paths the robot is to fol-low are predictable, and the tasks are repeti-tive and require little sensing capability.

10 E. Lustgarten, Vice President, paine, Webber, Mitchell!Hutchins, Inc., personal communication.

Table 7.—Examples of Current Robot Applications

Material Handling●

●

●

●

●

●

●

●

●

●

●

Depalletizing wheel spindles into conveyorsTransporting explosive devicesPackaging toaster ovensStacking engine partsTransfer of auto parts from machine to overheadconveyorTransfer of turbine parts from one conveyor to anotherLoading transmission cases from roller convveyor tomonorailTransfer of finished auto engines from assembly to hottestProcessing of thermometersBottle loadingTransfer of glass from rack to cutting line

Machine loading/unloading:●

●

●

●

●

●

●

●

●

●

Loading auto parts for grindingLoading auto components into test machinesLoading gears onto CNC lathesOrienting/loading transmission parts onto transfermachinesLoading hot form pressesLoading transmission ring gears onto vertical lathesLoading of electron beam welderLoading cylinder heads onto transfer machinesLoading a punch pressLoading die cast machine

Spray painting:● Painting of aircraft parts on automated Iine● Painting of truck bedŽ Painting of underside of agricultural equipment. Application of prime coat to truck cabs● Application of thermal material to rockets● Painting of appliance components

Welding:• Spot welding of auto bodies● Welding front-end loader bucketsŽ Arc welding hinge assemblies on agricultural

equipment● Braze alloying of aircraft seams. Arc welding of tractor front weight supports. Arc welding of auto axles

Machining:● Drilling alum inure panels on aircraft● Metal flash removal from casings● Sanding missile wings

Assembly:● Assembly of aircraft parts (used with auto-rivet

equipment)● Riveting small assemblies● Drilling and fastening metal panels● Assembling appliance switches● Inserting and fastening screws

Other:. Application of two-part urethane gasket to auto part• Application of adhesive● Induction hardening. Inspecting dimensions on parts● Inspection of hole diameter and walI thicknessSOURCE” Tech Tran Corp , Industrial Robots: A Summary and Forecast, 1983

Ch. 3—Programmable Automation Technologies ● 5 5.——

While spot welding, spray painting, andloading/unloading applications have been theprimary uses for robots, increasing sophistica-tion in programmability and in sensing is en-abling applications such as arc welding andassembly.

As an example of such an application, awelder at Emhart Corp. ’s United Shoe Man-ufacturing plant in Beverly, Mass., uses arobot to arc-weld frames for shoemaking ma-chinery* (see photo). He welds several dozenidentical frame units at a time; each frame unitrequires perhaps a dozen 2-inch welds to at-tach reinforcing bars to a steel sheet. The weld-er clamps the first sheet and reinforcing bars-———————

*OTA site visit, Emhart Corp., United Shoe ManufacturingPlant, Beverly, Mass., June 28, 1983.

onto a table. Using directional buttons on a“teach pendant”— a portable panel attachedto the robot controller-he directs the robotto the spot where it is to begin the first weld.He pushes a button to record that location.Still using the teach pendant, he moves the ro-bot to the end of the weld and records that lo-cation. Then he presses a button which in-structs the machine to “weld a straight linefrom the first point to the second. ” After re-peating this process for each of the dozenwelds, he gives the command for the robot tobegin welding, and the robot follows the pathit has been “led through ’’-this time with itswelding gun on. For each subsequent identicalframe unit, all that is required is to clampdown the parts in the same location as theoriginal set on which the machine was

Photo credit Emhart Corp

Welder Pete Bolger at Emhart’s United Shoe Manufacturing Plant uses a “teach pendant” to program a robot to weldparts of a metal frame, below left. After the robot is taught the correct steps, it can repeat those steps with its welding

gun on, while the operator can set up another frame on an adjacent table or perform other duties

2 7 - 4 5 2 0 - 8 4 - 5 : 0 L 3

56 ● computerized Manufacturing Automation: Employment, Education, and the Worldplace .

“taught,” signal the machine to begin, andthen inspect the welds after the machine com-pletes its program. The robot controller canstore several programs, so that the operatorcan use the robot to weld different types offrames in any order he chooses, as long as hesets up the steel plates and reinforcing barsin the appropriate positions.

Note that this application of a robot for arcwelding does not use sensors, even thoughthere has been extensive work done on devel-oping vision sensors that allow the robot to“see” the seam formed by the two pieces ofmetal, and to follow it automatically. Forthe fairly simple, straight-line applications atEmhart, sensors are not necessary. However,if the frame units were out of position by ahalf-inch, the welding robot would put a use-less blob of metal where it expected the jointto be. If a clamp was in the way of a pro-gramed weld, the robot would attempt to weldthrough the clamp, damaging the clamp anditself in the process.

The advantages of robots depend on wheth-er one is comparing them to hard automationdevices or to human workers. Clearly, the flex-ibility and programiability of robots is prom-inent in the first case, while in comparisonwith humans the advantages are likely to bethe robot’s greater consistency, endurance,and ability to tolerate hostile environments.

The disadvantages of robots also depend onwhether they are compared with other auto-mation or humans. In the former case, roboticdevices are sometimes more expensive than ahard automation device which is not program-mable, and they are not as fast—a typical robotmoves about as fast as a human, while dedi-cated automatic part-transfer devices canoperate at considerably greater speed. Theclear advantage of human workers over ro-bots, on the other hand, is in situations whereextensive sensing, judgment, or intelligenceis required, and/or where situations change sofrequently that the expense of programing arobot is uneconomical. For these reasons it isoften suggested that humans, robots, and hardautomation devices are best suited for low,

medium, and high production volumes, respec-tively, although there are many exceptions tothis-e. g., automotive spot welding. Each sit-uation must be evaluated individually.

The design of automated production proc-esses involves determining which tasks aremost suitable for a machine, and which aremost suitable for a human. Several technolo-gy experts have argued that some manufac-turers’ visions of robots as replacements forhuman workers will prevent the best alloca-tion of tasks between human and machine.One researcher argues:

A robot is a machine. It should be de-signed, controlled, and operated as a ma-chine. Any attempt to emulate human behav-ior with a robot is a misdirection. Take, forexample, the task of turning a bolt. A humanturns down a bolt in roughly half-revolutionincrements. At today’state-of-the-art, mostrobots are constrained to perform the taskin the same way. But robots need not be con-strained the way humans are. The most distalaxis of a robot should be capable of continu-ous rotation. The primary advantage that ro-bots will have in the manufacturing marketof the future will be based not on their abili-ty to mimic humans, but on their abilities toperform tasks in ways which humans can-not.11

General-purpose robots are already evolvingtoward special-purpose programmable devicesfor a particular task (e.g., assembly machines,painting machines), and this evolution maycontinue so that few robots in the future looklike the general-purpose “arm” of today.

Though they will not be covered in detailhere, robotics technology has a wide range ofnonmanufacturing uses including handling ofradioactive material, mining, undersea explor-ation, and aids for the handicapped.12

1lW. P. Seering, “Directions in Robot Design, ” Transactionsof the ASME, March 1983, pp. 12-13.

‘zSee, for example, T. N. Sofyanos and T. B. Sheridan, ArIAssessment of Undersea Tekoperators, Sea Grant College program, Massachusetts Institute of Technology, June, 1980; A.Seireg and J. Grundman, “Design of a Multitask ExoskeletalWalking Device,” Biomecham”cs of Meol”cal Devices, D. N.Ghista (cd.) (New York: Marcel Dekker Inc., 1981), pp. 569-639.

Ch. 3—Programmab/e Automation Technologies ● 5 7

Numerically Controlled Machine Tools

Numerically controlled, or NC, machinetools are devices which cut a piece of metal ac-cording to programed instructions about thedesired dimensions of a part and the steps forthe machining process. They consist of a ma-chine tool, specially equipped with motors toguide the cutting process, and a controllerwhich receives numerical control commands.

The U.S. Air Force developed NC technol-ogy in the 1940’s and 1950’s, in large part tohelp produce complex parts for aircraft which

were difficult to make reliably and economical-ly with a manually guided machine tool.

How They Work. —Machine tools for cuttingand forming metal are the heart of the metal-working industry. Using a conventional, man-ual machine tool, a machinist guides the shap-ing of a metal part by hand. He or she moveseither the workpiece or the head of the cuttingtool to produce the desired shape of the part.The machinist controls the speed of the cut,the flow of coolant, and all other relevant as-pects of the machining process.

Photo credit Cincinnati Milacron Corp

An operator supervises a large, computerized numerically controlled (CNC) machine tool. The minicomputer which controlsthe tool is at right; at left additional cutting tools are loaded into an automatic tool changing device

58 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace—

In ordinary NC machines, programs arewritten at a terminal which, in turn, punchesholes in a paper or mylar plastic tape. The tapeis then fed into the NC controller. Each set ofholes represents a command, which is trans-mitted to the motors guiding the machine toolby relays and other electromechanicalswitches. Although these machines are notcomputerized, they are programmable in thesense that the machine can easily be set tomaking a different part by feeding it a differ-ent punched tape; and they are automated inthat the machine moves its cutting head, ad-justs its coolant, and so forth, without directhuman intervention. However, most of thesemachines still require a human operator,though in some cases there is one operator fortwo or more NC machine tools. The operatorsupervises several critical aspects of the ma-chine’s operation:

1. he or she has override control to modifythe programed speeds (rate of motion ofthe cutting tool) and feeds (rate of cut) (seefig. 4). These rates will vary depending onthe batch of metal used and the conditionof the cutting tool;

2. he or she watches the quality and dimen-sions of the cut, and listens to the tool,replacing worn tools (ideally) before theyfail; and

3. he or she monitors the process to avoidaccidents or damage-e. g., a tool cuttinginto a misplaced clamp, or a blocked cool-ant line.

Typically, NC programs are written in a lan-guage called APT (Automatically ProgramedTools), which was developed during the initialAir Force research on NC (see fig. 7 for a sam-ple of an APT program). A number of modifiedversions of APT have been released in the lastdecade, and some of these are easier to usethan the original. But the essential conceptand structure of the numerical codes has re-mained the same. In large part because of themomentum it gained from its initial DOD sup-port, APT has become a de facto standard forNC machine tools.

Figure 7.—SampIe APT Computer Program

2.5”4

! 1 .0”

The path of the center of the cutter is shown as t movesabout the perimeter of the part.

(1)(2)(3)(4)(5)(6) S P .

(7) pl =(8) L1 =(9) cl =(lo) P2 =(11) L2 =(12) L3 =(13)(14)(15)(16)(17)(18)(19)(20)

PARTNO FLAT PLATE NO 12345678MACHIN/MODEL PTXCLPRNTCUTTER/.25FE DRAT/1OPOINT/ -.5, – .5,1POINT/0,0,0LINE/Pl , ATANGL,OCIRCLE/2,.5,.5PO IN T/0,1 ,0LlNE/P2, PAR LEL, L1LlNE/P2, PERPTO, L1FROM/SPGO/TO, L1GO RGl”/Ll ,TANTO,C1GOFWD/Cl ,TANTO, L2GOFWD/L2, PAST, L3Go LFT/L3, PAST, L1GOTO/SPFIN I

The APT computer program, above, directs a machine tool to cutaround the perimeter of a flat metal part with a semicircular end (seetop diagram) In the program, the first line Identifies the part, and line(2) calls out the postprocessor for the machine/control combinationthat IS to machine the part The postprocessor IS that part of thecomputer software program that tailors the tape Instruct Ions for theparticular machine/control combination Line (3) notes that the com-puter IS to print out the coordinates of all the straight-llne moves ofthe cutter Line (4) notes that the cutter iS to have a diameter of O 25Inches. Line (5) describes the feed rate in inches per minute Lines(6) through (12) describe the geometry of the part Lines (13) through(19) are motion statements and describe the path of the cutter Line(20) ends the part program

SOURCE: J. J. Childs, Princlples of Numerical Control (New York: Industrial Press,1982, pp. 134-135.

Since 1975, machine tool manufacturershave begun to use microprocessors in the con-troller, and some NC machines come equippedwith a dedicated minicomputer. Those calledcomputerized numerically controlled (CNC)

Ch. 3—Programmable Automation Technologies ● 5 9

tend to be equipped with a screen and key-board for writing or editing NC programs atthe machine. Closely related to CNC is directnumerical control (DNC), in which a largermini- or mainframe computer is used to pro-gram and run more than one NC tool simul-taneously. As the price of small computers hasdeclined over the past decade, DNC hasevolved both in meaning and concept intodistributed numerical control, in which eachmachine tool has a microcomputer of its own,and the systems are linked to a central con-trolling computer. One of the advantages ofsuch distributed control is that the machinescan often continue working for some time evenif the central computer “goes down. ”

In all types of NC machine tools the machin-ing processes are essentially the same-the dif-ference is in the sophistication and location ofthe controller. CNC controllers allow the oper-ator to edit the program at the machine, ratherthan sending a tape back to a programer ina computer room for changes. In addition, byavoiding the use of paper or mylar tape, CNCand DNC machines are substantially more reli-able than ordinary NC machines. The tapepunchers and readers and the tape itself havebeen notorious trouble spots. CNC and DNCmachines, through their computer screens,may also offer the operator more complete in-formation about the status of the machiningprocess.

Some NC tools are equipped with a featurecalled “adaptive control, ” which tries to au-tomatically optimize the rates of cut to pro-duce the part as fast as possible, while avoid-ing tool failure. As yet, there has been limitedsuccess with these devices.

Applications. –The diffusion of NC tech-nology into metalworking industry proceededvery slowly in the 1950’s and 1960’s, thoughit has accelerated somewhat over the past 10years. Figure 8 and table 8 detail the U.S. pop-ulation of machine tools. Numerically con-trolled machine tools represent only 4.7 per-cent of the total population,12 although this fig-

“’’The 13th American Machinist Inventory of MetalworkingEquipment 1983, ” American Machinist, November 1983, pp.113-144.

Figure 8.—Total Number of Numerically ControlledMachine Tools in U.S. Metalworking

1958 1963 1968 1973 1977a 1983

a 12th Inventory data collected over 3 years 1976 to 1978

SOURCE 13th American Machinist Inventory American Machinist November 1983

Table 8.—Estimated Total Machine Toolsin the United States

Total Metal- Metal-units cutting forming

Metalworking . . . . . . . . . 2,192,754 1,702,833 489,921Other industries . . . . . . . 380,000 275,000 105,000Training , . . . . . . . . . . . . . 74,000 70,000 4,000In storage and surplus . 250,000 200,000 50,000

Total . . . . . . . . . . . . . . . 2,896,754 2,247,833 648,921SOURCE: 13th Annual American Machinist Inventory and estimates, American

Machinist, November 1983

ure may be somewhat misleading: the newer,NC machine tools tend to be used more thanthe older equipment, and firms often keep oldequipment even when they buy new machines.Some industry experts have estimated that asmany as half of the parts made in machineshops are made using NC equipment. Never-theless, the applications still tend to be con-centrated in large firms and in smaller subcon-tractors in the aerospace and defense indus-tries.

Two examples from OTA’S case studies il-lustrate a range of uses for NC machine tools.A Connecticut machine shop with 48 employ-ees on the shop floor began using numericalcontrol technology around 1966, and now uses23 NC machines to produce contracted partsfor the electronics and aircraft industries. Bycontrast, one of the NC machine shops at alarge commercial aerospace manufacturer oc-

60 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace-.. .— — —

cupies 471,000 square feet—about the size of18 football fields–and includes 110 NC andCNC machine tools, as well as 230 conven-tional machine tools.

