Intelligent Control and Maintenance Management Integrated System Based on Multi-Agents for Coal-Prep

Jun.2006J. China Univ of mining tech (English EditionVol 16 No. 2Intelligent Control and Maintenance management Integrated system Based on MultiAgents for Coal-Preparation PlantMENG Fan-qin", WANG Yao-caihool of Information and Electrical engineering, China University of Mining Technology, Xuzhou, Jiangsu 221008, ChinaAviation Oil Engineering Department, Chinese Air Force Xuzhou College, Xuzhou, Jiangsu 221000, ChinaAbstract: This paper discusses the progress of computer integrated processing(CIPS)of coal-preparation and thenpresents an intelligence controlled production-process, device-maintenance and production-management system of coalreparation based on multi-agents(IICMMS-CP). The construction of the lICMMS-CP, the distributed network controlsystem based on live intelligence control stations and the strategy of implementing distributed intelligence control sys-tem are studied in order to overcome the disadvantages brought about by the wide use of the PLC system by coalpreparation plants. The software frame, based on a Multi-Agent Intelligence Control and Maintenance Management in-tegrated system, is studied and the implemention methods of IICMMS-CP are discussed. The characteristics of distrib-uted architecture, cooperation and parallel computing meet the needs of integrated control of coal-preparation plantswith large-scale spatial production distribution, densely-related processes and complex systems. Its application furtherimproves the reliability and precision of process control, accuracy of fault identification and intelligence of productiondjustment, establishes a technical basis for system integration and flexible production. The main function of the systemhas been tested in a coal- preparation plant to good effect in stabilizing product quality, improving efficiency and reduc.Key words: intelligence. controlled process; multi-agent system; computer integrated processing system; coalpreparation plantCLC number: TP2741 IntroductionHowever, in the present system, the process ofcoal-preparation consists of three subsystems: intelli-By coal-preparation, the content of ash and sul- gence control, device maintenance and productionfur in coal can be reduced and the discharge quantity management subsystems. The software and hardwareof soot, sulfur dioxide and other harmful substances of these three subsystems are separately designed andwill decrease. The ratio of coal being prepared to the implemented in the current system. They are isolatedoverall coal produced in developed countries is 50% from each other, which restrain the optimization of90%, but it is only 20%-30% in China. Small the hardware resources and processing control Thescale production, poor mechanical reliability and a computer control system of coal preparation plantslow level automatization result in a high cost of coal has a highly concentrated structure. Usually, the PLCpreparation, which then becomes the main factor re- network is used for coal preparation integrated constraining the progress of coal preparation in our coun- trol, while an isolated IPC system is used for controlof such processes as jigger, coal flotation and denseSince the 1990s, via advanced technologmedia cyclone. The separation of software andhave exerted intelligent control in coal-preparation hardware results in"Data Islands". Although dataspects like computer integrated control, jigging access can be obtained by data base connectivityprocess, coal flotation etc. We also carried out a sys- technology(such as ODBC, JDBC), the isolatedtematical study and applications of aggregate produc- data still cannot make optimization between prodition processes and CIMS aspects of coal preparation tion processes Possible. while the integration of thesepu吧e: That has lifted the production- prend appli- three中国煤化工 or to improve theCNMHaration plants it isprocess control and management level of coal prepa- evertor of developingration plants to a new levelCiMS of coal preparation plantsReceived 11 October 2005: accepte05ComespondingauthorTel+86-13952206436;Emailaddressmengfanqin@sohu.comMENG Fan-qin et alIntelligent Control and Maintenance Management Integrated System Based onThe application study of agent and multi agent preparation process, due to its features of multiplesystems(MAS), which started in the middle of the variables, targets and non-linearity, intelligence con-1980s, has become a major topic for research in dis- trol theory is used to manage processing control overtributed artificial intelligence. New plans and solution jigger, coal floatation and dense media cyclone o-l)between the agents in such a way that it can be ation The production management Agent can plan the coaloperation, optimize the target, adjust the process andto deal with incomplete and indefinite knowlanalyze the data.2. The device maintenance agentCooperation between them can improve the basic can provide for testing the state of the equipment,ability of a single agent and display its aggregatetesting of technical parameters, an