Gerovitch Automation

122

AUTOMATION

Planware has several types of kno wledg e, all encoded throu gh param eterize d heories. The first s knowledge of the sc hedul ing domai n, including the constraints on use of th e different types of resources, such as reusable or sharable resources. Another type of knowledge is algorithm nowledge, such as generate-and-test, branch -and-bo und, divide-and-conquer, dynamic programmi ng, and hill-climbing (see ALGORITHMS,DESIGN

Planware,http://www.kestrel.edu/HTML/projects/ arpa-plan2/. SciNapse. http://www.scicomp.com/about/ technology.htm1. Michael R. Lowry

AUTOMATION

AN D

CLASSIFICATION F) . By codifying them as pa rameterized theori es, algorithms can be automatically derived or a givenvery-high-level problem specification, given appropriate domain axioms.A third type of knowledge s implementation knowledge, which defines how higher-level onstructs su ch as sets can be encoded as more implementation-level constructs such as lists or bit-vectors.

All of these tools use advanc ed knowledge representation and automated reasoning capabilities. Although research tools today, they represent thedegree of programming automation that may become commercially available within a decade.

Bibliogvaphy 1956. Newell, A , , and Simon, H . A.“The Logic Theory Machine,” IR E Transactions on Information Theory, IT-2, 3 (March), 61-79. 1963. Simon, H . A. “Experiments with a Heuristic Compiler,” Journa l of the ACM, 10, 493-506. 1982. Martin, J . Application Development without Programmers. Upper Saddle River, NJ: Prentice Hall. 1986. Green, C., Luckham, D ., Balzer, R. , Cheatham, T., and Rich, C. “ Repor t on a Knowledge-based Software Ass istant,” in Readings in Artificial Intelligence and Software Engineering (eds. C. Rich and R. C. Waters ). San Francisco: Morgan Kaufmann. 1990. Rich, C., andWaters, R. C. The Programmer’s Apprentice. Reading, MA: Addison-Wesley. 1991. Lowry, M. R., and McCartney, R. D. (eds.) Automating Software Design. Cambrid ge, MA:MIT Press. 1993. Kant,

E.

Synthes is of Mathematical Modeling Soft ware ,” 10 , 3 (May), 30-41.

IEEE Software,

1995. Flener, P. Logi c Program Synthesis f ro m Incomplete Information. Boston: Kluwer Academic Publish ers. 1996. Smith, D. R., Parra, E. A , , and Westfold, S. J. “Synthesis of Planning an d Scheduling Softw are,” in Advanced Planning Technology (ed. A . Tate) , 226-234. Menlo Park, CA:AAAI Press. 1997. Browne, T . , Davila, D., Rugaber, S., and Stirewalt, K. “Using Declarative Descriptions to Model User Interfaces with MASTERMIND,” in Formal Methods in Hu ma n Computer Interaction (eds. F. Paterno and P. Palanque). New York: Springer-Verlag. 1998. Bibel, W., and Schmitt, P. Automated Deduction. A Basis for A pplicati ons. Boston: Kluwer Academic Press.

Automation

is the c onversion of a work process, a procedure, or equipment to automatic rather than human operation or control. Automation doesnot simply transfer hum an functions to machines, but involves a deep reorganization of the workprocess, during which both the human ndhemachine functions are redefined. Early automation relied on mechanical and electromechanical control devices; during the last 4 0 years, however, the computer gradually became the leading vehicle of automa tion. Modern a utomat ion is usually associated with computerization. This article examines the major phase s of historical develo pment and social and econo mic aspects of industrial automation, focusing on the computerization of production, engineering, and managerial asks. Other areas of compute r-basedautomation include administrative applications (4 . v.),communication via electronic mail (q.v.), anking applications, medical applications (q.v.), nd library automation (see DIGITAL LIBRARIES).