The U.S. machine tool population is signifi-cantly older than that of most other countries(see table 9), and this situation, suggesting rel-atively low levels of capital investment, hasbeen a source of concern for many in industryand government. In 1983, for the first time inseveral decades, the percentage of metalcut-ting tools less than 10 years old increased by3 percent, although the percentage of metal-forming tools less than 10 years old remainsat an all-time low of 27 percent.15

DOD has encouraged diffusion of NC tech-nology, which has moved beyond the aero-space industry—although not nearly as fastas most observers expected. There are severalreasons for the relatively slow diffusion of NCtechnology. They include high capital cost foran NC machine (perhaps $80,000 to $150,000and up, as opposed to $10,000 to $30,000 fora conventional machine tool).16 In addition, thesuccessful application of NC machine tools re-quires technical expertise that is in short sup-ply in many machine shops. Training is a prob-lem, as some users report requiring as muchas 2 years “to get an NC programer up tospeed. “17 Small machine shops typically do nothave the resources or expertise to train staffto use or maintain computerized equipment.Finally, “APT proved to be too complicatedfor most users outside the aerospace indus-try. . . . Most machine jobs could be specifiedin a considerably less complex world. ”18

However, intricate shapes such as those nowfound in the aerospace industry are nearly im-possible for even the most experienced ma-

140TA work environment case studies.“’’The 13th American A4achiiu”st Inventory of Metalworking

Equipment 1983, ” American Mac)u”m”st, November 1983.1eE. Lustgarten, Vice President, Paine, Webber, Mitchell,

Hutchins, Inc., personal communication.17A. M. Greene, “Is It Time for a New Approach to NC Pro-

gramming?” Iron Age, Sept. 24, 1982, p. 83. The need for sub-stantial training applies not only to NC machine tools but tovirtually all PA devices.

“Industry and Trade Strategies, unpublished paper preparedfor OTA, April 1983, p. 28.

chinist using conventional machine tools. WithNC, the parts can be more consistent becausethe same NC program is used to make the parteach time it is produced. A manually guidedmachine tool is more likely to produce partswith slight variations, because the machinistis likely to use a slightly different procedureeach time he or she makes a part. This maynot be a problem for one-of-a-kind or customproduction, but can cause headaches in batchproduction. The advantages in consistencydue to NC are seen by many manufacturersas an increase in their control over the machin-ing process.

NC machines tend to have a higher “through-put” than conventional machine tools, andhence are more productive. They are operating(i.e., cutting metal) more of the time than aconventional machine tool because all thesteps are established before the machiningbegins and are followed methodically by themachine’s controller. Further, on a complexpart that takes more than one shift of machin-ing on a conventional machine tool, it is verydifficult for a new machinist to take overwhere the first left off. The part may remainclamped to the machine and the part and ma-chine tool lie idle until the original machinistreturns. On NC machines, operators can sub-stitute for each other relatively easily, allow-ing the machining to continue uninterrupted.

As discussed previously, the capability ofguiding machine tools with numeric codesopens up possibilities for streamlining thesteps between design and production. The gometric data developed in drawing the producton a CAD system can be used to generate theNC program for manufacturing the product.

Flexible Manufacturing Systems

A flexible manufacturing system (FMS) isa production unit capable of producing a rangeof discrete products with a minimum of man-ual intervention. It consists of productionequipment workstations (machine tools orother equipment for fabrication, assembly, ortreatment) linked by a materials-handling sys-tem to move parts from one workstation to

Table 9.—Age of Machine Tools in Seven Industrial Nations- —

Metalcutting machines.

Metalforming machines

Year Units —O-2 yr 6-4 yr 0-9 yr >15 yr >20 yr Units o-2 yr O-4 yr O-9 yr >15 yr >20 yr

United States. . . . . . . . . . . . . . . 83 1,703,000 — 14 % 34 ”/0 — 320/o 490,000 — 90/0 180/0 — 37 “/0Canada . . . . . . . . . . . . . . . . . . . . 78 149,400 — — 41 — 37 61,400 — – 23 – 26Federal Republic of Germany. 80 985,000 – 15 34 480/o — 265,000 — 15 34 48 ‘/oFrance ., . . . . . . . . . . . . . . . . . . 80 584,000 – 16 35 – 32 177,000 — 16 35 – 32Italy . . . . . . . . . . . . . . . . . . . . 75 408,300 — — 41C — 29’ 133,000 — — 29 C — 25’Japan . . . . . . . . . . . . . . . . . . . . 81 707,000 150/0 — 35b 37 d — 211,000 180/0 — 41 b 31 d —United Kingdom . . . . . . . . . . . . 82 627,900 – 41 a — — 27 146,800 — 48 a — — 28

a 0-5 years old” bO-7 years old C0.8 years old. ’13 years old and UP. ’18 Years old and UP

SOURCE 13th American Machinist Inventory, Arnerlcan ?dachln(sf, November 1983 (Note Amerfcan Machlnlst used a variety of foreign sources for this table)

I

●

)

62 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace—

another, and it operates as an integrated sys-tem under full programmable control.19

An FMS is often designed to produce a fami-ly of related parts, usually in relatively smallbatches–in many cases less than 100, andeven as low as one. Most systems appropriate-ly considered to be an FMS include at leastfour workstations, and some have up to 32.Smaller systems of two or three machine toolsserved by a robot, which are sometimes calledflexible manufacturing systems, are more ap-propriately termed “machining cells. ”

How an FYWS Works.– Using NC programsand (often) computer-aided process planning,workers develop the process plan (i.e., the se-quence of production steps) for each part thatthe FMS produces. Then, based on inventory,orders, and computer simulations of how theFMS could run most effectively, the FMSmanagers establish a schedule for the partsthat the FMS will produce on a given day.Next, operators feed the material for each part

“M. E. Merchant, personal communication, Oct. 12, 1983.Adapted from a definition developed by the International In-stitution for Production Engineering Research.

into the system, typically by clamping a blockof metal into a special carrier that serves bothas a fixture to hold the part in place while itis being machined, and as a pallet for trans-porting the workpiece. Once loaded, the FMSessentially takes over. Robots, conveyors, orother automated materials handling devicestransport the workpiece from workstation toworkstation, according to the process plan. Ifa tool is not working, many FMSS can reroutethe part to other tools that can substitute.

Machine tools are not the only workstationsin an FMS; other possible stations includewashing or heat-treating machines, and auto-matic inspection devices. While most currentFMSS consist of groups of machine tools,other systems anticipated or in operation in-volve machines for grinding, sheet metal work-ing, plastics handling, and assembly.

The amount of flexibility necessary to de-serve the label “flexible” is arguable. SomeFMSS can produce only three or four parts ofvery similar size and shape—e.g., three or fourengine blocks for different configurations ofengines. One FMS expert argues, however,

IIlusfration credit: Cincinnati Milacron Corp

Schematic diagram of an FMS for producing aircraft parts. The lines indicate paths of automatic devices which bringworkplaces to the machines

Ch, 3—Programmable Automation Technologies ● 6 3—

Photo credit: Cincinati Corp

An FMS system for tank parts

64 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace

that in the current state of the technology, asystem that cannot produce at least 20 to 25different parts is not flexible. Indeed, some arebeing designed to manufacture up to 500parts. 20

The essential features that constitute aworkable “part family” for an FMS are:

●

●

●

●

A common shape. —In particular, prisma-tic (primarily flat surfaces) and rotationalparts cannot be produced by the same setof machines.Size.–An FMS will be designed to pro-duce parts of a certain maximum size,e.g., a 36-inch cube. Parts larger or muchsmaller than that size cannot be handled.Material. -Titanium and common steelparts cannot be effectively mixed, nor canmetal and plastic.Tolerance. –The level of precisionnecessary for the set of parts must be ina common range.

Applications. –For a manufacturer with anappropriate part family and volume to use anFMS, the technology offers substantial advan-tages over stand-alone machine tools. In anideal FMS arrangement, the company’s expen-sive machine tools are working at near fullcapacity. Turnaround time for manufacture ofa part is reduced dramatically because partsmove from one workstation to another quick-ly and systematically, and computer simula-tions of the FMS help determine optimal rout-ing paths. Most systems have some redundan-cy in processing capabilities and thus can au-tomatically reroute parts around a machinetool that is down. Because of these time sav-ings, work-in-process inventory can be dras-tically reduced. The company can also de-crease its inventory of finished parts, since itcan rely on the FMS to produce needed partson demand.

Finally, FMS can reduce the “economicorder quantity” for a given part-the batchsize necessary to justify setup costs. When apart has been produced once on an FMS, setup

20B JOhOSki, Mm~r, Manufacturing Systems Division, Cin-cinnati Milacron Corp., interview, Aug. 16, 1983.

costs for later batches are minimal becauseprocess plans are already established andstored in memory, and materials handling isautomatic. In the ultimate vision of an FMS,the machine could produce a one-part batchalmost as cheaply as it could produce 1,000,in cost per unit. In practice there are unavoid-able setup costs for apart and a one+part batchis uneconomical. Nevertheless, the FMS’s cap-ability to lower the economic order quantityis particularly useful in an economy in whichmanufacturers perceive an increased demandfor product customization and smaller batchsizes. *

A Midwestern agricultural equipment man-ufacturer, for example, uses an FMS to ma-chine transmission case and clutch housingsfor a family of tractors (see photo). They hadconsidered “hard automation ”-a transferline–to manufacture the parts, but expected

*ln defense production, several examples of the very high Costof spare parts have come to light. In part because the set upcost for producing parts is so high, the Pentagon’s contractorsmay charge thousands of dollars for producing one or two smallparts. The traditional solution to this problem is to make spareparts when the original equipment is bought, and keep thespares on hand. However, studies indicate that more than 95percent of those spare parts are never used. An FMS couldsubstantially reduce the cost for making a small number of partsbecause once a part has been made on the system and the tool-ing and production routing already established, setup costs arerelatively low.

Photo credit: Deere & Co,

A worker obtains status information from a computerterminal at one of the workstations of an FMS for

producing agricultural equipment

—. — — —Ch. 3—Programmable Automation Technologies ● 6 5

—

a new generation of transmissions within 5years, which would render the transfer line ob-solete. They chose an FMS instead because itcould be more easily adapted to other prod-ucts. In the system, a supervisory computercontrols 12 computerized machining centersand a system of chain-driven carts which shut-tle the fixtured parts to the appropriate ma-chines. The supervisory computer automati-cally routes parts to those machines with theshortest queue of workplaces waiting, and canreroute parts to avoid a disabled machine tool.About a dozen employees operate and main-tain the system during the day shift, and thereare even fewer people on the other two shifts.The system is designed to produce nine parttypes in almost any sequence desired. (Thus,it is rather inflexible according to the currentstate of the art.) It was, in fact, one of theearliest FMSS of substantial size to be de-signed. It was ordered in 1978, but not fullyimplemented until 1981.21

Another example of FMS application is asystem operating at Messerschmitt-Bolkow-Blohm’s plant in Augsburg, West Germany,to manufacture the center section of Tornadofighter planes at a rate of about 10 per month.The system includes 28 NC machine tools, au-tomatic systems for cutting-tool changing andworkpiece transport, and complete computercontrol. One observer reports:

The system has demonstrated remarkableefficiencies. They find that the machines inthe system are cutting metal, on the average,about 75 percent, or more, of the time—i.e.,machine utilization is 75 percent better. Leadtime for production of a Tornado is only 18months, compared to about 30 months forplanes produced by more conventionalmeans. The system reduced the number ofNC machines required (compared to doingthe same job with stand-alone NC machines)by 52,6 percent, required personnel by 52.6percent, required floor place [sic] by 42 per-cent, part through-put time by 25 percent, to-tal production time by 52.6 percent, toolingcost by 30 percent, total annual costs by 24

210TA work environment case study.

percent and capital investment costs by 10percent. 22

Finally, a Fanuc Ltd. factory near Mt. Fujiin Japan has received a great deal of attentionand is similarly impressive. The factory pro-duces industrial robots and various CNC tools.It has two automated storage and retrievalsystems (these are described in the next sec-tion) as well as an automated materials han-dling system to deliver materials to worksta-tions. Automatic pallet changers and robotsare used to load and unload machine teds fromthe automatic materials handling vehicles, andthe plant makes extensive use of unmannedmachining at night. The 29 machining centersare attended by 19 workers during the dayshift, while at night no one is on the machiningfloor, and one worker monitors the operationfrom a control room. Several other areas ofthis factory are not automated, however—no-tably, assembly and inspection.23

The chief problems related to an FMS arisefrom its complexity and cost. Several years ofplanning are needed for such a system, and in-stalling and maintaining an FMS is likely torequire a higher degree of technical expertisethan manufacturers may have available. Final-ly, because FMS is a system of interdependenttools, reliability problems tend to magnify. Inparticular, the materials handling portions ofFMS are notoriously troublesome. (See below.)

Despite the advantages claimed for FMS,the systems are still relatively rare. Observersestimate that there are 20 to 30 of such sys-tems in Japan, 20 each in Western and East-ern Europe, and 20 to 30 in the United States.24

The reasons for this scarcity of application in-clude the complexity, newness and cost of the

~XM@ E. Me~h~t, “tint Status of, and potential for, Au@mation in the Metalworking Manufacturing Industry, ” Amdsof the CIRP, vol. 32, no. 2, 1983.

“Ibid; D. Nitzan, “Robotics in Japan-A Trip Report, ” SRIInternational, February 1982.

“’’C AM, An International Comparison, American Mdu”rI-ist, November, 1981, pp. 207-226; W. Dostal, A. W. Kamp, M.Lahner, and W. P. Seesle, “Flexible Manufacturing Systemsand Job Structures” (Mitteilungen aus der arbeitsmarkt andberufsforschung), 1982. Reliable statistics on FMS are difficultto obtain because of conflicting definitions of an FMS and theearly stage of the technology’s development.

66 . Computerized Manufacturing Automation: Employment, Education, and the Workplace— —

systems. One American manufacturer estimatedthat FMS cost $600,000 to $800,000 per machin-ing workstation, with a minimum expenditureof $3 million to $4 million.2s In addition, the in-house costs of planning for installation of anFMS–a process which often takes severalyears-are likely to substantially increase theinvestment in an FMS.

Automated Materials Handling Systems

Automated materials handling (AMH) sys-tems store and move products and materialsunder computer control. Some AMH systemsare used primarily to shuttle items to the workareas or between workstations on automatedcarts or conveyors. Automated storage and re-trieval systems (AS/RS) are another form ofautomated materials handling, essentiallycomprising an automated warehouse whereparts are stored in racks and retrieved on com-puterized carts and lift trucks. For the pur-poses of this report, this category includesonly those materials handling systems whichare not classified as robots.

How AMH Systems Work.–There are awide variety of formats for automated materi-als handling. They include conveyors, mono-rails, tow lines, motorized carts riding ontracks, and automated carriers which followwires embedded in the floor of the factory.Each AMH system is unique, and each is de-signed for the materials handling needs of aparticular factory. The common characteristicof these devices is that they are controlled bya central computer.

There are three general applications for AMH.The first is to shuttle workplaces between sta-tions on an FMS. In this case, the AMH sys-tem operates on commands from the FMS con-troller. For example, when the controllerreceives a message that a machine tool has fin-ished work on a certain workpiece, the control-ler orders the AMH system to pick up theworkpiece and deliver it to the next worksta-tion in its routing. The materials handling por-tion of the FMS is one of its trickiest ele-

26B. Johoski, Manager, Manufacturing Systems Division, Cin-cinnati Milacron Corp., interview, Aug. 16, 1983.