intelligent opinionpacity.When it is used to deal with an actual problem, of device malfunction and device repair management.Agent technology can establish the complicated system in the form of mini-subsystems, which are auto-IICMMS-CPnomous, easily managed but communicate and coop-erate with each other o!. Therefore, the mAs are usedDevicentelligeProductionduring industrial processes to set up distributed testMaintenanceanagementcontroling and controlling systems dealing with complexprocessing control problems. In 1995, Schrott and hiscolleagues raised the concept of a Multi-Agent DisFig 1 Configuration of IICMMS-CPtributed Real-Time System(MAD-RTS), consistingIn the multi agent systems, each single agentof a compiler, operating system kernel and communi- carries out different functions and covers its owncation tools. They established an experimental system range of functions. The interactive methods and reali-controlled by an elevator; then, based on MAS, they zation of negotiation and cooperation between eachconstructed a distributed field bus control system for Agent is a complex problem. Given the nature of thea model plant. Several articles [7-9] discuss the ap- production process of coal preparation plants, in anplication of MAs in different industrial processes. IICMMS-CP process based on MAS technology, weThese studies prove the feasibility and validity of need not only be concerned with the control, mainteMAS technology in industrial process control systems. nance and management of each agent, but also theirMAs has an unequalled superiority in solving interaction in this distribution environment - .Incomplex systems. The production of a coal prepara- 1997, Bonasso and others presented the three-layertion plant is a complex industrial process with several agent structure in their study of autonomous robotsinput and output variables, most of which are non- In 1998, Schreckenghost and others applied it to in-linear, indefinite and spatially distributed. To con- telligence controls over the life-support system in thestruct an integrated system of intelligent control, space station, proving that it was suitable for distrib-equipment maintenance and production management uted process control and decision-making systemsfor this industrial process has always been technicallyhe ICMMS-CP based on MAS adopts thedifficult. The autonomous, distribution and adaptive three-level hierarchical organization structure, shownfeatures of Agent offer a new way of resolving this in Fig. 2. The bottom level is the agent of a closedproblem. This paper discusses the Mass Agent tech- loop control and the data-interface layer(DACMnology and realization of the IICMMS-CP (Intelligent Agent). As for the intelligent process control(IPCIntegrated Control, Maintenance and Management device maintenance(DM) and production manageSystem of a coal preparation plant). It makes a spe- ment(PM), the basic task is the closed loop controlcific study of the distributed testing and control net- and data-collection. The bottom Agent implementswork of a coal preparation plant, the distributed intel- the basic control over the production process, dataligent control, the general structure of the Mainte- collection on the state of the device and technologicalnance Management system and their realization parameters. The second level is the task Agent (TAmethods. It is designed in order to realize the decen- Agent). This level schedules the bottom Agent so ittralization of the control function of a coal prepara- can reach the expected target given the task and modetion plant and its intelligence process control as well prescribed at the previous level. The top level is theas the computerization and the integration of equip- Agent of planning and adjustment(PA Agent). It dis-ment maintenance and management.trib中国煤化工 en the general tary calling the next2 IICMMS-CP ArchitecturelevelCN target values ansystem modes given the general state of the system,The functional system of IICMMS-CP consists their effects and sets value accordingly. For example,of three parts: the integrated production control (IPC), in density and liquid level control of the magnetitedevice maintenance (DM) and production manage- medium of dense media cyclone, the Agent of PAmen, M% The IPC refers to the fact that in the coal- level assigns control objectives, designates the con-J. China Univ of Mining Tech (English Edition)Vol 16 No.2trol mode of the dense media cyclone, schedules theIn the"computer integrated manufacturing platTA Agent to implement the density and liquid level form and integrated automation systems of coal minecontrol objectives according to the production man- and coal-preparation plant"program, we developed,agement plan and the state of the device. The ta via the standard field bus and distributed structure,aAgent distributes this control process into tasks of distributed computer control network platform, acdensity andnd liquid level fuzzy control, density data cording to the requirements of the production envi-on the bottom Agent to complete the task 3collection and control over the opening of by-pass ronment and the current condition of devicesvalves according to the established model and calls technology. This platform consists of fieldbuswork interfaces, intelligent 1O stations, gateways,intelligent stations (IS), control stations(CS)and aIICMMS-CPIICMMS-CP terminal station(CPS), shown in Fig 3MD AgentstatemodelUser requestPM IPC BAM1/OO1O1/Otask、 statecommandfeed backgrading sifterPM IPC BAM H匚 DACMageis Coal bunkFig. 