Intelligent manufacturing-The Challage for Manufacturing Strategy in China in the 21st century

Intelligent Manufacturing : the Challenge for Manufacturing Strategy in China in the 21st century what we will do YANG Shuzi ,LEI Ming , GUAN Zailin ,XIONG YoÜIt School of Mechanical Science and Engineering, Huazhong University of Science and Technology, Wuhan ,430074,China Abstract This paper first introduces the project of intelligent manufacturing in china and the research state of the I- IMRC (Intelligent and Integrated Manufacturing Research Centre) of HUST (Huazhong University of Science and Technology) ,then reviews the recent advances in object-oriented and distributed artificial intelligence and puts forth the view that these advances open up the prospect of systems that will enable the true integration of enterprises. In an attempt to identify domain reguirements and match them with research achievements ,the pa- per examines the current literature and distinguishes 14 features that are common. It argues that effective enter- prise-wide support could be greatly facilitated by the existence of intelligent software entities with autonomous processing capabilities ,that posses coordination and negotiation facilities and arc organized in distributed hier- archical states. Keywords : intelligent manufacturing ,system integration ,distributed artificial intelligence ,Active Intelli- gent objects ,Multi-agent system ,autonomous agents. 1. Introduction All industrial countries are exposed of increasingly severe national and international competition. This pro- duces surplus capacities With considerable presure on prices due to the excess supply,forcing manufactures to reduce costs through a variety of measures or to increase the utility of their products. As increasing attention is being paid to specific customer requirements ,there has been a massive increase in the variety of products and parts with smaller batch sizes ,tighter delivery schedules and extreme demonds in terms of product quality and custom service. One of the key icons of the 1980s was the characterization of manufacturing systems as being fragmented into "islands of automation". Automation in various production processes has resulted in "isolated is lands of automation" connected by incompatible interface. The coordination of various corporate activities from marketing ,planning , research and development to lo- gistics ,design and production and delivery costs an inordinate amount of resourecs. Bridges connecting these automated islands in the present system simply convey information and material but are far from harmonizing their functions in optimum efficiency. 2 ISPIE Vol. 2620 O-8194-2012-3/95/$6.00 Downloaded from SPIE Digital Library on 21 Nov 2010 to 202.120.6.23. Terms of Use: http://spiedl.org/terms Administrator 线条 Administrator 线条 Administrator 注释 facilitated 推动 Administrator 线条 Administrator 线条 Administrator 注释 surplus capacities 容量过剩 fragment （成）碎片 incompatible 不兼容的 harmonizing 协调，和谐 Administrator 高亮 Administrator 线条 Administrator 线条 Administrator 线条 Administrator 线条 Many studies have documented the fundamental changes that are now rocking the manufacturing world. [1][41[5]Among the trends which virtually all observes have noted are globalization of manufacturing (mar- kets ,supply chains ,and technology) ; increasingly complex and competitive markets ;shortened product life cy- des ; increasing requirements for quality ; increasing customization of products ;faster paced advances in technol- ogy ; rapidly expanding options in materials and processes ;and increasing skill requirements. In China , the highly skilled human resources are also shortage. With the more and more fierce competition in international market ,enterprise circles ,academic circles and governments have r.eached a common sense that only those enterprise which possess high technolgies and adopt modern organizationed management methods and means can survive in the globe"bussniess wars". The concept of Intelligent Manufacturing System (IMS) is defined by Japanese as"an advanced production system which ,through interface of man and intelligent machinery ,integrated the entire manufacturing processes from order booking through design , production and distribution and maximizes output in flexible responsiveness"[5]. The IMS goes beyond current emphases ,on flexibility and computer integraation. In partic- ular , intelligent manufacturing systems will possess the innate ability to respond ,promptly and correctly , to changes in requirements. They differ from conventional manufacturing systems (even advanced ones ) in their inherent capability to adapt to changes without external intervention. The remainder of this paper is organized as follows :A brief introduction to the China project on intelligent manufacturing is given in sect. 2; the general description of object-oriented distributed artificial inteilgence is presented in sect. 3 ;the common features are presented in sect. 4; the architecture for object-based intelligence appear in sect. 5 ; finally the concluding remarks are given in sect. 6. 