Photo credit’ Cincinnati Milacron Corp

Automated guided vehicle (also known as a “robotcart”) follows wires embedded in the floor of thefactory in order to shuttle workplaces from one part

of the plant to another

ments—part transport needs tend to be logis-tically complicated, and the AMH systemmust place the part accurately and reliably formachining. Many AMH systems, such as con-veyors or tow chains, are serial in nature—i.e., there is only one path from Point A toPoint B. This has caused FMSS to cease op-erating when a cart becomes stuck or a criti-cal path becomes unusable. FMS designershave responded to this problem by designingAMH systems with backup paths, or by usingsystems such as the wire-guided vehicle men-tioned earlier, which can be routed arounddisabled carts or other obstacles.

The second major application of AMH is fortransporting work-in-process from one man-ufacturing stage to the next within a factory.This application is similar in concept to AMHuse for a flexible manufacturing system, al-though serving an entire factory is more com-plex. There is more area to cover, more poten-tial obstacles and logistical difficulties in

CH. 3—Prograrnrnable Automation Technologies ● 67—— ——

establishing paths for the AMH carriers, anda wider range of materials to handle. For thisreason, whole-factory AMH systems are notyet widely used. However, General Motors hasrecently agreed to purchase automatic guidedvehicles from Volvo which allow automobilesto proceed independently through the plantwhile being assembled. The “robot carts” canbe programed to stop at appropriate worksta-tions, and the cart system essentially replacesan assembly line. Volvo uses about 2,000 ofthe carts in its own plants in Europe.26 Fiatalso uses such carts in Italy.

The final application for AMH is in auto-mated storage and retrieval systems. Thesesystems are often very tall in order to conservespace and to limit the number of automaticcarrier devices needed to service the facility.In many cases the structure housing the AS/

26+. Walter, “Volvo Will Build Robot Carts, ” The DetroitNews, Sept. 27, 1983.

Photo credit Cincinnati Milacron Corp

An automated storage and retrieval system (AS/RS),with a computer terminal showing its status. An“automatic stacker crane” (top, center) operates under

computer control

RS is built separately adjacent to the main fac-tory building. Design of an AS/RS depends onthe size of the products stored, the volume ofmaterial to be stored, and the speed and fre-quency of items moving in and out of the sys-tem. Advocates of AS/RS cite advantages forthe system, as compared to nonautomatedsystems, which include lowered land needs,fewer (but more highly trained) staff, more ac-curate inventory records, and lower energyuse.

Applications. –In theory, AMH systemscan move material quickly, efficiently, and re-liably, and keep better track of the locationand quantities of the parts by use of the com-puter’s memory, thus avoiding much paper-work. They can minimize loss of parts in a fac-tory, which is a common problem in materi-als handling.

Deere & Co., for example, uses an extensiveAS/RS to store materials and inventory at oneof its tractor plants.27 The system’s computer-ized controller keeps track of the productsstored on the shelves, and workers can orderthe system to retrieve parts from the shelvesby typing commands at a computer terminal.After they are retrieved from the AS/RS, theparts can be automatically carried by over-head conveyors to the desired location withinthe plant complex.

IBM’s Poughkeepsie plant is planning anAMH conveyor cart system for transportinga 65-pound computer subassembly fixturebetween assembly and testing stations. Themanufacturing manager reports that the deci-sion to adopt this system was prompted bylogistical difficulties in keeping track of manysuch fixtures among a great variety of work-stations, as well as by worker health problemsrelated to transporting the fixtures manually. *

AMH systems often have reliability prob-lems in practice. A Deere & Co. executive re-lated an anecdote at a recent National Re-

2’G. H. Millar, vice president, Engineering, Deere & Co., ad-dress to National Research Council seminar on “The Futureof Manufacturing in the United States, ” Washington, D. C., Apr.13, 1983.

*OTA si~ visit, IBM Corp., Poughkeepsie, NY, June 9, 1983.

68 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace—.—.. — —— -— —

search Council symposium.28 Deere’s AS/RSwas systematically reporting that they hadmore engines stored on the racks than otherrecords indicated. After long weeks of search-ing for the problem they finally found theculprit: A leak in the roof was allowing waterto drip past the photocell that counted the en-gines as they were stored. In essence each dripbecame an engine in the computer’s inventory.

Although Deere’s experience is doubtlessnot widely applicable to AS/RSs, the notionthat AMH systems seem to present unex-pected logistical and mechanical problemsdoes seem to be generally accurate. Eventhough these systems are a key aspect of flex-ible manufacturing systems and of computer-integrated manufacturing, materials handlinghas long been a neglected topic in industrialresearch. Materials handling system manu-facturers have only recently “caught up” toother industrial systems in level of sophistica-tion, and few companies have so far installedsophisticated AMH systems. Because of thisrelative lack of sophistication, materials handl-ing for FMS and CIM, especially for a com-plex application such as delivery of multipleparts to an assembly station, may be one ofthe biggest problems facing integrated auto-mation. 29

Other CAM Equipment

While they will not be addressed in detail,there are several other kinds of programmableautomation equipment used in manufacturing.They include:

Ž Computer-aided inspection and testequipment. — For mechanical parts, themost prominent such device is the Coor-dinate Measuring Machine, which is aprogrammable device capable of automa-tic and precise measurements of parts. A

28G. Milk, op. cit.“B. Roth, professor of mechanical engineering, Stanford

University, “Principles of Automation, ” address to the UnileverSymposium on Future Directions in Manufacturing Technology,Apr. 6-7, 1983; and J. Apple, senior vice president, Systecon,Inc., “Retrieval and Distribution Systems-A Pivotal Part ofFuture Process Planning, ” address to Technology TransferSociety Symposium on Factory of the Future, Oct. 26-28, 1983.

Photo credit National Bureau of Standards

A Coordinate Measuring Machine uses a tiny butprecise probe (center) to automatically

measure parts

great variety of inspection and test equip-ment is also used for electronic parts.IBM’s Poughkeepsie plant, mentionedabove, performs the vast majority of itstesting of microprocessor modules withautomatic devices built in-house. In ad-dition, robots can be used as computer-aided inspection and test devices; severaltwo-armed, gantry-style robots are usedat IBM to test the wiring for computercircuit boards.* In the test, thousands ofpairs of pins on the circuit board must betested to make sure that they are correct-ly wired together. Each arm of the robotis equipped with an electronic needle-likeprobe, and by touching its probes to each

*OTA site visit, IBM Corp., Poughkeepsie, N. Y., June 9*1983.

Ch. 3—Programmable Automation Technologies ●

— —69—

pair of pins and passing an electronic sig-nal through the probes, the robot’s con-trol computer can determine whether thecircuit board’s wiring is “OK.”

● Electronics assembly. —Increasingly, pro-grammable equipment is used to insertcomponents— resistors, capacitors, di-odes, etc. —into printed circuit boards.One such system, Dyna-Pert, manufac-tured by a subsidiary of the EmhartCorp., is capable of inserting 15,000 partsper hour. A programmable machine as-sembles spools of electronic parts in theright order for insertion into the circuitboard, and another machine inserts thecomponents.

● Process control. –Programm able control-lers (PCs) are being used extensively inboth continuous-process and discrete-manufacturing industries. PCs are small,dedicated computers which are used tocontrol a variety of production processes.They are useful when a set of electronicor mechanical devices must be controlledin a particular logical sequence, as in atransfer line where the conveyor belt mustbe sequenced with other tools, or in heattreatment of metals in which the sequenceof steps and temperature must be con-trolled very precisely. Until the late1960’s, PCs were comprised of mechanicalrelays, and were “hard-wired’ ’–one hadto physically rewire the device to changeits function or order of processes. ModernPCs are computerized, and can typicallybe reprogrammed by plugging a portablecomputer terminal into the PC. A comput-erized PC is not only more easily repro-grammed than a hard-wired device, but isalso capable of a wider range of functions.Modern PCs, for example, are often usednot only to control production processesbut also to collect information about theprocess. PCs and numerical control de-vices for machine tools are very similarin concept —essentially, NCS are a special-ized form of PC designed for controllinga machine tool.

Programmable Automation andManufacturing Management

Several kinds of computerized tools arebecoming available to assist in managementand control of a manufacturing operation. Theessential common characteristic of computer-ized tools for management is their ability tomanipulate and coordinate “data bases”-stores of accumulated information about eachcomponent of the manufacturing process. Theability to quickly and effectively get access tothese data bases is an extraordinarily power-ful management tool–what was a chaotic andmurky manufacturing process can becomemuch more organized, and its strengths andweaknesses more apparent. The following sec-tion describes some of these tools, as well asthe notion of “computer-integrated manufac-turing, ” which is not a tool or technology initself but rather a strategy for organizing andcontrolling the factory.

Management Information Systems

Manufacturers use and store information ondesigns, inventory, outstanding orders, capa-bilities of different machines, personnel, andcosts of raw materials, among other things.In even a modestly complex business opera-tion, these data bases become so large and in-tricate that complex computer programs mustbe used to sort the data and summarize it ef-ficiently. Management information systems(MIS) perform this function, providing reportson such topics as current status of production,inventory and demand levels, and personneland financial information.

Before the advent of powerful computersand management information systems, someof the information which MIS now handle wassimply not collected. In other cases, the col-lection and digestion of the information re-quired dozens of clerks. Beyond saving labor,however, MIS bring more flexible and morewidespread access to corporate information.For example, with just a few seconds of com-puter time a firm’s sales records can be listed

70 . Computerized Manufacturing Automation: Employment, Education, and the Workplace—

by region for the sales staff, by dollar amountfor the sales managers, and by product typefor production staff. Perhaps most important-ly, the goal for MIS is that the system shouldbe so easy to use that it can be used directly bytop-level managers.*

Computer-Aided Planning

Computer-aided planning systems sort thedata bases for inventory, orders, and staff, andhelp factory management schedule the flow ofwork in the most efficient manner. Manufac-turing resources planning (MRP) is perhapsthe best-known example of computer-aidedplanning tools.** MRP can be used not only totie together and summarize the various databases in the factory, but also to juggle orders,inventory, and work schedules, and to opti-mize decisions in running the factory. In somecases these systems include simulations of thefactory floor so as to predict the effect of dif-ferent scheduling decisions. MRP systemshave applicability for many types of industryin addition to metalworking.

Another kind of computer-aided planningtool is computer-aided process planning(CAPP), used by production planners to estab-lish the optimal sequence of production opera-tions for a product. There are two primarytypes of CAPP systems—variant and gener-ative.

The variant type, which represents the vastmajority of such systems currently in use, re-lies heavily on group technology (GT). In GT,a manufacturer classifies parts produced ac-cording to various characteristics: e.g., shape,size, material, presence of teeth or holes, andtolerances. In the most elaborate GT systems,each part may have a 30- to 40-digit code. GTmakes it easier to systematically exploit sim-ilarities in the nature of parts produced and

*sometime9 the terms “rnanagernent information system”and “data base management system” (DBMS) are used inter-changeably. MIS tends to refer to a more powerful and comprehensive DBMS aimed for use by relatively high-level staff.

**TWO forms of MRP me mentioned in industry literature.The earlier version was materials requirements planning, a morelimited form of computer-aided planning system for orderingand managing inventory. Manufacturing resources planning issometimes known as MRP II to distinguish it from this earliernotion.

in machining processes to produce them. Thetheory is that similar parts are manufacturedin similar ways. So, for example, a processplanner might define a part, using GT classi-fication techniques, as circular with interiorholes, 6‘ ‘ diameter, 0.01 ‘ ‘ tolerance, and soforth. Then, using a group technology-basedCAPP system, the planner could recall fromcomputer memory the process plan for a partwith a similar GT classification, and edit thatplan for the new, but similar, part.

Generative process planning systems, on theother hand, attempt to generate an ideal rout-ing for a part based on information about thepart and sophisticated rules about how suchparts should be handled, and the capabilitiesof machines in the plant. The advantage ofsuch systems is that process plans in variantsystems may not be optimal. A variant sys-tem uses as its foundations the best guessesof an engineer about how to produce certainparts. The variants on that process plan maysimply be variations on one engineer’s badjudgment.

Though generative CAPP may also dependon group technology principles, it approachesprocess planning more systematically. Theprinciple behind such systems is that the ac-cumulated expertise of the firm’s best processplanners is painstakingly recorded and storedin the computer’s memory. Lockheed-Georgia,for example, developed a generative CAPPsystem called Genplan to create process plansfor aircraft parts (see photo for an example ofa process plan developed by Genplan). Engi-neers assign each part a code based on its ge-ometry, physical properties, aircraft model,and other related information. Planners canthen use Genplan to develop the routing forthe part, the estimated production times, andthe necessary tooling. Lockheed-Georgia offi-cials report that one planner can now do workthat previously required four to eight people,and that a planner can be trained in 1 year in-stead of 3 to 4.*

*OTA site visit, Lockheed-Georgia, Mar. 10-11, 1983. Genplanwas derived from a generative CAPP system developed by Com-puter Aided Manufacturing-International-a consortium forprogr ammable automation research (see ch. 8).

—.

— —

.

. .

... ,.

L . ,. . . . ..+

.

. . ~ ( / ( .

b “ ’ ”

Ch. 3—Programrnab/e Automation Technologies ● 71.—

’ . . , , ( ,

Photo credit Lockheed. Georgia Corp

An excerpt from a process plan developed by the “Genplan” system

Computer-Integrated Manufacturing

Computer-integrated manufacturing (CIM,pronounced “sire”) involves the integrationand coordination of design, manufacturing,and management using computer-based sys-tems. Computer-integrated manufacturing isnot yet a specific technology that can be pur-chased, but rather an approach to factory or-ganization and management.

Computer-integrated manufacturing wasfirst popularized by Joseph Harrington’s bookof the same name, published in 1974. One sys-tems expert recounts the history of the con-cept in this way:

ICIM] came about from: 1) The realizationthat in many cases automation for discreteactivities in manufacturing, such as designor machining, in fact often decreased the ef-

fectiveness of the entire operation–e.g., de-signers could conceive parts with CAD thatcould not be made in the factory; NC machinetools required such elaborate setup that theycould not be economically programmed orused. 2) Development of large mainframecomputers supported by data base manage-ment systems (DBMS) and communicationscapabilities with other computers. TheDBMS and communications allowed func-tional areas to share information with oneanother on demand. 3) The dawning of the mi -crocomputer age which began to allow ma-chines in the factory to be remotely pro-grammed, to talk to each other and to reporttheir activity to their ultimate source ofinstruction .30

soD Wi~no9ky, ~oup vice president, GCA COW., Industri~Systems Group, personal communication, October 1983.Wisnosky is a former director of the U.S. Air Force IntegratedComputer-Aided Manufacturing Program (ICAM).

. ‘, - ~ ‘, .’ } – &j LI – ,, : ,,1, >,

72 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace

Though there is no quantitative measure ofintegration in a factory, and definitions ofCIM vary widely, the concept has become alightning rod for technologists and industrial-ists seeking to increase productivity and ex-ploit the computer in manufacturing. For ex-ample, James Lardner, vice president of Deere& Co., sees the current state-of-the-art manu-facturing process as a series of “islands ofautomation, ” in which machines perform tasksessentially automatically, connected by “hu-man bridges. ” The ultimate step, he argues,is to connect those islands into an integratedwhole through CIM and artificial intelligence(described in the next section of this chapter),replacing the human bridges with machines.In this essentially “unmanned factory, ” hu-mans would then perform only the tasks thatrequire creativity, primarily those of concep-tual design. Lardner’s vision is echoed bymany other prominent experts.