Phase I: Mechanization and Rationalization of Labor The izatio n of machine tools for gan mechan during the Industrial Revolution atproduction the endof bethe 18th century with the introduction of the Watt steam engine, the Jacqu ard loom, the lathe, and the screw machine. Mechanization replaced human or animal power with machine power; those mechanisms, however, were not automatic but controlled by factory workers. The factory system, with its large-volume, standardized production, and division of labor, replaced the old work organization, where broadly skilled craftsmen and artisans prod uced small quantities of diverse products. In he late 19th century Frederick W. Taylor rationalized th e factory system by introducing the principles of “scientific manag ement. Heviewed the body of eac h worke r as amachin e ”

Websites Amphion,http://ic-www.arc.nasa.gov/ic/projects/ amphion/. Mastermind. http://ww.cc.gatech.edu/gvu/ user-interfaces/Mastermind/.

whose movements had o beoptimized n order o minimize time required to compl ete eachask and thus increase overall productivity. “Scientificmanagement” strictly separated mental work frommanual labor:

AUTOMATION

workerswerenot o think but o follow detailed instructionsprepared for them by managers. The rationalized factory system gave birth to a new man agerial class and large clerical bureaucracies. The Taylorist principles served as a basis for Henry Ford’s system of mass prod uction . In 1913 the Ford Motor Company introduce d a moving assembly line, drastically cutting assembly time. The assembly ine imposed a strict order onproduction by forcing work ers tokeep pacewith the motion of the conveyor belt. Mass production relie d on he standardization of components and final products and routinization of manufacturing a nd assembly jobs. The Ford assembly line became a symbol of efficiency of American manufacturing; for workersand social critics, however, it epitomized the monotony and relentless pressure of mechanized work.

Phase

II:

Automation

of Production

In 1947 the Ford Company brought the term “automation” into wide circulation by establishing the first Automation Department,harged with designing electromechanical, hydraulic, andpneumaticpartshandling, work-feeding, and work-removing mechanisms to connect standalone machines and ncrease the rate of produc tion. In 1950 Fordput into operation he first “automated” engine plant. Although early automation was “hard,” or fixed in the hardw are, and did not involve automa tic feedback control, this conce pt provoked great public enthusiasm for “u nmanne d factories” controlled by “buttons that push hemselves, as well as causing growingconcern about the rospects ”

of mass unemployment. To meet U S Air Force demands for a high-performance fighter aircraft whosecomplex structuralmembers could not be manufacture d by traditional machining metho ds, a technology of Numerical Control (NC) of machin e tools was developed in the early 1950s. N C laid foundation for programmable, or “soft,” automation, in which the sequence of processing operations was not fixed ut could be changed or each ew product style. Commercial NC machines for batch production appeared in the mid-1950s. Designed to military specifications, early N C equipment proved too complex and therefore unreliable, as well as prohibitively expensive, and was applied mostly in th e state-subsidized aircraft industry. The abstract, formal approach of N C , based on mathematical modeling of the machining proce ss, superseded the record-playback techniqu e of direct machin e imitation of workers’ a ctions. While the record-playback approach relied on the skill and discretion of the worker, NC technologyallowed engineers and man agers to exercise greate r control over the production process.

Phase (CAM)

Ill:

12 3

Computer- Aided Manufacturing

The first industrial applications of dig ital computers occurred in t he electrical power, dairy, chemic al, and petroleum refinery industries for automatic process control. In 1959, TRW installed the first digital computer designed specifically for plant process control at Texaco’s Port Arthur refinery. Early applications were open-loop control systems: gathering data from measuring devices and sensors throughout the plant, the compu ters monitored technological processes, performed calculations, and printed out “operat or guides ”; subsequent adjustments were made by human operators. In the 1960s closed-loopfeedback contro l systems appeared. These computers were connected irectly to servo-control valves and made adjustments autom atically (see CYBERNETICS). In he ate 1960s, with the development of time sharing (4.v.) on large mainframe computers (q.~.), standalone N C machines were broug ht under Direct Numerical Control (DNC)of a central computer.DNC systems proved vulnerable to frequent failures due to malfunctioning of the ce ntral compu ter an d the nterference of factory power cables with t he data tran smission cables of the D N C system. With the in trodu ction of microprocessors (q.v.) n the 1970s, centralized D N C systems in manufacturing were largely replaced by C ompu ter Numerical Control (CNC) systems with distributed control, in which each N C machin e was controlled by its own microcom pute r. This blending of information and produ ction technologies produced a new breed of machinistprogrammer who could operate C N C equipment by generating and debugging N C programs, thus breaking down he traditional distincti on between whitecollar and blue-collar jobs. Robotics combined the techniques of N C and remote control to replace human work ers with numerically controlled mechanical manipulators. The first commercial robots appea red in th e early 1960s.Robots proved very efficient in performing specialized tasks that demanded high precision or had to be done in hazardous environments. To approa ch the humanevel of flexibility, robots were supplied with sophisticated techniques of fee dback , vision an d tactile sensors, reasoning capabilities, and adaptive control. In the 1980s industrial app lications of robots slowed down, as their increasing complexity resulted in growing costs and insufficient reliability. Hierarchical NumericalControlSystems ombined DNC and CN C features: they linked each standalone computer controller to a central co mputer that maintained a large library of C N C programs and monitored