2. The project of intelligent manufacturing in China Intelligent manufacturing is the direction of the development of future manufacturing technology. Recent- ly , several international joint research projects on intelligent manufacturing have been put into action succes- sively. In 1993 ,the project of intelligent manufacturing was organized by National Natural Science Foundation of China. The research project will mainly focus on the following three aspects: a. The fundamental theory and technical architecture of IMS ,including the synergistic decision model in the distributed parallel enviroment ,IMS architecture ,the multi-agent cooperative solving strategy ,the distribut- ed heterogeneous association knowledge base and data base ,etc. b. Intelligent and Integrated Cell Technologues ,including ID (intelligent design)/IPP (Intelligent process planning)! IM (intelligent manufacturing) ,IINC (intelligent and integrated qumerical control) ,IIQC (intelli- gent and integrated guality control) ,and the virtual-reality technology ,etc. c. Intelligent Machines , including intelligent machining centre (IMC) and intelligent numerical control milling-machine or lathe. In 1992 ,The IIMRC (Intelligent and Integrated Manufacturing Research Center) was set up in HUST,it concentrates its main effort on the study of manufacturing automation and intelligence and has achieved re- markable progresses in modern manufacturing theory,integrated CAD/CAPP/CAM ,intelligent prediction and diagnosis ,CIMS ,NC technology,integrated quality control and intelligent compensative control. In 1992,an intelligent manufacturing project, "Intelligent Grinding Control of the Leadscrew",was fin- ished. By means of compensating technology with artificial neural networks ,the grinding precision of 3-meter SPIE Vol. 2620 / 3 Downloaded from SPIE Digital Library on 21 Nov 2010 to 202.120.6.23. Terms of Use: http://spiedl.org/terms Administrator 注释 document 证明 virtually 事实上的 innate 先天的 inherent 先天的 reminder 剩余的 architecture 结构 synergistic 增效的，协作的 heterogeneous 不同的 Administrator 线条 Administrator 线条 Administrator 线条 Administrator 线条 Administrator 高亮 Administrator 线条 Administrator 线条 Administrator 线条 Administrator 线条 Administrator 高亮 Administrator 高亮 Administrator 线条 Administrator 线条 long leadscrew is up to the first level according to chinese standard. [21] Since 1992. the cooperation in the field of IMR (intelligent manufacturing research) between the govern- ment of the Republic of Singapore and the government of the People's Republic of China has been launched. NS- FC (National Natural Science Foundation of China) actively organize the internationci or regional cooperation in the development of intelligent manufacturing. An Asian &Pacific IMR Coalition will be set up. 3. Object-oriented distributed artificial intelligence There has recently been a heightened awareness of the role of advanced manufacturing technologies in the manufacturing information system ; The actual implementation of intelligent manufacturing systems in most companies ,however,is based on myriad (and often incompatible) hardware and innumerable software applica- tions. The latter can be grouped into three distinct categories: (1 ) The first category groups together applications in design ,drafting ,process planning ,engineering ,etc. It can be labelled"Computer aided Desing and Engineering". ( 2 ) The second category , labelled"Manufacturing Planning and Control ", consists mostly of applications relation to production scheduling , aggregate capacity planning ,shop floor control ,etc. (3) The third category can be labelled"Manufacturing Automation",this category incorporates numerically controlled (NC) machines ,computer-aided inspection , automated materials-handling , and warehouses and flexi- ble machines systems. It has been argued that decision support in manufacturing assumes that an information transition ,associat- ed with changing an entity's measured attributes ,is accompanied by a material transition ,changing the physical attrbutes of the machined workpiece. That is ,the logical transformation of operand data by one or more proces- sors into a finished information product is accompanied by a concurrent physical transformation fo the wok— piece into a financial product. [16] The intelligent functions that intelligent manufacturing systems should perform have been classified as :re- action , learning and problem solving. Reaction is the most primitive form of intelligent function in which the manufacturing system does some form of pattern recognition and provides some response to it. Learning re- quires that the manufacturing system recognises significant experiences ,data or generated plans and incorpo- rates this new information into it control structure ; actually ,learning incorprates a capability