Experts differ in their assessment of howlong it might take to achieve this vision—vir-tually no one believes that it is attainable inless than 10 to 15 years, while some expertswould say an unmanned factory is at leastthree decades away. More importantly, thereare other technologists who argue that the vi-sion may, in fact, be just a dream. For exam-ple, Bernard Roth, professor of mechanical en-gineering at Stanford University, argues thatfactories will, in reality, reach an appropriateand economical level of automation and thenthe trend toward automation will level off. Ina sense, the difference between these twoviews may be a difference of degree ratherthan kind. For many factories, the “appropri-ate” level of automation might indeed be veryhigh. In others, however, a fair number of hu-mans will remain, though they may be signif-icantly fewer than is currently the case.

Integrated systems are often found to re-quire more human input than was expected.31

Indeed, as one engineer explains:32

31This phenomenon has been noted in a variety of places, in-cluding OTA work environment case studies, and the OTAAutomation Technology Workshop, May 29, 1983.

32B, Bums, Manufacturing Technology Group Engineer,Lockheed-Georgia Co., cited in “Considering People Before Im-plementation, ” CAD/CAM Technology, fall 1983, p. 6.

There is much talk about the totallyautomated factory—the factory of the fu-ture—and night shifts where robots operatethe factory. Whereas these situations willdevelop in some cases . . . many manufactur-ing facilities will not be fully automated.Even those that are will involve humans insystem design, control, and maintenance—and the factory will operate within a corpor-ate organization of managers and planners.

These two views do have important signifi-cance for how an industrialist might now pro-ceed. Many who hold the vision of the un-manned factory seem to emphasize technolo-gies, such as robotics, that can remove hu-mans from manufacturing. Those who do notshare the vision of “unmanned manufactur-ing’ tend to argue that there are more practi-cal ways to enhance productivity in manufac-turing, including redesigning products for easeof fabrication and assembly.

How CIM Works.—There are two differentschemes for CIM: In vertically integratedmanufacturing, a designer would design aproduct using a CAD system, which wouldthen translate the design into instructions forproduction on CAM equipment. Managementinformation systems and computer-aided plan-ning systems would be used to control andmonitor the process. A horizontal approach tointegration, on the other hand, would attemptto coordinate only the manufacturing portionof the process; i.e., a set of computer-aidedmanufacturing equipment on the factory flooris tied together and coordinated by computerinstructions. A flexible manufacturing systemwould be a good example of such horizontalintegration. * Vertically integrated manufac-turing is what is most commonly meant byCIM, however, and many experts would con-sider horizontal or “shop floor” integration tobe only partial CIM. Figure 9 is a conceptualframework for CIM which illustrates the roleof some of the PA technologies at variouslevels of factory control.

*Vertical and horizontal integration of programmable auto-mation equipment should not be confused with vertical andhorizontal integration in the markets for selling this equipment;this will be discussed in chapter 7.

Ch. 3—Programmable Automation Technologies ● 7 3— — — — — — —— —

Plan

Figure 9.— Programmable Automation Factory Hierarchy (Simplified)

\Computer ,Integrated

\ Factory / Strategizelmanufacturing

Integratedsoftware

System network

Robotics

/ /

— Station — Move/manipulate

Any manufacturingprocess Process Make

SOURC GCA Corp Industrial Systems Group

A vertically integrated factory usually im-plies maximum use and coordination of all PAtechnologies, and can involve much more cen-tralized control of manufacturing processesthan a nonintegrated production process.Communication and shared data bases areespecially important for CIM. For example,CAD systems must be able to access datafrom inventory on the cost of raw materials,and from CAM systems on how to adapt thedesign to facilitate manufacture. Computer-aided manufacturing systems must be able tointerpret the CAD design and establish effi-cient process plans. And management com-puter tools should be able to derive up-to-datesummary and performance information fromboth CAD and CAM data bases, and effectively help manage the manufacturing operation.

Some parts of the above requirements arealready possible, while others seem far on thehorizon. Factory data bases now tend to becompletely separate, with very different struc-tures to serve different needs. In particular,the extensive communications between CADand CAM data bases will require moresophistication in both CAD and CAM,research on how to establish such communica-tions, and finally, major changes in traditional

v I m p l e m e n t

factory data structures in order to implementsuch a system.

Ap@’cations. -CIM sounds like utopia tomany manufacturers because it promises tosolve nearly all of the problems in manufac-turing that were identified in the section on“the manufacturing process” at the beginningof this chapter, and in particukir it promisesto dramatically increase managerial controlover the factory. Design changes are easy withextensive use of CAD; CAP and MIS systemshelp in scheduling; FMS and other CAMequipment cut turnaround time for manufac-ture, minimize production costs, and greatlyincrease equipment utilization; connectionsfrom CAD to CAM help create designs thatare economical to manufacture; control andcommunication is excellent, with minimalpaper flow; and CAM equipment minimizestime loss due to setup and materials handling.

@Many of the companies which make exten-sive use of computers view their factories asexamples of CIM, but on close examinationtheir integration is horizontal-in the manu-facturing area only–or at best includes pri-marily manufacturing and management. Boe-ing, however, has made substantial strides

74 . computerized Manufacturing Automation: Employment, Education, and the Workplace

toward a common design and manufacturing systems, has embarked on ambitious plans fordata base system in their CAD/CAM Inte- integration at several of its factories, includinggrated Information Network (CIIN). Similar- ity Erie Locomotive Plant, its Schenectadyly, General Electric, as part of its effort to Steam Turbine Plant, and its Charlottesvillebecome a major vendor of factory automation Controls Manufacturing Division.

Technical Trends and Barriers: Future Applications

While the possibilities for application of ex-isting programmable automation tools are ex-tensive, the technologies continue to developrapidly. They depend on and share the extraor-dinary rate of growth in technical capabilitiesof computer technologies as a whole.

There are five themes in the directions fordevelopment in each of the technologies. They

increasing the power of the technologies—i.e., their speed, accuracy, reliability, andefficiency;increasing their versatility—the range ofproblems to which the technologies canbe applied;increasing the ease of use, so that they re-quire less operator time and training, canperform more complex operations, andcan be adapted to new applications morequickly;increasing what is commonly called theintelligence of the systems, so that theycan offer advice to the operator and re-spond to complex situations in the man-ufacturing environment; andincreasing the ease of integration of PAdevices so that they can be comprehen-sively coordinated and their data basesintimately linked.

This section first summarizes the principalresearch efforts and directions for develop-ment of the five technologies on which thisreport primarily focuses: CAD, robotics, NCmachine tools, FMS, and CIM. Next, it sum-marizes issues in several technical areas whichhave a large potential impact on all the tech-nologies: artificial intelligence, standards andinterfaces, human factors, materials, and sen-

sors. Chapters 8 and 9 describe the institution-al context for research and development (e.g.,sources of R&D funding), and compare R&Dprograms on an international basis.

Trends and Barriers in Five Technologies

Computer-Aided Design

There are at least three generations of CADequipment, two of which are widely availablecommercially, with the third still largely inprototype applications and R&D labs. Thefirst are the 2-D computerized drafting sys-tems mentioned earlier in the chapter, whichstreamline the process of drawing and, espe-cially, editing the drawings of parts, plans, orblueprints. The second generation are 3-DCAD systems, which allow the user to drawan image of a part using either wireframemodels or “surfacing” (displaying the surfacesof objects).

The third generation, commercially availablewithin the past few years but still in their in-fancy me the so-called solid modelers. Suchsystems (actually an expanded 3-D capabili-ty) can be used not only to draw the object inthree dimensions but also to obtain a realisticvisualization of the part. Users can rotate,move and view the part from any angle, and,in some cases, derive performance characteris-tics. Engineers at IBM’s Poughkeepsie plant,for example, use an advanced CAD system ofthis type to design cabinet arrangements forIBM mainframe computers. Because the sys-tem “constructs” a sophisticated solid modelof an object, it can be used to visualize suchdesign issues as component clearance prob-lems. One can even “pull out a drawer” tomake sure it does not hit a cable, for instance.

Ch. 3—Programable Automation Technologies ● 7 5——.—— . —. — — ——

Photo credit Coputervision Corp

An exploded view of a part assembly from a CADsystem with solid modeling capabilities

There seems to be a consensus among manu-facturing managers and researchers that suchthird-generation 3-D CAD systems are a criti-cal element in the progress toward effectiveand powerful use of programmable automationin factories. The increased sophistication of3-D systems greatly improves the ability ofsuch systems to communicate design specifi-cations to manufacturing equipment. Whilethird generation 3-D systems are technicallyfeasible now, there are nontechnical barriersto implementation of 3-D systems, in partbecause of the complexity of the systems andthe problems encountered in switching from2-D to 3-D systems. 33 In fact, there is a needfor a fourth generation, a CAD system whichoffers more “intelligent” design assistance andcan be easily linked to other programmableautomation systems for manufacturing andmanagement.

Indeed, there seem to be three relatedthemes in current CAD research:

1. improving the algorithms for represent-ing objects using the computer so design-ers can create and manipulate complex

— - —“R. Simon, Computervision Corp., personal communication.

Oct. 6, 1983.

2.

3.

objects in an efficient and intuitively clearfashion;adding “intelligence’ to CAD systems sothat they prevent design errors and facil-itate the design process; anddeveloping effective interfaces betweenCAD systems and manufacturing andmanagement.

Improving Algorithms. –Representingshapes in computer memory and manipulatingthose representations has been and remainsa difficult challenge for computer researchers.As the power and complexity of CAD systemsincrease, their computing needs grow rapid-ly. One of the problems in manipulating com-plex shapes with the computer is illustratedby the experimental CAD system used forcomputer cabinet design at IBM: One of itscreators reported that a typical manipulationof a complex object—say, generating an im-age of the cabinet from a different viewingangle, with all hidden lines removed—mighttake several minutes of computer processingtime. * Although the system is still useful,clearly quicker response is needed for the de-signer to have optimal flexibility from a CADsystem. A shorter response time can comefrom a faster computer or from more efficientways of representing and manipulating shapesin computer memory.

Although faster computers are unquestion-ably on the horizon, much of the current re-search on CAD involves attempts at more ef-ficient representations. The efficiency of a cer-tain scheme also depends on how easy it is touse. A wide variety of schemes are beingstudied, none of which has a clear overall su-periority. One scheme, called “constructivesolid geometry, “ involves assembling imagesby combining simple shapes, such as blocks,cylinders, and spheres. The other is boundaryrepresentation, in which an object is con-

*OTA Site visit, D. Grossman, IBM Corp. Yorktown Heights,N. Y., June 8, 1983. “Hidden lines” in images are those edgesof a solid object that one cannot see from a given viewing angle.Grossman reports that when CAD is used for such mechanicalmodels as the computer cabinets discussed here, each modelconsists of a polyhedron with roughly 40,000 separate faces forthe computer to store, manipulate, and determine whether theywould be ‘‘hidden’ or not.

76 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace. — . ——. —.

structed as a set of individual surfaces. Forexample, one system being developed by agroup at the University of Utah is based on“splines.” The designer manipulates on thescreen the equivalent of the thin metal stripsused in models of boats or planes. He or shecan expand them, curve them, cut them, andso forth to create the model. 34

There is some concern that not enough timeand effort in industry is being devoted to ex-panding the technologies, particularly the al-gorithms available for “solids modeling, ” i.e.,for true three-dimensional representations ofobjects. Thus the “experience base” of indus-tries experimenting with 3-D systems is verysmall, and such experience is necessary to re-fine the systems and determine the needs ofmanufacturing industries.

Adding “Intelligence” to CAD.–In the in-dustry there is much discussion of “smart”CAD systems which would not permit certainoperator errors. For example, they would notpermit the design of an object that could notbe manufactured, a case without a handle, ora faulty circuit board. Further, they would fa-cilitate the designer’s work by such functionsas comparing a design to existing designs forsimilar objects, and storing data on standarddimensions and design sub-units, such as fas-tener sizes and standard shapes. Such systemsmight also increase the ability of CAD sys-tems to simulate the performance of products.There is much concern over “bad design” inindustry, and intelligent CAD systems areconsidered one way to improve the situation.

Though such systems have become ratheradvanced in electronics applications and offersome hope of becoming more so, there is as yetlittle in the way of “smart” CAD systems formechanical applications. A few systems canbe programed to question a designer’s choiceof certain features that are nonstandard-a22-mm screw hole in a shop that only uses 20-and 30-mm holes, for instance. Some research-ers feel that it will be possible to use an “ex-pert” system (see the next section’s discussion

34R. Reisenfeld, professor of computer science, University ofUtah, personal communication.

Photo credit University of Utaf

A CAD image of a part from a s y s t e mbased on “splines”

of artificial intelligence) for developing a“smart” CAD system.

CAD as Part of Computer-Integrated Man-ufacturing. –Perhaps the most important re-search theme involves connecting computer-aided design to other computerized systemsin the factory. Such connections would mean,for example, that design information could beforwarded directly to machine tools that makethe part, that designers could draw on pre-vious designs as well as data on their perform-ance and cost, and that designers would haveup-to-date information on the manufacturabili-

—— .— —

ty and cost of their designs. Such comprehen-sive connections between design and manufac-turing are currently far beyond the state-of-the-art.

There has, however, been modest progresstoward interfaces between CAD devices. TheInitial Graphics Exchange Standard, devel-oped at the National Bureau of Standardsunder U.S. Air Force sponsorship, allows dif-ferent CAD systems to exchange data (see ch.8). However, while interfaces between comput-er-aided design systems are becoming easier,there is as yet little progress in allowing CADand CAM systems to communicate. In somecases these devices can be wired together intoa computer network, but establishing an effec-tive interface requires sophisticated softwareto manipulate manufacturing information sothat it is useful for designers, and vice versa.

Movement toward design-manufacturingconnections is impeded by a strong traditionof separatism among design engineers andmanufacturing engineers. A common descrip-tion of the relationship is, “The design engi-neer throws the set of drawings over the wallto manufacturing. ” There is evidence thatsuch barriers are beginning to break down,slowly, as the need for communication hasbecome apparent, and as engineering schoolshave begun to broaden the connections be-tween design and manufacturing curricula.

There are many research efforts whose ulti-mate goal includes such connections betweenCAD and other manufacturing systems. Theseresearch programs include the Air Force’s In-tegrated Computer Aided Manufacturing proj-ect, as well as the National Bureau of Stand-ards (NBS) National Engineering Laboratoryand a joint West German/Norwegian effort(see ch. 9). The heart of the latter effort is anattempt to use a very advanced geometricmodeling system developed by the TechnicalUniversity of Berlin as the basis for develop-ing software which would allow design to beconnected to all aspects of the manufacturingprocess. In addition, users of PA, such as GEand IBM, are also working on interface issues.However, full integration still seems at leasta decade off.

Ch, 3—Programmable Automation Technologies ● 7 7— . . —

Robotics

Robotics research is currently an area of in-tense interest in both industry and univer-sities. There are a dozen or more universitieswith significant ongoing research projects inrobotics, and perhaps 3 dozen industrial firmsand independent laboratories. Governmentlabs at NBS, the National Aeronautics andSpace Administration (NASA), and several atDOD, are also involved.

In part because of the technical immaturi-ty of robotics technology, and in part becauseit is a complex and interdisciplinary technol-ogy, there are many discrete areas of researchproblems and possible directions for extensionof capabilities. The problem areas include:35

● Improved positioning accuracy for the ro-bot arm. —Increased accuracy is essen-tial for many applications of robots, par-ticularly in assembly operations and othercases where a robot is programed offline.While current robots are precise (they canreturn to the same position on each cyclefairly reliably, within perhaps 0.005 inch),their accuracy (the ability to arrive at apredetermined point in space), is not near-ly so reliable. Several techniques are be-ing used to increase accuracy in robots.Though the traditional answer has beento increase the stiffness and mechanicalprecision of the manipulator arm, such ap-proaches can greatly increase the weightand cost of the unit. Software calibration,a technique being developed at the NBS,involves adjusting the robot electronicsto compensate for inaccuracies in itsmovements. Another technique involvesusing machine vision systems to “watch’the robot in action and correct its move-ments as they occur-this technique couldpotentially improve both accuracy andprecision. Of the two, software calibrationis far simpler technically and is likely tobe available far sooner.