2 The multi-agent system architecture of IICMMS-CPThe user interface of IICMMS-CP(UI AgentIs distributionprovides the information interactive interface forCoal-Preparation Plantmanual intervention. With the Ul, an operator cangive a set of control parameters, revise control mod-Can busEtherneels, fault diagnosis models, production planning mo-dels and the management plan for devices. He canalso directly control a device.Fig 3 The field-bus network of coal preparationIn the above hierarchical organization structure,the agent in each level only needs to be responsible toThis platform divides the network into two nethe previous agent. It is the logical structure of a tree segments: CAN and industry Ethernet according todiagram. But in the distributed system, each agent real-time requirements. The industry Ethernet de-implements the exchange, interaction and negotiation creases the uncertainty of Ethernet data transmissionof the information using a uniform communication time by using the gateway to isolate the frequent,protocol and message-reaction mechanism in the con- massive data transmission node forcibly In the sys-trol network, based on the field buscontrol station and Iicmms-cp terminal station. The3 Distributed Control Network Environ- message-transmission is accomplished via the Can-ment of IICMMS-CPbus field bus and industry Ethernet. Generally, thelogic control Agent is accomplished by PLC; theThe distributed intelligence control and mainte- data-collection Agent by intelligent IO, other tasknance management system needs to construct a real- Agents by Is, negotiation and the schedule agent bytime, distributed computation environment. In the control stations, the adjustment agents by IICMMSIICMMS-CP, based on MAS each agent possesses the CP terminal station, but all kinds of agents can beproperty of great autonomy and self-government, executed in a station at the some timehile the distributed environment provides an interactive mechanism for negotiation and cooperation 4 System Implementation of IICMMS-CPbetween agents. The control system based on the fieldBased on Agentbus(FCs) provides such a real-time, distributed con-trol computation environment. The communicationscale coal mines coal preparation procprotocols and the technological standards of the field esses中国煤化工 Jigger, a dense mebus provide a mutually cooperated, interoperable in- dia cteraction mechanism for each agent. The agents are ter all l.CNMHGation of slime wa-ns of rawincorporated in network devices and the general tar- year. In its ICMMS-CP system, the device mainte-get is accomplished by the interaction and coopera- nance mainly implements the fault detection, alerts iftion between each network device in a distributed main devices fail, suggests repairs and manages de-computation environmentvices. Management inspects the production optimiza-MENG Fan-qin et altelligent Control and Maintenance Management Integrated System Based ontion index, the product structure and the status of the fashion and revisions made as needed. SM evaluatesprocess. The intelligence process control executes the if a control order can be implemented or not given theintelligence control of dense media cyclone and jig. data provided and reacts accordingly. TS can adoptdifferent strategies given the real-time state of the4.1 Three tier agent architecturesystem. For example, in a process such as reducingthe ash content of the jigger products, TS can chooseThe implementation of the IICMMS-CP func- to adjust bed thickness, adjust the coal feed rate ortions of intelligence control, device maintenance, change the periods of the hydraulic and pneumaticfault analysis and decision-making of production pressure according to the current coal granularity andmanagement are carried out in a three tiered structure, washability. At the same time, it can adjust the conshown in the flow-chart presented in Fig. 4trol strategy according to different control effects. PDcan divide the general target into sub-targets againPlan and decision.and plan the new tasks after a target failsMaking AgentIn IICMMS-CP, the three TS Agents, dense media cyclone, jigger and concentration and filtrationlan/monitoring can be used, respectively, 1)to implement the fuzzycontrol of liquid level and density of the dense mediacyclone, 2)control the rotation speed of the coalfeeder, the fuzzy control of the bed thickness in thefirst and second periods of the jigger and 3) controlthe hydraulic and pneumatic valve, 4)control the oncentration and filtration precess of slime watersubtask4.2 Communications between agentsIn the Multi Agent system, communication be-tween Agents is the basis for cooperation. We