in the manufac- turing system to modify its knowledge base. Finally ,problem solving activities in manufacturing include mod- elling the problem domain form one state to another,generating plans for such transitions ,monitoring the exe- cution of the plans ,etc. Generation of such plans is a non-trivial task,because most problems of practical inter- est in manufacturing would be NP-complete. Hence problem-solving techniques shoule attempt to develop heuristics that generate' reasonably good' but not necessarily optimal solutions. In the area of distributed factory control YAMS(Yet Another Manufacturing System) is a prototype system that apportions tasks by negotiation [ii]. YAMS views the problem of factory control as a search througha space the dimentsions of which include the equipment available at a site ,the products to be manufactured and the available resources such as time ,inventory and storage space. YAMS models a manufacturing enterprise as a hierarchy of workcells, or functional groups of machinery. This model corresponds closely to the traditional manufacturing view of a corporation as a hierarchical system of plants ,FMS and machines. YAMS employs the contract net protocol of negotiation iii order to accommodate the stochasticity inherent in manufacturing. Each 4 ISPIE Vol. 2620 Downloaded from SPIE Digital Library on 21 Nov 2010 to 202.120.6.23. Terms of Use: http://spiedl.org/terms Administrator 线条 Administrator 注释 implementation 执行 myriad 无数的 incompatible 不兼容的 aggregate capacity 总功率 incorporate 合并 apportion 分配 hierarchy层次 stochasticity 协调 inherent 固有的 Administrator 线条 Administrator 线条 Administrator 高亮 Administrator 打字机 Administrator 打字机 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 高亮 Administrator 线条 Administrator 线条 Administrator 线条 Administrator 线条 node in the contract net corresponds to a workcell of the manufacturing company ,This hierarchy models compo- sition ,not control. Each node is a negotiation entity that can communicate not only with its parent and children ,but also with its siblings. Each node has a library of process plans describing the processes it knows how to per- form. For example ,assume that the Engine—Plant is a contractor to execute the make—engine 'process. Af- ter consulting its own process library , it finds that the first step in making an engine is the process , makeblock' , and that it does not know to perform this operation. It broadcasts a task announcement for the block process and receives bids for other nodes. After an evaluation of the bids it awards the most successful bidder. In order to support the characterisation and management of a heterogeneous system (such as the factory control system ) YAMS adopts the CSP formalism for concurrent processing as an analytical communications model. In the context of flexible manufacturing systems ,Shaw and Winston [8][9] view each FMS as a multi-a- gent system in which each agent is a flexible manufacturing cell and develop a system that includes a bidding process based on the contract-net protocol for introducing an agent's performance. In addition ,by viewing a DAI system as an adaptive system capable of learning to improve its performance ,they used a genetic algorithm as a competitive learning process. They applied these methods to the scheduling problem of flexible manufactur- ing systems (FMS). Within a process-based perspective of the manufacturing system Parthasarathy and LEI [ioJ[iJ extend the Actor model of concurrent computation and develop an actor+ model of the manufacturing system. The super- script plus (+ ) in actor refers to the model's ablilty to simulate physical flows and transformations in addi- tion to information flows and conputations. In addition. in order to incorporate the ability to reason about time in actor , they extend the work on point and interval representations of time and temporal reasoning .In apply- ing the concepts of actors in a manufacturing system ,Partyasarathy faced the following problem :the creation of actors ,which is a concept crucial to the Actor model .initially seems irrelevant to manufacturing systems , since machines and robots cannot create copies of themselves. In order to overcome this difficulty they consider the production environment as a highly dynamic one , in which machines are versatile and capable of performing a variety of operations with different set —ups. Then ,once they assign a unique address (or identifier ) to each ma- chine ,robot ,etc. of the manufacturing system. they can define an actor in the actor+ model as any distinct oper- ation or transformation that can be performed at any of the addresses within the system. Hence the concept of actor creation can be retained and operations can' create' new operations. 4. Elements of models 4. 