“(J. S. Albus, “ Industrial Robot Technology and Productivi-t~ Impro\’ement, ” Exploratory 14’orkshop on the Social im-pacts ofllobotics: Summary and Issues (Washington, D. C.: U.S.Congress, Office of Technology Assessment, OTA-BP-CIT-I 1,February 1982), pp. 62-89.

78 . computerized Manufacturing Automation: Employment, Education, and the Workplace——.— — — - —

Increased “grace, dexterity, and speed. ”—The physical structure of the manipu-lator—its material, actuation mechanism,and joints—has remained substantiallythe same for several decades. Severalgroups of researchers, sponsored byNASA and the Defense Advanced Re-search Projects Agency (DARPA), amongothers, are working on lighter structuresfor the robot arm. These would most like-ly consist of composite fiber materialssimilar to those now used extensively onaircraft-about one-sixth the weight ofsteel. Though the technology for suchstructures exists, composites are extremely expensive and the cost is holding backfurther use in robotics. Other directionsfor progress in robot structures includefundamentally different designs for themanipulator arm. A Swedish group hasdeveloped an arm which is structured insome ways like a human spinal column,while other research is directed towardusing “tendons” to effect movement ofthe arm, as in the human musculoskeletalinteraction.

Cost is not the only drawback to the useof lighter structural materials. In addi-tion, the robot’s controller must becomemore sophisticated in order to direct themotions of a lightweight, and inherentlysomewhat flexible, robot arm. For in-stance, computer scientists and mathema-ticians must develop control algorithmsthat will prevent backlash–i.e., the “play’or vibration that occurs when the arm ismoved quickly from one position toanother.

Finally, gripper design needs to bemade more flexible. Directions for prog-ress in grippers include both developing“hands” that can be used to manipulatea wider variety of objects, and also devel-oping “quick-change” grippers so that therobot can autonomously exchange one“hand” for another.

● Sensors, including vision, touch and force.–Because sensors can be applied to awide range of programmable automationdevices, they will be addressed in a sepa-

rate section later. One problem relevantto robots is the development of controlsystems that can accommodate sensoryinformation. Systems are only now begin-ning to become available that can acceptfeedback from various kinds of sensors.In part, these systems have developed inconjunction with new generations of robotprograming languages, to be discussedbelow. A continuing tension in devel-opment of robots is whether one shouldstructure the robot’s environment so thatit does not need extensive sensing, or tryto provide sensors to enable it to copewith an unstructured environment.

● Model-based control systems. —The mostadvanced and versatile controller for arobot would be one that had an internalmodel of its environment. In other words,it would have a store of information aboutthe three-dimensional world, what the ob-jects it worked with were supposed tolook or feel like, and the rules for howphysical objects interact with each other.Although this problem has intriguedmany technologists, who view it as oneof the ultimate solutions for expandingrobot versatility and “intelligence, it isextraordinarily difficult to impart such in-formation to the machine, and even todecide how one might structure suchinformation.

Software. -Methods for programing ro-bots are becoming easier and more effi-cient, although there is still substantialwork needed in this area. Two languageshave been released recently–IBM’s AManufacturing Language (AML) andRAIL, by Automatix–which are consid-erably more powerful than traditionalrobot languages, and which permit moresophisticated programing techniques,similar to advanced general-purpose com-puter languages such as PASCAL orADA. Most other programing languagescurrently available are rather cumber-some and inflexible by computer-industrystandards. At the same time, teaching-by-guiding programing is becoming lesspractical for complex applications; it

Ch. 3—Programmable Automation Technologies ● 7 9. — — — — . — —

delays production and has very limitedcapabilities for editing the program orusing sensory information.

There is still much progress to be madein human interfaces with robots—the de-sign of languages and programing sys-tems that can be most easily and effec-tively used by humans. One technique forimproving human interfaces, which hasjust become available, is the use of CADto program robots and simulate their op-eration. The ability to visualize the ro-bot’s path may permit more effectiveplanning and and debugging of programsso that production need not be stoppedin order to test a robot program.

Interface standards. — Standards need tobe developed for communication of infor-mation between robots and machine tools,sensors, and control computers. Whilesuch standards are a tractable problem,programmable automation producers, aswell as the computing industry, are onlybeginning to make progress in establish-ing standards, and the standards-makingprocess is long and intricate. In the mean-time, efforts to establish interfaces be-tween robots and other automated de-vices are hindered by a lack of standards.Researchers at NBS report that somemanufacturers refuse to divulge details ofthe operation of their equipment that

Photo credit Computervision Corp

A CAD-based simulation of a robot’s operation

80 ● computerized Manufacturing Automation: Employment, Education, and the Workplace

would enable the equipment to be linkedto other computers. (See the following dis-cussion of standards and interfaces.)

● Mobility. –While techniques for limitedmovement along rails are already avail-able for robots, the more general problemof developing a robot that could navigateits way through a cluttered factory is farmore difficult. There is some argumentabout whether such mobility is even nec-essary for the factory-some would assertthat such technology is too esoteric forthe factory, and the plant should insteadbe organized so that mobility is unneces-sary. There is substantial research inmobility, however, in large part sponsoredby DOD agencies for specific battlefieldapplications.

Numerically Controlled Machine Tools

Although machine tools are a well-estab-lished technology, there continues to be a needfor substantial improvements in the tools andtheir controllers. A “Machine Tool TaskForce, ” operating under the auspices of DOD’sAir Force Materials Laboratory, issued a re-port in 1980 calling for hundreds of improve-ments and new research efforts. Among theones most relevant to this study are thoselisted in table 10.98

“Machine Tool Task Force, “Technology of Machine Tools:A Survey of the State of the Art” (Livermore, Calif.: LawrenceLivermore National Laboratory, October 1980)

The bulk of machine tool R&D takes placein the laboratories of machine tool and con-troller manufacturers. A smaller but signifi-cant amount of work is undertaken at univer-sity mechanical engineering departments, withfunding from industry or the Federal Govern-ment. The chief research problems can bedivided into those involving the machine itself,and those involving the controller.

The Machine Tool.–Many of the develop-ment needs for machine tools involve deviceswhich facilitate the use of the tools under com-puterized control. For example, chip removal–disposal of the metal shavings that accumu-late in large volume during machining-is abig problem in industry, a problem that getsbigger as machines get more efficient andmore automated. Various schemes have beenused for chip control and disposal, none ofwhich are entirely satisfactory. Many engi-neers believe the answer is not to create thechips in the first place—by forging the partclose to its final shape, for example, or machin-ing with lasers instead of cutting tools. While“near net shape” forging is becoming moreprevalent, laser machining is still immature,and not yet practical for widespread appli-cations.

Another problem in machine tools, whetherautomated or manual, is tool wear. A drill bitor grinding wheel has a fixed useful life, afterwhich the quality of cut begins to decline andthe tool eventually fails. The traditional solu-tion to tool wear is simply for an experienced

Table 10.—Machine Tool Task Force Recommendations for Improving Machine Tool Controls

Hardware (H) orsoftware (S)

Area of improvement development Objective

Toolsetting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . H/s Reduce setup time, improve accuracyDiagnostics and sensors . . . . . . . . . . . . . . . . . . . . . H/S Allow more of the important parameters to be sensed and

monitored for failure identificationFixturing/clamping . . . . . . . . . . . . . . . . . . . . . . . . . . H More versatility of fixture and less setup timeNC programing and instruction . . . . . . . . . . . . . . . S Develop improved new computer subroutines to simplify

and reduce time for programingProgrammable controls . . . . . . . . . . . . . . . . . . . . . . H Integrate machine processes into computerized system;

enhance conventional machine operations; provideinterfacing devices and flexibility

Interface standards . . . . . . . . . . . . . . . . . . . . . . . . . . H/S Improve upgrading and growth-retrofit potential;interchangeability

SOURCE Machine Tool Task Force, Technology of Machine Tools, October 1980

Ch. 3—Programmable Automation Technologies ● 81

machinist to listen to the machine tool and,ideally, sense when noise and vibrations be-come abnormal. In situations where that is notpossible, particularly on CNC machine tools,machinists replace the tool after a specifiedperiod of tool life. In addition, the Japaneseare said to run their machines at slower speedsin order to minimize tool failure during un-manned machining. However, tools can fail atalmost any point in their use—a drill bit mayfail after it only drills a few holes, or it maylast for hours. This variability makes pre-scheduled replacement difficult and inefficient.There has been some progress in developingdevices that can sense tool wear and reportwhen tool change is needed. The NationalBureau of Standards, for example, has a pro-totype device that “listens” to the vibrationsproduced by the tool and can be “taught” torecognize abnormal vibrations.

The rate at which a machine tool can cutmetal depends on many factors—the type ofmetal, the depth of cut, the condition of thetool, and so forth. Controlling the speed of cutor the feed rate so as to cut metal at optimumremoval rates has been a continuing researchand development problem in the industry. Aswith sensing tool wear, the traditional answerhas been for experienced machinists to adjusta cutting speed or feed rate dial on the ma-chine. In the past decade, various “adaptivecontrol’ devices have been developed whichvary the “speeds and feeds” of the machinetool based on motor load, for instance. How-ever, these devices have had uneven reputa-tions for effectiveness and reliability.

Finally, a great deal of effort is now beingdevoted to increasing precision in machinetools. The Navy’s precision manufacturingprogram will be described in chapter 8.

Related to improvements in precision, along-term goal for machine tool technology in-volves measurement of parts during machin-ing. With such a scheme, quality problemscould be identified and corrected during man-ufacturing rather than afterwards, therebyreducing waste. NBS has done some prelim-inary research on such a system of on line

metrology at machine tools, although commer-cial use of such systems is limited to very sim-ple and predictable part geometries.

Machine Tool Controllers. -As with allother forms of programmable automation,there is continuing demand for and researchon simplifying programing of NC machinetools; the same holds true for the need tosimplify and set standards for interfaces—be-tween machine tool and controller, betweenmachine tools, and between machine tools andother automation devices. A critical issue isthe development of effective interfaces be-tween CNC machines and other computerizeddevices, so that, for example, CNC machinescan derive their cutting instructions from thestored dimensions of a design produced withCAD. This is now possible only in specificlimited situations, where tremendous efforthas been devoted to developing the interfacesfor a particular application.

Flexible Manufacturing Systems

Flexible manufacturing systems for the ma-chining of prismatic parts are becoming moreprevalent, and are a relatively establishedtechnology. FMS for rotational parts are justbeginning to be available, while the range ofother possible applications for FMS—grind-ing, sheet metal working, or assembly—are notbeyond the reach of current technology, butare only at early stages of development.

Many of the chief R&D problems for FMSinvolve logistics: design and layout for theFMS, and computer control strategies thatcan handle sophisticated combinations ofpowerful machine tools. In addition, there isa need for more sophistication in simulationsystems for the FMS so that their efficiencycan be optimized.

There are a variety of enhancements to FMShardware which seem to be on the horizon. Inaddition to all of the developments describedunder the individual technologies, these in-clude automatic delivery and changing of cut-ting tools, and systems for automatic fixturingand refixturing of material to be processed.

82 . Computerized Manufacturing Automation: Employment, Education, and the Workplace

Improving the reliability and versatility ofmaterials handling systems is also an impor-tant need for FMS. As mentioned earlier, thelevel of sophistication in materials handlingtechnology often does not match that of otherPA technologies, and the AMH system maybe the “weak link” in the FMS.

Computer-Integrated Manufacturing

Computer-integrated manufacturing receivessubstantial attention in industry discussionsand trade journals, though there is relativelylittle active R&D at this level of the comput-erized factory. This is at least partly becausethere is not yet substantial demand for CIMsystems. GE and IBM have begun to work oncomputer-integrated manufacturing, as havesome Japanese firms, particularly Hitachi, anda coalition of laboratories in West Germanyand Norway. The Automated ManufacturingResearch Facility at NBS is perhaps thelargest test bed for CIM techniques. It isdescribed in more detail in chapter 8.

As with FMS, one of the key issues in CIMdevelopment involves the logistics of a com-plicated factory. Several groups, includingNBS, the U.S. Air Force ICAM project, andComputer-Aided Manufacturing International(CAM-I), have been working on “architec-tures” for such an automated factory. Figure10 is an example of such a conceptual frame-work for CIM which forms the foundation fordetailed work on factory control architectures.

One of NBS’s major contributions in auto-mation R&D has been in developing strategiesfor the interface of programmable automationdevices. Their emphasis has been on what theycall a “mailbox” or decentralized approach tofactory communication and control. In sucha system, the control of the factory is distrib-uted at different levels among the various PAdevices (see fig. 10). For example, a factory-level computer might send a message to a production-level computer—’ ’Make 150 of partnumber 302570. ” The production-level com-puter would then send a message to the “mail-box” of a certain work cell-” Execute produc-tion plan for part 302570, 150 times. ” In turn,

the work cell controller would send messagesto the mailbox of the machine tools and robotsin the cell, to execute certain programs storedin their memory.

The “mailbox” approach differs from a cen-tralized, or “star,” approach to automated sys-tems control in which a central computer di-rectly controls each action of every machinein the factory. The advantages of the mailboxsystem are that it simplifies standards and in-terface problems—the only interface standardnecessary is for the location of the mailbox inwhich to deposit messages. This allows onerobot to be substituted for another, for exam-ple, with relative ease. The mailbox approachallows different PA devices to operate usingdifferent languages and proprietary operatingsystems, as long as they are able to interpretmessages from the computer controller.

Hierarchical arrangements for automatedmanufacturing, such as those illustrated in fig-ure 10, tend to involve a large number of sep-arate computers, each with separate databases. Techniques for “distributed data basemanagement, ” that is, managing and manip-ulating data in several computer systemssimultaneously, need to be developed in orderfor a hierarchical arrangement to be practical.Similarly, techniques and standards for estab-lishing communication between computerizeddevices, both in-plant and between plants,need to become much more sophisticated.

A group of researchers at Purdue Universi-ty, in collaboration with several large manu-facturers, is attempting to exploit currentlyavailable technology to design an actual fac-tory with maximum computer integration.The leader of that effort argues that the tech-nology for CIM is available, and that technicaladvances, though welcome, are not necessary.Rather, he argues that factors holding back“fully” automated manufacturing are primarily:

1.

2.

3.

the lack of standards for interfaces, com-munication networks, and programinglanguages;a need for more powerful data-base man-agement systems;the need for detailed mathematical

Ch. 3—Programmable Automation Technologies ● 8 3—

Figure 1O.—NBS

.., .. . . . . .-- ,.. ,,, ..

,,, ! .-.

<. .

- .

C* . . . . .. . . . .

. . . ” “ . m . -

S!.G. . . -

‘ . n . . ”

., .-. .. . .. . .W. .

w ’“.. ,,.,

. . ~ :..’””-. . . ..,.,-. .,. ,.

.,

. .

Scheme for Distributed

Orders

Shopcontrol

1

ACellcontrol

Factory Control

w “ “Q___SOURCE PJatl~nal Burpab of Standards

4.5.6.

models of physical and chemical proc-esses;shortages of technical personnel;shortages of computer power; andmanufacturing management who areunaware of the detailed technical benefitsof automation.37

Technical Trends and Barriers:Cross-Cutting Issues

The following issues are not primarily con-nected to a particular automation technology,

17 T, \\’illiams, Purdue I.aboratory for Applied Industrial Con-t ml. “ Information Systems Technology and Automation: ItsI’resent Da3’ Status and a Prognosis, ” paper developed for theAmerican %ciet~’ of Nlechanical Engineers Wrinter Annualhleeting. Boston, Nov. 15, 1983.