12 4

AUTOMATION

production. This approach aspired to replace the human operato r’s xpertise by engineering knowledge formalized inC N C programs. In such systems, human operators generally no longer programmed C N C equipmenton heshop floor, andproductionwas brought under remote upervision of a central manage ment-controlled compute r. Flexible Manufacturi ng ystem s (FMS) combin ed D N C equipment with machinesor automated loading, unloading, and transfer of workpiece s. These systems permitted varying process routes andsequences of operations, allowing automa tic machin ing of different products in small batche s in the same system. Centralized FMS have often prove d too compl ex, howev er, and they are increasingly ubdivided into smaller flexible manufacturin g cells (FMC) ha t include several C N C machines, robots, and transfer devices controlled by a single computer, the “cell control ler.”

Phase IV : Automated

Engine eri ng

In the 1960s large aerospace manufacturers, such as

processed hug e amo unt s of dat a genera ted in mass production and mass marketing, became rimary a target of a utomat ion a nd job reduction in the 1960s and 1970s. By 1970 the profession of bookkeeperwas almost completely eliminated in the USA. In the mid1960s the first manag ement-inform ation systems (MIS) appeare d, providing management withdat a, models of analysis, and algorithms for decision-making; eventually they became a standard tool for operation control, management control, and strategic planning.

Phase VI: Computer-Integrated Manufact uri ng (CIM ) In the late 1980s an integration of the automa ted factory and the electronic office (4.v.) began. CIM combines flexible automa tion robot s, numerically controlled machines, and flexible manufacturing systems), CAD/ CA M systems, and management-information systems to build integrated production systems that cover the complete operations of a manufac turingirm, including purchasing, logistics, maintenance, engineering, and business operations. CIM emphasizes horizontal links

McDonnell-Douglas and(CAD) Boeing, developed proprietary computer-aided design systems, which provided computer graphics (q.v.) tools for drafting, analyzing, and modifying aircraft design s. In 1970 Comp uterVision Corporation introduced the irst complete turn key commercial CA D system for industrial designers, which provided all the necessary har dwa re and software in one package. In the 1970s, combined CAD/ CA M systems emerged which used the parameters of a geometrical model created with the help of CAD to generate programs for C N C machin e tools and develop manufacturing plans and schedules. While CA D systems are often packaged and standardized, CA M (Computer-Aided Manufacturing) ppli cations tend to be industry-specificand proprietary. With the introduc-

betwee n different organizational units of a firm and provides the possibility of sharing dat a and compu ting resources, making it possible to break the traditional institutional barriers between departments and create flexible functional group s to performasks more speedily and efficiently.

tion of Compu ter-Ai ded Engine eringCAE) systems for standard techniques of en gineering analysis, the whole range of engineering tasks-from concep tual design to analysis to detailed design to drafting and documentation to manufacturin g esign-became aut omat ed. The distinction between blue-collar and white-collar jobs was further blurred, as engineers, clerks, and managers became integrated in an automat edoffice.

pessimists view machin es as instru ments of subjugation and control by a ruling elite, argue that automation leads to the degradation of human beings, and depict the future as a grim technological dystopia. Both sidesview autom atic technologyas an autonomous force determ ining he direction of huma n history. Automation itself, however, is a social process shaped by various social and econom ic forces. This process may take various directions and mayhavediverse consequences depending on hesocioeconomic and organizational choices made during automation.