1. Common features Since an important step in the formalisation of concepts is a model of the models used to describe computa- tional support in both the office and production environments must represent the inherent characteristics of the latter. The paper argues that the computational models used to support office and production management work exhibit a number of common features. These features are classified in two groups that correspond to the two dif- ferent levels if the system designed :the level of the single , independent element ,and the system level. Here we distinguish between the system architecture (the global ,top—level design of the entire system) and the internal structure (the features of single computational elements). Fourteen features of the abstract models are exam- SPIE Vol. 2620 / 5 Downloaded from SPIE Digital Library on 21 Nov 2010 to 202.120.6.23. Terms of Use: http://spiedl.org/terms Administrator 线条 Administrator 线条 Administrator 线条 Administrator 注释 formalism 拘泥形式 perspective 透视图全景 versatile 通用的，万能的 Administrator 线条 Administrator 高亮 Administrator 高亮 med ; seven for each level. The features referring to the internal structure are :specialization; representation ; effectuation ; learning; adaptability ; planning and intentionality. Specialization. This refers to ensuring that each computational element can efficiently solve a part of a given problem , that it has a specific area-expertise ,for example , robots that handle specific production pro- cesses or automated office assistants that perform prespecified calculations. Another issue here relates to knowl- edge organization ; distinctions have been made between static and dynamic knowledge. It seems ,however , that the open system perspective ,in which there may not be any global control ,goals shared knowledge or Success criteria is the most appropriate for IMS. However ,some of the systems revieved adopt a more centralised ap- proach ,using blackboards as a common repository for knowledge and information exchange. Representation. This feature attempts to capture the requirement that each element represents se- mantically loaded information. Data abstraction ,encapsulation and inheritance are crucial in this context. In some cases ,the computational elements represent' experts' and lead to approaches similar to cooperating expert systems. In other cases (actor in the production environment) ,mechanisms form object oriented programming are borrowed ,such systems represent software elements as agents or objects of the real world. Effectuation (transformation) . For example , in an office system a certain document gets filled in ,checked ,approved and signed ,hence it follows a pre-specified procedure ,during which it changes' states'. The actor+ model views change as the heart of manufacturing activity ,the manufacturing system conducts three basic activities conversion ,transportation and storage. According to Parthasarathy and KIM ,their transforma- tion model classifies conversion activity into three basic types :geometric transformations (associated with size and shape dimensions) ; quality transformations (associated with changes In the structure and properties of raw materials) ; and union transformations (involving the combination of two or more physical resources ,such as assembly transformations ). A topic arising here relates to the reactive versus deliberative nature of effectua- tion ; The term' deliberative' implies that an agent possesses reasonably explicit representations of its own belief- 5. plans and/or goals that it uses in deciding which action it should select at a given time. On the contrary ,non- deliberative behavior implies that the atent's beliefs ,plans and/or goals are embedded (or precompiled) into the atent's structure. Learning. The office and factory are environments in which inormation dynamically changes, hence they generate problems which could only be solved by advanced Learning abilities. Learning is defined here as the ability to perform new tasks that could not be performed before. For example ,agent FMS adopt a learn-by- being-told strategy ,in which successful bidder FMS spawn additional sgents : the learning technique uses genetic algorithms. On the other hand ,the learn action in the KNOs system facilitates learning[15] , using this action, a certain KNO takes an imported operation and adds it to its opertation list ,removing of operations can be done with the unlearn action. Adaptation. Here differentiate between adaptation and learning in that by learning a computational el- ement incorporates new operations. while by adapting it changes its current behaviour ,in order to account for newly acquired information and performs old tasks better ,faster ,more accurately ,etc. Planning. Planning capabilities are needed when ,prior to taking action a data structure represcnting the intention to take the action is developed;such a data structure corresponds to a plan or a schedule. Planning and scheduling activities are needed to organize efficiently both office and production work. 6 / SPIE Vol. 2620 Downloaded from SPIE Digital Library on 21 Nov