/

-----

but rather have a large potential impact on thecurrent and future capabilities of automationtechnologies as a whole.

Artificial Intelligence*

Artificial intelligence (AI) is a loose con-glomeration of research areas united by thecommon goal of designing machines which canperform tasks we would generally regard asrequiring intelligence. It is significant for pro-grammable automation because many expertslook to AI techniques as the key to automat-ing parts of the manufacturing process here-

*A forthcoming OTA report, “Information Technology Re-search and Development, ” will discuss artificial intelligence inmore detail.

84 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace— .— —— —

tofore thought to be too complex for automa-tion.

The core of basic, long-term AI research in-cludes work on imparting such intelligentcharacteristics as learning, reasoning andplanning to computers. Building on some ofthis work are several more applied researchareas: the most sophisticated end of the robot-ics field; the development of systems for im-age processing–deciphering images from videocameras or touch sensors; development of tech-niques to allow the computer to understandnatural language (e.g., English, as opposed tocomputer languages such as FORTRAN),both written and spoken; and expert systems,programs which can, through a sophisticatednetwork of rules, advise or make decisions inspecific situations much as a human expertwould.

While robotics and sensors (image under-standing) have been largely covered elsewherein the chapter, natural language and expertsystems both have significant potential ap-plications for manufacturing in this decade.

Commercial systems for processing bothwritten and spoken language have receivedsubstantial attention in the past 5 years. Thehope for both kinds of systems is that, byallowing people to give commands and com-municate with computers in everyday lan-guage, widespread use of the computer will besubstantially easier. Fewer people would needto learn specialized computer languages, andfewer computer experts would be needed as in-termediaries between computers and thosewho wish to use the computer as a tool.

The primary application for computer proc-essing of written language has been in the de-velopment of so-called natural language frontends for data-base management systems(DBMS). In such a system, the data base–e.g., sales records for a company—and theDBMS itself used to manipulate and sum-marize that data, remain essentially the same.However, the natural language “front end”allows users to type questions in relativelyfree-form English, and translates those ques-

tions into the specialized language used by theDBMS.

For example, without a natural languagefront end, a plant manager who sought theanswer to the question, “Which products inthe 2000 series were sold in volumes of morethan 1,000 last year?” would probably referthe question to a programer, who would writea short program in a computer language toprocess the request.* With a natural languagefeature added to the DBMS, that plant man-ager could type his request, more or less ex-actly as he would say it, into a computer ter-minal, and the requested information wouldappear on the screen.

In general, though, scientists have foundnatural language understanding to be a muchgreater challenge than originally expected.Because there are so many ambiguities andunclear references, understanding everydaylanguage requires substantial informationabout the context of a given statement orquestion, and the world in general. Organiz-ing such information to allow natural languageunderstanding by computer has proven to bean extremely difficult task, in part because ofour very incomplete understanding of howpeople store and manipulate such information.

In practice, this means that constructing anatural language front end for a DBMS re-quires weeks or months of work in writingcode that sets forth for the computer the vari-ous meanings of the terms used for a particu-lar application, and the possible ways they canbe combined. Although many such systemscan interpret relatively freeform questions,they are also severely constricted in subjectarea. In other words, the same system whichcan interpret questions about a firm’s salescannot decipher queries about the design ofits products.

*ln one DBMS language, such a (very simple) program fightlook like:

OPEN SALES 83:FIND ALLFIND PRODNO GT 1999 AND LT 3000FIND SALESVOL GT 1000LIST PRODNO, SALESVOL

Ch, 3—Programmable Automation Technologies ● 8 5. — —

Systems for computer processing of spokenlanguage present a different set of problems.These involve techniques for analyzing theelectronic signals produced by a human speak-ing into a microphone, and comparing themto signal patterns stored in computer memory.Voice recognition systems have been devel-oped which are capable of interpreting perhaps100 words, spoken distinctly and usually bya speaker to which the system has been“trained” to listen. Such systems can be used,for example, to allow workers to give the com-puter simple commands when they do nothave a hand free to use a keyboard. Variousother uses have been proposed for such sys-tems, from directing the motions of a robot tooperating a CAD system, although the limitedvocabulary and lack of flexibility of such SYS-tems has hindered widespread use. Rapid ad-vances in hardware and software for voicerecognition may expand their capabilities inthe next few years.

Both for written and spoken language under-standing systems, the limited breadth andflexibility of applications is a consistenttheme. In fact, a general solution to the prob-lems of natural language understanding-e. g.,one that could impart to a computer the lan-guage understanding capabilities of a 5-year-old child—is probably at least two decades off.

Finally, expert systems can allow the use ofcomputers in situations normally thought tobe so complex that they require human judg-ment or ‘‘intuition. AI researchers havefound that in a narrowly defined problem area,it is sometimes possible to simulate much ofthe judgment of human experts by breakingdown that judgment into hundreds of rules forthe information to look for under different cir-cumstances, and how to weigh that informa-tion.

Expert systems are typically composed ofhundreds of rules, gathered by painstaking in-terviewing of human experts, about exactlyhow they make their judgments. The interviewand development process for an expert systemtypically takes several years to complete,although it is becoming less time-consuming

as techniques for developing the systemsbecome more refined. The interview tech-niques used by expert system developers allowthe systems to capture many of the subtletiesof how an expert arrives at judgments.

Two of the classic expert systems, for ex-ample, are “Dipmeter Advisor, ” developed bySchlumberger to offer advice on oil geology,and Stanford’s MYCIN, which advises doctorson antibiotic drug treatment. Figure 11 is anexcerpt from the rules upon which the MYCINsystem is based. The system asks the doctora series of questions about the patient, andthen offers a diagnosis and treatment (see fig.12). It can explain the rules it used to arriveat its recommendation, and physicians canalso improve the system by adding new rules.

Some of the more recent applications of ex-pert systems, closer to manufacturing, are asystem which guides mechanics in the repairof locomotives (Delta/CATS-l, developed byGE), and a system which helps configurecomplex computer systems (XCON, formerly“R-l, ” developed by Carnegie-Mellon Univer-sity and Digital Equipment Corp.).

Though the commercialization of expert sys-tems is only beginning, many industrialistshave high hopes for their use in manufactur-ing. A myriad of applications have been pro-posed, including systems which could mimicthe performance of a human machinist; sys-tems for advising designers and preventingdesign errors; systems which would act as alinkage between manufacturing and design

Figure 1 l.– Sample Rule from the MYCINExpert System

Rule 27If 1) the gram stain of the organism is gram

negative, and2) the morphology of the organism is rod,

and3) the aerobicity of the organism is

anaerobic,Then: There is suggestive evidence (0.7) that

the identity of the organism isBacteriodes.

SOURCE R Davis and C Rich, Expert Systems Fundamentalstutorial at 1983 conference of the American Associationfor Artificial Intellgience

86 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace—

Figure 12.—Sample “Advice” From the MYCINExpert System

[Rec 1] My preferred therapy recommendation is asfollows:

In order to cover for items <13456>:Give: GENTAMICINDose: 128 mg (1.7mg/kg) q8h IV (or IM) for 10

daysComments: Modify dose in renal failure

In order to cover for item <2>:Give: CHLORAMPHENICOLDose: 563 mg (7.5 mg/kg) q6h for 14 daysComments: Monitor patient’s white count

Do you wish to see the next choice therapy?● * NOSOURCE: R. Davis and C. Rich, Expert Systems: Fundamentals, tutorial

at 1983 conference of the American Association for ArtificialIntelligence.

data bases; and even systems for overall fac-tory control.

Two researchers describe the characteristicsof a problem area which makes a good domainfor expert systems as one where:38

There are recognized experts; the expertsare provably better than amateurs; the tasktakes an expert a few minutes to a few hours;the task is primarily cognitive; the skill is(routinely) taught to neophytes; the task do-main has a high payoff; the task requires nocommon sense.

It is unclear in this early stage of applica-tion of expert systems just how widely appli-cable these tools will be. While the successesto date have been impressive, each of the cur-rent systems has been the result of many yearsof effort in top AI laboratories. Furthermore,they have succeeded in very carefully selected,and very carefully restricted problem areas.For example, GE’s system for diagnosing loco-motive problems cannot be used to diagnoseautomobiles, or even to diagnose differentbrands or configurations of locomotives with-out major alteration.

With current high levels of interest in ex-pert systems, and evolving tools and tech-niques to streamline their development, itseems likely that these tools will be used in

SER. Davi9 ~d c. Rich, ‘‘Expert Systems: Fundamentals,tutorial at American Association for Artificial Intelligence 1983annual conference, Washington, D. C., Aug. 22, 1983.

several areas of manufacturing. However, itis unlikely that expert systems will in the nearfuture meet the many expectations which theirrecent successes have generated. It is easyboth to underestimate the development effortand skills needed to construct such tools, aswell as to imagine new applications in areaswhich are too broad or ill-defined for currenttechnology to handle.

Sensing this problem, one recent NationalResearch Council committee report warned of“unrealistic expectations:”39

In an extremely narrow context, someexpert systems outperform humans (e.g.,MACSYMA), but certainly no machine ex-hibits the common sense facility of humans atthis time. Machines cannot outperform hu-mans in a general sense, and that may neverbe possible. Further, the belief that such sys-tems will bail out current or impending disas-ters in more conventional system develop-ments that are presently under way is almostalways erroneous.

One of the dangers of high expectations forexpert systems and other areas of AI is thatif these expectations are unmet, there couldbe a backlash and loss of interest in AI. Thefield has already suffered from two or moresuch cycles of high expectations and loss ofinterest and credibility. Indeed, AI has longbeen an area in which claims and hopes aremore prevalent than concrete successes,though current workers in this area seem tobe rather more cautious.

Manufacturers are not alone in their highhopes for AI, as evidenced by Japan’s recent“Fifth Generation” computer project, andDOD’s new Strategic Computing project.Both of these programs are long-term, ambi-tious R&D in computer hardware and softwarein which AI plays a primary role (also see ch.8). Another major goal of both programs is thedevelopment of “supercomputers.” Thoughthe definition of supercomputers changes as

“Committee on Army Robotics and Artificial Intelligence,Manufacturing Studies Board, National Research Council, Ap-plications of Robotics and Artificial Intelligence to Reduce Riskand Improve Effectiveness: A Study for the United StatesArmy (Washington, D. C.: National Academy Press, 1983), p. 63.

Ch. 3—Prograrnmable Automation Technologies ● 8 7

the technology develops, a current workingdefinition is machines that can process morethan 100 million instructions per second. AIand supercomputers tend to be discussed to-gether and often confused with each other.However, though both AI and supercomput-ers are at the frontiers of computer science,they are essentially separate research areas atthis time. It is likely, though, that future AIapplications will require advanced computerarchitectures-not high-speed number crunch-ers as much as machines designed to processSymbols and logic..

Although the infusion of DOD funds into AImay expand and advance the field, defense ap-plications may also continue to monopolize thesmall pool of U.S. AI researchers. Despite thefact that DOD is making concerted efforts toencourage commercial spinoffs from the Stra-tegic Computing project, most of the atten-tion of the AI R&D community will be focusedon military applications rather than commer-cial manufacturing applications.

Much of the current wave of commercializa-tion of Al is based on AI research done asmuch as a decade or more ago. In many cases,commercial applications have recently becomefeasible because of the continuing declines incosts of computing power. While one can ex-pect further improvements in available Al-Based tools over the next few years, these im-provements may be small in comparison tothis initial harvest. The more fundamentalproblems of AI, involving natural languagesystems of general applicability, versatile andunstructured machine vision, and—ultimate-IY-–generally intelligent, perhaps “learning”machines, are still very much a long-term re-search issue.

Standards and Interfaces

The need for standards in both languagesand interfaces is strong and consistent through-out programmable automation technologies.Without standards, it is very difficult to com-bine equipment of different vendors, and it ismore difficult to proceed incrementally towardcomputer-integrated manufacturing.

The demand for standardization in lan-guages is particularly strong from users of

automation devices, because of the increasedconfusion and need for additional training thatresult from the many different programinglanguages. 40

More likely than one standard language formanufacturing, however, may be a set ofstandard languages for each application. Forexample, there might be a standard languagefor arc-welding robots, another for materialshandling systems. These could be either for-mally adopted or de facto standards (i.e., theybecome commonly accepted through usage orthrough the influence of major vendors orusers in the field). For example, many ofIBM products and techniques are treated asstandards because the company has domi-nated the computer field. However, domina-tion by a single firm in programmable auto-mation systems is not as likely (see ch. 7). Inaddition, DOD has created many de factostandards, APT among them, through its pro-curement practices. It remains to be seen towhat extent DOD’s latest attempt at a stand-ard computer language, ADA, will be appli-cable to manufacturing systems.

1n addition to standards for programing lan-guages, standards for interfaces between com-puterized devices will greatly facilitate inte-grated PA systems. The recent developmentof standards for “local area networks, initial-ly aimed to connect personal computers in of-fices, may also be useful in the factory. ” Suchstandards define the hardware connections forhooking devices together, as well as the “pro-tocols” that ensure that different systems caninterpret each others’ messages. However, thecontent of those messages depends on the ar-chitecture of the factory —i.e,, the differentlevels of control and the kind of informationit is necessary to communicate. As discussedearlier, efficient architectures for integratedfactories are only beginning to be worked outin manufacturing laboratories such as theAutomated Manufacturing Research Facilityat NBS.

4“OTA automation technology workshop, May 29, 1983.“’’Local Area Network Facilitates Factory Automation, ”

Tooling and Production, May 1983, p. 94. These networks arebased on a professional association-developed standard knownas IEEE802.

2 5 - 4 5 2 0 - 8 4 - 7

88 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace— — .- .

Manufacturers and others often argue thatpremature standardization will stifle innova-tion. It can tend to “freeze” a technology ata particular point in its development, and dis-courage further innovations which may be in-consistent with the standard. In addition,there is sometimes a strategic concern thatstandard languages cause more competitionby permitting easier combination of PA de-vices from different manufacturers. One NBSofficial has argued that parts of the computer-ized controllers for machine tools, for exam-ple, are technically ripe for standardization butthe machine tool manufacturers do not seemto support such a move.42

Apart from any resistance to standards,there is the fact that implementation of stand-ards is voluntary in the United States, whichis not the case in many other countries. As aresult, development of a successful standardtakes years of negotiation among manufactur-ers and users. To complicate matters, recentcourt decisions43 have held organizers of stand-ards efforts liable if a standard can be shownto hurt a particular company. This has madeprogress toward adoption of standards inmany areas even more cautious and slow-going.

NBS staffers contribute to standards effortsby serving on and helping to coordinate themany private sector standards committeesworking on automation issues. Relevant ef-forts are being conducted by the Electronic In-dustries Association, the Robotic IndustriesAssociation, the American National Stand-ards Institute, the American Society for Test-ing and Materials, the Institute of Electricaland Electronic Engineers, and the Interna-tional Standards Organization.

. - — —“R. Hocken, NBS, personal communication, Aug. 12, 1983.

For example, some NC controllers only understand a numberto be “two” if it is written as “2”. Others require it to be writ-ten as “2.000” or “2. EOO”. This can cause difficulty in tryingto move programs from one machine to another, even if themachines ostensibly use the same language.

4‘1 n American Societ-v of Mechm”cal Engineers v. Hwvdrole\’-e) Corp. 102 S. Ct. 1935 ( 1982), the Supreme Court held thata standard-setting organization was liable for the antitrust viola-tions of participants in the standards-making process when theyacted with the apparent authority of the ASME.