Phase V: Automate d Management Among the earliest applications of i nform ation technology was the automatio n of information-pro cessing tasks. The first stored -prog ram digital computer purchase d by a nongo vernme nt custo mer was UNIVAC (q.v.), installed by G E in 1954 to automat e basic trans action processing: payroll, inventory control and mate rial scheduling, billing and order service, and general

cost accounting. Large clerical bureaucracies, which

Social and Economic Dimensions of Automation Views of automa tion range betwee n two extremesunabashed optimism and utmost pessimism. The optimistsbelieve n a technological utopia, an imagined bright future in which machin es will relieve people of all hard work and bring prosperity to humank ind. The

The Productivity Par adox While productivity in major industries in the USA rose sharply during production autom ation in the 1950s and 60s, its grow th has slowed significantly since the 1970s, precisely at the time of widesprea d computerization of the factory and the office. The link between

AUTOMATION

125

computerization and productivity remainsproblematic. The advantages mostommonly associated with computer-aided manufacturing nclude increased production rates, better product quality, more efficient use of materials, sho rte r lead times, reduced work hours, and improv ed work safety-all factors leading to higher productivity. Among its main disadvantages, analysts usually cite the high cost of designing, build-

machin ery to the use of st andard design and analysis procedures that tell the computer how to design and build a needed par t. Man agemen t evolved from direct supervision of labor to “man age ment by number s,” based on numerical ata reports and pre-programmed compu ter alg orithms for decision-making. When operators must step in and take control in case of an emergency at an automatically controlled nuclear

ing, and maintaining computerized equipment;ulnerability to down time; elatively low flexibilityompare d with humans; and work er displacement and emotional stress-all leading to lower productivity. It is particularly difficult to compa re directly productivity before and after co mputerization, since it brings with it not merely technological, but also organizational change which transforms the entire nature o f production and brings with it the most benefits and losses.

power plant, would they possess the necessary skills if their training and daily experience mainly concerned work with a computerized ontrol system?

As manufacturers whontroduced omputer-aided manufacturing systems ffirm, the largest payoff from computerization comesnotfrom speeding up old operations but rommaking work organization more flexible and efficient. On theotherhand, if compu ters are used to conserve old inefficient organization, comp uteriz ation can only accelerate negative trends . As John Bessant has remarke d, “When ou put a compu ter nto a chaotic factory the only thing you get is computerized chaos” (quoted n Ayres, 199 1-1 992, Vol. 4 , p. 94). Most successful manufacture rs streamline operations before computerization, following the dictum, “Simplify, the n auto mate !” Efficient compu terization takes far more han merelynstalling a computer: it requires changes in the entire workstyle.

Worker Displacemen t, Skill, and Working Conditions A leading concern among workers, labor leaders, and

social critics has been the issue of worke r displacement-a loss of wo rk, transfer t o a different job, o r geographic dislocation-due to automa tion. Suchategories as welders, carpe nters, insulators, machinists, and clerical staff have been most heavily affected. At the same time, automation creates new highly-skilled jobs in programming, ope rating, and maintaining computerized production machinery. Workers need xtensive retraining programs, however, to prepareor such jobs.

Because of the high cost of downtime, efficient maintenance and fast repai rs become crucial in automated production, which places a great burden of responsibility and tight tim e constraints on mainte nance and repair crews. Computerized equipment can be sed to enhanc e the flexibility of work organization, leaving one in charge of planning one’s work time, but it may also be used to mpose a strict and inflexiblework regimeon factory and office worke rs bylosely monito ring their performa nce. As a result, auto mation can make work either easier or more exhausting and stressful, depending on he type of work organization.

Technocentric vs. Human-Centered Approaches Historically thepredominantapproach oautomation has been echnocentric: a goal of au tomatio n is to reduc e and ultimately entirely eliminate human participation in produ ction and eventually arrive at an unmanned factory. From this standpoint, workers are seen as a source of potential err ors, disturbance, and unreliability; on the other hand, automatic machinery isviewedas inherently more precise, reliable, and controllable. The technocentric approach extends the principles of Taylorist work organization t o modern information-processing and production systems. It is based on further subdivision of labor, with more complex and intelligent tasks trusted to flexible computer systems and simpler tasks left to low-skilled workers who assume aesidual role. Skill gradually passes rom people tomachines,and control functions are also transferred in the same direction.