Human Factors Research

In the past few years, makers of all com-puterized equipment have become aware of aneed to design systems for optimal usefulnessand productivity for their human operators.There are various terms used to describe thefocus of such efforts: “user friendly” qualitiesand “man-machine interface, for example. *In part to help market their equipment, com-puter manufacturers have found that there aresteps they can take to improve the human fac-tors aspects. Human factors experts arguethat research and testing of the effectivenessof a product must be undertaken throughoutits design cycle. “Human engineering, whichwas seen as the paint put on at the end of aproject, is becoming the steel frame on whichthe structure is built. ”44

Although many experts agree on the impor-tance of human factors, it has often been aneglected topic in research. It is frequentlyregarded as too basic for industry to examine,and too applied for university research efforts.Although DOD has pursued man-machine in-teraction research for decades, only recentlyhas human factors become a subject of sys-tematic study outside of DOD. Psychologistshave developed testing procedures to helpdetermine the human effectiveness of differentdesigns. Recently, human factors of computersystems has become a strong and growingsubfield of cognitive psychology.

Designers of programmable automationequipment have lagged behind the trendtoward concern about human factors in com-puterized systems, in part because of thenewness and small size of the market for manyautomation devices. In addition, some PA de-vices such as robots or portions of FMSS areoften designed with the intention of minimalcontact with humans.45 Several systems de-

● A Variety of termsby used by researchers industry todescribe human factors and related subjects. Some others notmentioned in this section include software psychology, userscience, and human-computer interaction.

“B. Shneiderman, “Fighting for the User, ” ASIS Bulletin,December 1982, p. 29.

“H. M. Parsons and G. P. Kearsley, “Human Factors andRobotics: Current Status and Future Prospects, ” HumanResources Research Organization, October 1981.

signers have noted, in fact, that the systemswith the worst human factors seem to be thosewhich were designed to be unmanned, but laterdetermined to need an operator or monitor.Computer-aided design is somewhat differentfrom other PA technologies in this respect.Because of the larger size of the market andthe recent attempts to develop lower cost sys-tems for noncomputer users, CAD designershave begun to pay attention to designing sys-tems that operators can use more easily andproductively.

There are essentially two levels of humanfactors research. The first, sometimes knownas “ergonomics, ” aims to make people morephysically comfortable and productive whileworking at a machine. For example, it includesresearch on the ideal levels of light, color ofdisplay screen, size and configuration of key-board, etc. A second level looks at more fun-damental questions in “human-machine inter-face, ” such as how to distribute control be-tween operator and machine, how to designsoftware for optimum productivity, and howto maximize operator satisfaction. Most suchwork has been directed toward general pur-pose computers or word processors ratherthan programmable automation.4G

These research areas are related to largerquestions in industrial psychology and man-agement concerning less tangible issues suchas the impact of technology on the work envi-ronment and/or on the design of jobs. Therehas been little systematic work in the UnitedStates in these areas, although there is sub-stantial research in some European coun-tries. ”

Ch. 3—Programmable Automation Technologies ● 8 9

Sensors

The vast majority of programmable automa-tion devices are limited in their capabilitiesbecause they are “unaware” of their environ-ment. To use anthropomorphic terms, they donot “know” what they are doing, exactlywhere their parts are, or whether somethingis wrong in the manufacturing process. * Thisproblem is especially acute when manufactur-ers hope to use PA devices to perform tasksnormally performed by people. A minor ad-justment or observation which would be easyand obvious for a human-e. g., righting a partwhich arrived upside-down-is nearly impossi-ble with most current robots.

Hence, computerized devices that can ac-quire information about the environment area lively area of inquiry. While many of thesedevices are used in conjunction with robots,they can also be used with other CAM equip-ment—e.g., NC machine tools or AMH sys-tems–or independently. There are roughlythree categories of applications for sensor sys-tems: 1) inspection, in which parts or productsare examined and evaluated according to pre-established criteria; 2) identification, in whichparts, products or other objects are classifiedfor purposes of sorting or further processing;and 3) guidance and control, in which sensorsprovide feedback to robots or other CAM de-vices on their position and the state of the partor product.

One can simplify the range of sensor tech-nologies by dividing the devices into threeclasses according to their complexity. Whileall of the devices are used for guidance and

4’)There has been substantial work, however, in the design ofs~’stems for teleoperators —remote manipulators controlled bya human operator. Such work is often aimed for ultimate ap-plications in unmanned space missions or underseas, handlingof radioactive material, or battlefield applications, See, for ex-ample, T. N. Sofyanos and T. B. Sheridan, An Assessment ofi~ndersea Telcoperators {Cambridge, Mass.: MIT Sea Grant Col-lege f+-ogram, ,June 1980).

“see, for example, H. H. Rosenbrock, Professor of ControlEngineering, University of Manchester (U. K.) Institute ofScience and Technology, “Robots and People, ” Measurementand Control, March 1982, pp. 105-1 12; P. Brodner and T. Mar-tin, “Introduction of New Technologies into Industrial Produc-tion in F. R. Germany and its Social Effects-Methods, Results,

Lessons Learned, and Future Plans, ” Proceedings of the EighthTriennal World Congress of the International Federation ofAutomatic Control, Kyoto, Japan, Aug. 24-28, 1981, pp.3433-3445; Swedish Work Environment Fund, Programme ofActivities and Budget, 1981/82–1983/84. For further detail seech. 5.

*There are some exceptions where PA devices do have signifi-cant information about their environment. One obvious excep-tion is the Coordinate Measuring Machine, built specifically} tomeasure the dimensions of objects, Another is in factories whichcod each part, for example, with optical codes similar to thoseused on groceries. optical code readers at each machine can iden-tify the part in process. Finally, many PA devices do have in-ternal sensors. For example robots and machine tools have sen-sors which provide feedback on the positions of their joints,

90 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace— — .— —

control applications, usually only the mostcomplex (i.e., vision and touch sensors) canhandle inspection and identification tasks.

The simplest devices provide binary infor-mation—e.g., a weight sensor, photocell, orsimple electrical switch can indicate whethera part is or is not present. These simple sen-sors are relatively cheap, technologically ma-ture and easy to implement. They are alreadyused widely in manufacturing equipment, andtheir use will undoubtedly continue to growfor applications in which binary informationis useful.

At a moderate level of complexity, the infor-mation sensed is analog (continuously vary-ing). For example, a proximity sensor can de-termine the distance of an object. A popularproximity sensor used as a safety device onindustrial robots is the same as the one usedon Polaroid SX-70 cameras. It calculates dis-tance by emitting inaudible sound waves andcalculating how long they take to bounce offthe closest object and return. For safety pur-poses, these can be used to stop the motionof the robot if a human enters its ‘‘work enve-lope. ” Other sensors in this moderate level ofcomplexity include devices which can electro-mechanically sense force and torque—e. g., ina robot arm or a machine tool spindle. Thesecan be used, for example, to allow a robot grip-per to apply just enough force to a delicate ob-ject. Finally, many devices for measurementfall into the moderate-complexity category.Optical sensors, for example, can be used tomonitor the diameter of a driveshaft on alathe, or for noncontact sensing of the dimen-sions of hot metal as it emerges from forgingprocesses.

Most of these moderately complex sensingtechniques are fairly well-developed, and canbe applied relatively easily albeit with somecustomization. There is a moderate amount ofIt&D under way to increase the quality of in-formation from these devices (e.g., their sen-sitivity and speed), and to increase their rangeof applicability (e.g., development of sensorsfor measuring arbitrary prismatic shapes onmachine tools). In addition, the coordination

of these sensors with each other and withCAM tools is a very difficult problem. Com-puter scientists are attempting to developprocessing techniques that can quickly inter-pret force and torque data from the variousjoints of a robot, for example, and providefeedback to the robot’s controller.

At the most sophisticated end of the sens-ing techniques, visual and touch sensors dealwith information that is not only analog butalso needs substantial processing to be useful.Vision and touch sensing technologies are onlyin their infancy, and have just begun to havepractical uses in the factory. Of the two, vi-sion by far receives the most attention.

The chief technical problem in machinevision systems is in interpreting the picturesgenerated by a video camera. In a typicalvision system, a frame—i.e., one complete im-age frozen in time— is typically composed of256 by 256 picture elements, or pixels. If eachpixel is either black or white, then there aremore than 65,000 bits of information that thecomputer program must process. In general,

steps in the process include:48

Segmentation.—The areas of the imagemust be clarified and divided into seg-ments or “blobs, ” representing the fea-tures of the object and its background.There are two general schemes for begin-ning the interpretation of’ the data. Onemakes use of discontinuities-prirnarilydetecting the edges of the object in theimage. The other approach relies on simi-larities in the image, i.e., areas of the im-age that are of similar intensity.Recognition.–The system must comparethe features (segments) it has identifiedwith those stored in its memory, attemptto find an object in its memory that issuitably close to the one in the image, andhence label the object and its features.Interpretation. –This step varies depend-ing on the machine vision application. For

,,. N!, \\”right, “Vision of the Future, ” unpublished manu-s(’ript, Carnegie-lvlellon University Robotics 1 nstitute, Januar?’1 9H:;.

Ch. 3—Programrnable Automation Technologies 91

robot guidance, the interpretation stepmight be to identify the object, then cal-culate its coordinates so that the robotcan grasp it. For an inspection applica-tion, the interpretation of the imagemight be to determine whether the objecthas the right dimensions or is free ofdefects.

In the vast majority of current vision sys-tems, each picture element in the 256-by-256element image is either black or white. In moreadvanced systems just beginning to be usedin industry, each pixel can be one of severalshades of gray. These systems, often called‘‘gray-scale, are potentially more powerful intheir ability to identify objects and cope withuneven lighting, but they also require muchmore computer power and algorithms for proc-essing data which are only beginning to beworked out. Systems for processing color im-ages are another order of magnitude more

complex, and there is no active work on suchsystems yet.

The systems described above essentiallyprovide 2-D information, although certaintricks can be used to infer the 3-D character-istics of an object. Some researchers have usedmore than one camera in order to obtain 3-Dinformation much as the human eye does,though such schemes are in very early stages.One very promising method to obtain 3-D in-formation is the use of “light striping” sys-tems. In such a system, a laser or other lightsource flashes a very precise band of light ontoan object, and the camera records the imageat that instant. By examining how such struc-tured light bends over a 3-D object, the sys-tem can infer the dimensions and distance ofthe object.

Current machine vision is in a very earlystage. The range of objects that can be iden-

Photo credit National Bureau of Standards

“Light striping” system can determine the shape of a 3-D part by flashing a very precise line of light (from slots on right,below gripper) and photographing how that light is bent by the object. TV monitor shows view of camera above gripper

92 Computerized Manufacturing Automation: Employment, Education, and the Workplace

tified, the speed of the interpretation, and thesusceptibility of the systems to lighting prob-lems and minor variations in texture of objectsare all examples of serious problems with cur-rent technology. Successful applications ofcurrent machine vision technology tend to bevery specific, ad hoc solutions, often usingclever “tricks” or manipulations of the manu-facturing environment. As one report notes,“The vision systems of today, and those forthe rest of the decade will not promise greatgenerality. These sorts of tricks will be an im-portant part of the field for many years tocome. ’49 Nevertheless, many useful applica-tions are possible with existing technology andmachine vision is currently a rapidly growingfield. In certain specific applications, especial-ly very tedious tasks such as inspection of elec-tronic circuit boards, machine vision systemscan outperform humans.

An example of a successful machine visionapplication is shown in figure 13. Here, a sys-tem developed by Octek Corp. counts stacksof cups prior to packaging. The system first“grabs” an image from its camera under con-trolled lighting conditions, defines the edgesof the cups, attempts to eliminate shadowsand other confusing data, and counts thenumber of cup lips. Similar programs havebeen developed to inspect cassette tapes andcircuit boards.50

While there is considerably less research ef-fort under way on touch sensors, there are sev-eral groups of researchers, for example, work-ing on a touch sensor based on a carbon-impregnated rubber pad which changes itselectrical conductivity under pressure. Thispad could be attached to a robot gripper, andit would send to the computer processor an im-age or “footprint” of the object being grasped.Once this image has been obtained, inter-preting it involves virtually the identical proc-ess used for vision processing.

“Ibid.5[’D. L. Hudson and J. D. Trombly, “Developing Industrial

Applications for Machine Vision, ” Computers in Mechm”calEngineering, vol. 1, April 1983, pp. 18-23. Note that this ap-plication of machine vision, like many inspection applications,is not used in connection with a robot.

Figure 13.— Machine Vision Process

Two steps in a machine vision process developed to count stacksof paper cups for a cup manufacturer. The cups are lit by a highlydirectional fiber optics light source, which makes their edges standout. The top photo shows part of the system’s segmentation process,in which it assesses the intensity of light for each cup lip. The bottomphoto indicates the interpretation function, in which the system hascounted the number of cups in the stack. Note the shadows andrelative unevenness of light which complicate even this simplemachine vision application.

SOURCE: Octek Corp.

There seem to be two schools of thought onsensors for industrial robots. One argues that,if enough care is taken in organizing the man-ufacturing environment, complex sensors areunnecessary. Parts can be carefully fixturedso they are in the proper orientation and posi-tion, and sensors in the simple or moderatelevels of complexity will suffice. The otherpoint of view is that robots should be able toadapt to the chaos of manufacturing, andshould ideally have senses—vision, in partic-ular-which rival those of a human.

Materials Trends

Plastics, ceramics, and composites are re-placing metals in a wide variety of products

-4-- —Ch. 3—Programmable Automation Technologies ● 9 3

— — . — . —

These trends and others mean that theamount of metal-removing activity is going todecrease. Thus, there is some chance that useof plastics and ceramics will eventually renderobsolete some new metalcutting equipment.This possibility has not yet been examinedsystematically by the metalworking communi-ty. Robots, because of their flexibility, are lesslikely to be affected than machine tools. How-ever, there are certain factors that tend tomake widespread obsolescence of new metal-working equipment unlikely. First, metalwork-ing machines have useful lives of 30 years ormore, and the users of this equipment movenotoriously slowly in replacing machine tools.As new materials technologies do reduce theamount of metalworking, it is the vast stockof older, manual machine tools that is likelyto be useless rather than the newer equipment.Second, it is expected to take at least two dec-ades for ceramics to displace a significantamount of metalworking.

Future of the Technologies

Capabilities

Building on the “Trends and Barriers” sec-tion of this chapter, tables 11 through 15 sum-marize the main problems for PA technologiesand the projected times for solution. Thoughthese projections must be considered extreme-ly tentative, they provide a sense of the rela-tive scale and complexity of the problems.

Because the amount of time between labora-tory solution of a problem, first prototype ap-plications, and the widespread and easy availa-bility of that solution is significant, the tablesinclude a separate estimation of each. Projec-tions for applications and availability are evenrougher than the projection for a technical so-lution, since they depend on many social, eco-nomic, and market conditions.

These projections were compiled by analysisof existing sets of projections51 and by inter-views with technology experts. Projections oftechnological developments are inherently con-troversial, and experts do not always agree.Some experts will view these estimates aseither too optimistic or too pessimistic. Dur-ing the interviews with technology specialists,for example, several pointed out that some ofthe “key problems” listed in the table may

“See, for example, Manufacturing Studies Board, NationalResearch Council, Applications of Robotics and Artificial In-telligence to Reduce Risk and Improve Effectiveness: A Study-for the United States Army (Washington, D. C.: NationalAcademy Press, 1983); R. E. Garrett and R. M. Mueller, “Stra-tegic Analysis/Technology Trend Report, ” Control Data Corp.,May 15, 1981; D. Grossman and J. T. Schwartz, “The NextGeneration of Robots, ” in Frontiers in Science and Technology(Washington, D. C.: National Academy of Sciences, 1983, pp.185-209.

94 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace

Table 11 .—CAD: Projections for Solution of Key Problems

Current (1984)Hardware:1. High-resolution, color display of designs,

with rapid generation of imagesa . . . . . . . . A

Both hardware and software:2. Low-cost, powerful microcomputer-based

workstations forb

a) electronics design . . . . . . . . . . . Ab) mechanical design . . . . . . . . . . . . . . . .

3. Independent CAD workstations linkedby network, with access to super-computer for powerful analysis andsimulation . . . . . . . . . . . . . . . . . . . . A

Software:4. Three-dimensional solid modeling

systems, resulting in:c

a) more realistic images . . . . . . . . . . . A

b) enhanced ability to connect withmanufacturing equipment . . . . . . A

5. Comprehensive, powerful computer-aidedengineering systemsd for mechanicaldesign . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6. Extensive design/manufacturingintegration . . . . . . . . . . . . . .—

1985-86

●

●

A

●

A

a - 1987-90

■

●

■

■

●

A

●

1991-2000 2001 and beyond—

awhl Ie color displays are currently aval [able, they tend to sacnflce etther resolut Ion (the fineness and clarlty of the picture) or the speed WI th which the lmaOeS canappear on the screen New techniques for displays, such as the use of dedicated microprocessor chips (sometimes termed “slllcon engines”) to generate ~mages,promise to Improve this situation.

bMICrOCornputer.baSed Workstations for CAD are now being marketed, but In the judgment of techntcal experts consulted by OTA, they are etther not powerful enou9h

andlor not Inexpensive enough to be useful In a w!de variety of applicationsCCAD expefls repo~ that many systems for 3-D SOI id model i ng are avai Iable now, but they are not being used because of their lar9e aPPetlte for com Puter Power, and

because their capacity to link design data to manufacturing equ!pment is inadequate Part (b) of this entv refers to this abillty to store and manipulate design dataabout the physical characteristics of a parl i n such a way that it can be transmitted to manufacture ng equi Pment with only minimal i ntermed!ate steps

dTh ls entw refers t. modules powerful enough t. allow extensive interactive testing, simulation and refinement of designs In a wide range of appllCatlons Such SyStemS

are strongly product-dependent, whi Ie they may be near avai Iable for certatn products now (e g , integrated circ u its, certal n port!ons of al rcraft and motor veh Icles),they are much less advanced In other industries and applications

‘This entry denotes the “window from design to production” wh!ch would, for Instance, allow designers to examine the production implications of design choicesThese Include the costs and necessary production processes, as well as the history of manufacturing slmllar Items at the plant Such comprehensive connectionswould allow much more substantial Integration of CAD, CAM, and computer. based management

A Solutlon in laboratories● first commercial applications9 solution widely and easily available (requlrtng minimal custom engineering for each appi!catlon)

SOURCE OTA andysls and compilation of data from technology experts

never be solved at all-the development ofstandard languages for robots (table 12 item10) for example, depends as much on marketfactors and political considerations among ro-bot vendors and users as it does on technicalissues. Similarly, development of artificialintelligencebased systems which could controlmuch of a factory without human intervention(table 15, item 5) depends on fundamental ad-vances in the field of artificial intelligence thatare by no means assured.

On the other hand, it is possible that someof the advances in the accompanying tablesmay occur significantly faster than the tablesindicate. This might be particularly likely if

the Federal Government and/or industry wereto choose to make dramatic increases in R&Dfunding for PA technologies. Chapter 8 dis-cusses R&D funding in more detail.

However, industry observers report withvirtual unanimity that the application of pro-grammable automation in most industries islagging significantly behind the technologies’development, and that there appear to beabundant, relatively easy opportunities for useof current automation technologies. Hence themain stumbling blocks in the near future forimplementation of PA technologies are nottechnical, but rather are barriers of cost, or-ganization of the factory, availability of ap-

Ch. 3—Programmable Automation Technologies ● 9 5—-——

Table 12.— Robotics: Projections for Solution of Key Problems

Current (1984)

Hardware:1. Lightweight, composite structures and

new forms of drive mechanisms . . . . . A

Both hardware and software:2. Force sensors . . . . . . . . . . . . . . . . . . . . . . . . . A s

3. Versatile touch sensors ... . . . . . . . . . .4. Coordinated multiple arms . . . . . . . . . . . . . .5. Flexible, versatile grippers . . . . . . . . . . . . .

Software:6. Precise path planning, simulation and

control with CAD . . . . . . . . . . . . . . . . . . . . . . A

7.3-D vision in structured environmentswhich have been planned to simplify thevision task . . . . . . . . . . . . . . . . . . . . . . . . . . . A

8.3-D vision in unstructured complexenvironments which have not beenplanned to simplify the vision task ... . . .

9. Robust mobility in unstructuredenvironments . . . . . . . . . . . . . . . . . . . . . . . . .

10. Standards clarifying different versionsof robot languages, and helping ensurea common language for similarapplications. . . . . . . , . . . . . . . . . . . . . . . . . . .

ò= Solution in laboratories. = f i rst commercial appl icat ions

1985-86

●

A

A

1987-90

■

■

●

●

A

●

■

A

A

1991-2000

solution widely and easily available (requiring minimal custom engineering for each application)

SOURCE OTA analysis and compilation of data from technology experts

Table 13.—NC Machine Tools: Projections for Solution of Key Problems

Current (1984)

Hardware:1. Systems which can automatically and

reliably remove a wide variety of metalchips produced in cuttinga . . . . . . . . . . . .

Both hardware and software:2. Reliable, widely applicable adaptive

control to optimize speed of metalremoval . . . . . . . . . . . . . . . . . . . . . . . . . . . . A

3. Tool wear sensors applicable to widerange of cutting tools . . . . . . . . . . . . . . . A

4. Systems for measurement of parts of avariety of shapes and sizes while theparts are being machined . . . . . . . . . . . . . . .

Soft ware:5. Controllers to accommodate ties to robots A

6. Model-based machining in which themachine tool operates substantiallyautomatically based on data aboutmetal processes and the part to beproduced . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7, Widely applicable 3-D verification of NCprograms using CAD-based simulations . .

1985-86

●

A

●

A

A

1987-90

A

●

●

●

1991-2000

2001 and beyond

■

■

� ✍✍ ✎� �

2001 and beyond

■

■

asy~tems currently exl~t for automatic removal of metal chips, but despite much Interest and research, they are neither very reliable nor QenerlCally applicable (1 e ,

they can only be used for certain kinds of metals or cutting processes)A = solutlon In laboratories.= first commercial applications~= solutfon widely and eas!ly available (requiring minimal custom engineering for each application)

SOURCE OTA analys!s and compilation of data from technology experts

● Computerized Manufacturing Automation: Employment, Education, and the Workplace—

Table 14.—FMS: Projections for Solution of Key Problems

Current (1984)

Hardware:1. Generic fixtures for holding a variety of

work-in-process parts . . . . . . . . . . . . . .

Both hardware and software:2. FMS for:a

a) cylindrical parts production . . . . . . . . A O

b) sheet metal parts production . . . . . . . A ,c) 3-D mechanical assembly . . . . . . A

d) electronics assembly. . . . . . . . . . . . . A .

3. Materials handling systems which canhandle a variety of parts in any sequencenecessary . . . . . . . . . . . . . . . . . . . . . . . . . . A O

Software:4. Automatic diagnosis of breakdowns

in the FMS. . . . . . . . . . . . . . . . . . . . . . . . . . . .5. Standardization of software

interfaces between computerizeddevices in an FMS . . . . . . . . . . . . . . . . . . . . .

1985-86

A

1987-90

A

A

1991-2000

● 9

●

●

2001 and beyond

aAlmost all FMS currently running are used to machine t!c Darts, (e g , enaine blocks) which are those whose outer shaoe consists Drlmarilv of flat surfacesThe projections m this entry ref~r to FMS for quite different appiications~a) m~chining of’cylindrical parts, such as rotors and driveshaf~s (or “parts of rotation, ”I n machlnl ng Jargon, since they are generally made on lathes), b) stamping and bending of sheet metal parts, such as car body panels, c) assembly (as opposed tofabrication of indiwdual parts) of three-dimensional products, such as motors, and d) assembly of electronic devices, such as circuit boards. While machines currentlyexist for automatic Insert Ion of electronic parts into c!rcu!t boards, an electronics FMS would integrate the insertion devices with soldering and testing equipment.

A – solution In laboratories● = first commercial appllcattons,■ solutlon wdely and eas!ly available (requiring mlnlmal custom engineering for each application)

SOURCE OTA analysis and compilation of data from technology experts

Table 15.—CIM: Projections for Solution of Key Problems

Current (1984)

Software:1.

2.

3.

4.

5.

Well-understood, widely applicabletechniques for scheduling and logisticsof complex materials handling systemsthat would allow full factory integration . .Standard communication systems(networks) . . . . . . . . . . . . . . . . . . . . . . . . . . . . A

Standardization of interfaces betweenwide range of computerized devicesin an integrated factory. . . . . . . . . . . . . . . . .Data base management systems whichcould sort, maintain and update all datain a factory . . . . . . . . . . . . . . . . . . . . . . . . . . .Computerized factories which couldrun on a day-to-day basis with only afew humans in management, designfunctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

A = solution in laboratories.● = first commercial applications

1985-86

●

1987-90

A

A

1991-2000 2001 and beyond

■

AI

■ = solution widely and’ easily available (requiring minimal custom engineering for each application).

SOURCE OTA analysis and compilation of data from technology experts

Ch. 3—Programmable Automation Technologies ● 9 7—

propriate skills, and social effects of thesetechnologies. These issues are more fully ad-dressed in later chapters of this report.

Many in industry would argue that CIM isinevitably the future of manufacturing. Its ad-vantages in cost, quality, flexibility, and con-trol will, they assert, mandate its adoption.Many parts of computer-integrated manufac-turing can be put together on an ad hoc basisnow, and as the tables show, prototype solu-tions for many of the key problems already ex-ist. However, several key aspects of the puz-zle are as yet unsolved (the development of in-terface standards for computerized tools, inparticular), and for CIM to be practical eachof its elements must be mature, versatile, andrelatively easily available commercially.

As noted earlier in this chapter, CIM doesnot necessarily imply manufacturing withouthumans. In fact, one of the biggest challengeson the road to CIM is learning to use humansin effective ways, to develop machines withwhich humans can work effectively, and toidentify the points in the production processwhere maintaining human involvement mayenhance flexibility, responsiveness, diagnosticpower, and creativity. The extent to whichthat effective use of people in manufacturingwill develop, and the extent to which CIM willremove humans from manufacturing environ-ments, are still open questions.

Automation technology researchers reportprogress on virtually all of the technical prob-lems, although the degree of progress oftendepends on research funding, commercial de-mand for related products, and inclinations ofresearchers. The technical barriers to increasedsophistication in programmable automationare largely due to the complexity of the man-ufacturing environment, and to the fact thatmany manufacturing processes—e.g., machin-ing, scheduling, design-have not been clear-ly understood in a way that can easily becomputerized.

These projections of future technical capa-bilities, along with various other projections,imply that the remainder of the 1980’s will bea time when applications will to some extent

catch up with developed PA tools. Some tech-nical improvements will doubtless be madeduring this time. But most prognosticatorsseem to see the 1990’s as a period when theapplication of basic PA tools will become widespread, and a number of major technical im-provements will be available, particularly forrobots and FMSs (see tables 12 and 14). Whilethis may, in some cases, suggest that the1990’s seem far enough away to solve almostany technical problem, it also seems to in-dicate that the next decade will bring quitesubstantial increases in the power and poten-tial uses of programmable automation.

Future Levels of Use ofProgrammable Automation

The rate of growth in use of programmableautomation in the United States, known as the“diffusion” of the technologies, depends onfactors both in the larger economy and at thelevel of individual firms and products. Someof the more general factors include availabil-ity of capital and skilled labor, internationalcompetition, and the amount of attentionAmerican firms devote to improvements inmanufacturing processes. The last factor maybe the most critical. Manufacturing engineer-ing in the United States has been largely neg-lected both in engineering schools and inindustry .52

Prompted in part by international competi-tion, however, the mood among American in-dustrialists (to the extent there is a “mood”in such a diverse group) seems to be chang-ing. Increasingly, established managementpractices are being questioned in conferencesand industry journals, and many industrialmanagers are closely examining improve-ments in manufacturing processes, particular-ly robots.* The extent to which this change

“See, for example, E. N. Berg, “Manufacturing’s AcademicRenaissance, ” The New York Times, Oct. 30, 1983.

*DeSpite the generally rising interest in robotics, a signifi-cant group of people remain powerfully skeptical about the useof robots as one of the primary steps to enhance productivity.In a 1983 survey by the Institute of Industrial Engineers, forexample, members of the society-who are closely involved withmanufacturing processes on a day-today basis—rated roboticsrelatively low in effectiveness of a group of productivity-

98 ● Computerized Manufacturing Automation: Employment, Education, and the Workplace

in mood will effect lasting and significantchange in manufacturing, however, is uncer-tain. Many management specialists believethat such lasting change must include discard-ing powerfully entrenched habits in industry,particularly financially-oriented managementstrategies that discourage risk-taking anddownplay quality relative to cost.53

In addition to these more general questions,a large number of factors come into play whenan individual firm chooses to use or not to useprogrammable automation. Some of the tech-nical factors include: the applicability of thetechnology to the problem at hand, whichtends to vary according to the particular man-ufacturing processes used in each factory; therange of tasks to which a given tool can be ap-

—

enhancing methods. Only 29 percent had undertaken increas-ed use of robotics in the past 5 years; 22 percent rated the stephigh in effectiveness, 48 percent moderate, and 26 percent low.Measures which received higher effectiveness ratings includ-ed capital investment for new or automated machinery general-ly, worker training, improvement of inventory control, and“systems innovations. ” The interpretation of these surveyresults could differ; some would argue that increased familiarityof industrial engineers with robotics will lead to higher percep-tions of effectiveness. In addition, robotics was not viewed asunproductive by very many respondents; it simply appearednot to be the productivity tool of first choice. (See “ProductivityToday: An Inside Report, ” The Institute of IndustrialEngineers, Norcross, Ga.)

53See, for example, R. H. Hayes and W. J. Abernathy, 4’Man-aging our way to econonic decline, ” Harvard Business Review,tJuly-August, 1980, pp. 67-77.

plied; the cost of customization, particularlyfor new technologies where few standards ex-ist and almost every application is a proto-type; the ease of use of the tool; the reliabili-ty of the equipment; the compatibility of pro-grammable automation with machines alreadyin place; and finally, the capacity of differentPA systems for upgrading and expansion.

Organizational factors can also have a sig-nificant effect on firms’ automation decisions.For example, one researcher found thatprevious experience with automation was akey factor in successful applications,54 and in-dustry observers report that many unsuccess-ful attempts to use programmable automationhave been due to premature jumps into com-plex systems. There can also be substantialresistance to change on the part of workers ormanagement. Many manufacturers report,however, that production workers tend toaccept technological changes such as automa-tion, while strong resistance tends to comefrom middle managers who fear program-mable automation will diminish their degreeof control or eliminate their jobs.55 Chapter 5discusses organizational issues of PA imple-mentation in more detail.

“J. Fleck, “The Adoption of Robots, ” PmcA”ngs of the 13thInternational) Symposium on Industrial Robots and Robots 7,Apr. 17-21, 1983.

MOTA Automation Technology Workshop, May 29, 1983.