Another risk ishe dangerof employees losing essential worki ng skills as work becom es increasingly mediate d by the computer. With automation, the worker has gone throu gh a series of transformations-from a di-

The technocentric approach f aces a fundamental paradox: it aspires to replace human skill with highly flexible computerizedmachinery,but this machinery requires even more human skill to operate, maintain, and repair it . Instead of “freeing” produ ction from the “ huma n elem ent, ” automation onl y increases the importance of highly qualified, versatile, and motivated

rect prod ucer of goods and services to the operator of production quipment oheprogr ammer of the computer that operates and controls that equipment. Engineering changed romhands-on tinkering with

worke rs. Accidents at th e nuclear powe r plants at Three Mile Island and Chernobyl testify hat autom ationdoes not eliminate the possibility of hum an erro r; itonly makes this error morecostly.

126

AUTOMATION

The Taylorist logic of seeking productivity by accelerating the pace of work may not apply in a computerized workplace. With computerization, companies do not simply automate, but “inform ate” their operations. Computer-based control of p roduction becomes an information-processing task; workers urn into analyzers of info rmatio n rather than simple machin e minders. Improving the quality of this analysis, instead

nical system s” approa ch, elaborated in Britain. Based on group assembly instead of a conventional assembly line, this newesignave workersmore initiative, flexibility, and control over pro duct quality. In the 1980s major American manufacturers began experimenting with worker nvolvement in decision-making, a recent example being GM’s Saturn project. The human-centered approach finds a source of productiv-

of speeding up worke rs’ movem ents, becomecrucial s problem of au tomation .

ity inmore efficient utilization f human abilities, rat her than in the utopian efforts to eliminate people from production.

An alternative approa ch aspires to chang e the workforce from being part of th e manufacturi ng probl em into part of the solution. Instead of taking skills, responsibility, and control away from the worker and absorbing hem into themachine,human-centered CIM systems mobilize th e in tellectual resources of all employees. Leading Japanese ompanies, uch as Matsushita and Toyota, achieved much greater productivity gains from autom ation than their American compet itors by decentralizing control and reorganizing the factory layout into producti on islands controlled by semi-autonomous multi-skilled teams responsible for all

operations. Reversing the Taylorist trend of subdivision of labor, the human-centered approach ntegrates functions and skills in flexible teams, wh ere workers can rotate obs and choose the optimal order and pace of w ork. Instead of being fo rced to follow instructions handed to them rom above, workers are motivated to play a greater role indecision-ma kingby programming C N C equipment on the shoploor. In the ate 1960s and early1970sonly a handful of American com panies , such as Procter & Gamble,Cummins Engine, and Gaines Foods, realized that greater productivity did not come automatically with more sophisticated equipment but required profound organizational change. In1974 Volvo built a highly produ ctive plant at Kalmar, Sw eden, which implemented the “soci otech-

Bibliography 1967. Bright, J. R. “Th e Development of Automatio n,” in Technology in We stern Civili zation, Vol. I1 (eds. M . Kranzberg and C. W . Pursell, Jr.), 635-654. New York: Oxford University Press. 1984. Noble, D. F. Forces o f P roduction: A Social History of New York: Knopf/Random House. Industrial Automation. 1988. Zuboff, S. In the Age of the Smar t Ma chin e: New York: Basic Books. The Future of Work and Power. 1989. Forester, T. (ed.) Computers in the Hum an Cont ext :

Inform ation T echnolog y, Productiv ity, and People. Cambrid ge, MA:MIT Press. 1992. Ayres, R. U., Haywood, W ., and Tchijov, I . (eds.) 4 Vols. London: Computer Integrated Manufacturing. Chapman & Hall. 1994. Allen, T . J., and Scott Mort on, M . S. (eds.) Information 1990s: Research Technology and the Corporation of the Studies. New York: Oxford University Press. 1996. Kling, R. ed.) Computerization and Controversy: 2nd Ed. San Diego, CA: Value Conflicts and Social Choices,

Academic Press. 1997. Rochlin, G. I. Trappe d in the Net: The Unantici pated Consequences of C omputer ization. Princeton, NJ: Princeton University Press. Slava Gerovitch

AUXILIARY MEMORY See MEMORY: AUXILIARY.