makeTA Agentuse of the predefined message system to provide forcommunication between agents. We manage thetransmission via CAN and TCP/IP protocoL.The following messages are defined无sSM Updating message: when the Agent of theExcaitionSM type detges in the state of the system it,eventAgentinforms the system of this message. Then the Agentsof fault and production process identification evaluateFig 4 Three tier control architecturethe state of the system device and proceed accordinglyPlan and decision-making(PD): in the processFault message: the Agents of fault and producof implementing the control task, a general target can tion process identification inform the system of anbe divided into multiple sub-targets via a task-classfault message, which contains the information of thefication plan network. Each sub-target can be further faulty device or process, the type of fault and advicedivided into multiple tasks, to which the resources in dealing with the problemss will be guided and supervised. In the judgment notify the expected target or task. Up until now, thisand identification of the fault pattern, the type of fault kind of message is handled manually. It implementscan be identified and maintenance or on-line adjust- parameter settings, runs specific programs and cor-ment tasks proposed according to fuzzy knowledge rects faultsand inference rulesFault recovery message: when some specificTask execution serial (TS): at this level, the tasks fault can be corrected on-line, or some checks arefrom the plan or operators are divided into sub-tasks needed to confirm the fault, the Agent of fault evaluaa series of specific tasks to be executed are con- tion sends this message and TA and SM will executestructed(see Fig. 4)Execution management(SM): at the third tier中国煤化工sms8closed loop control and data-collection modules are provCN MHGtem Besides theseconstructed and integrated into a dynamic implement- there will be inquiring and answering informationtation network which accomplishes the tasks assigned between the agents at each level in the running proc-at the previous leveless. We define six types of such messages: HA-PMThe implementation methods described above HA-TS, HA-SM, PM-TS, PM-SM and TS-SM.Dueare robust At each step errors can be found in timely to the different methods of dealing with these mesJ. China Univ, of Mining Tech (English Edition)Vol 16 No. 2sages and the expression patterns between each level, 5 Conclusionwe define the information structure and format Forexample, the execution Agent corresponds directlyThe Multi-Agent Intelligence Control and Mainwith the meter and actuator and deals with the infor- tenance Management Integrated System of coal-mation of digital quantity and switching value, while preparation plants is proposed. The characteristics ofthe Pa level makes use of the fuzzy numbers and distributed architecture, cooperation and parallelfuzzy inferences to handle fault identificationcomputing meet the needs of integrated control of thefore, the execution Agent neeal-preparation plants with a large spatial prquantities into fuzzy numbers when it answers PA. distribution, densely-related processes and complexThe following example may clarify the point. The systems. Its application further improves the reliabilagent may change the value of a specific liquid level ity and precision of process control, the veracity ofinto"extremely low", low","middle", "high"and fault identification, the intelligence of production ad-extremely high". In order to manage such informa- justment and establishes a technical basis for systemtion exchange, the information interface agent was integration and flexible production of coal-preparation plantsReferences[1] Shi Y Q, Tan C, Gong D W. CIMS system integration of coal preparation plant. Journal of China University of MiningTechnology, 1999, 28 (5): 479-481(In Chinese)[2] Xu ZF, Wang Y C. The Study of computer control system used in the coal preparation plant Coal Mine Design, 2000(1): 43-44.(In Chinese[3] Ma FQ, Wang Y C, Wang Z K, Computer integrated production system for coal preparing plant. Coal Mine Automatization,1997(1):8-10.( n Chinese)[4] Yang H C, Sun J Z, Suo L S, et al. Agent-based intelligent control-maintenance-technical management system and applicationin pumping station control. Computer Engineering and Application, 2002, 17(2): 47-48.(In Chinese)[5] Wang P Z. The trend and suggestions in coal preparation automation. Coal Preparation Technology, 2000(6): 45-47 (In Chi[6] Wang J P, Chen H, Xu Y, et al. Study of agent-based distributed hierarchical intelligent control. Control and Decision, 200116(2):177-180.( In Chinese)[7] Dong H Y, Jia S f, Li H, et al. Distributed intelligent control of train based on multi agents. Journal of the China Railway So-ciety, 2004, 26(6): 61-65 (In Chinese[8]ouclhadjD.amulti-agentsystemfortheintegrateddynamicschedulingofsteelproduction[eb/ol].http://www.asap.csott.ac uk/publications /pdf /Ouclhadjthesis Pdf, 2003[9] Wang JX, Wei Z, Lu Q Model of distributed computer control system based on the multi-agent system. Journal of Hefei Uniersity of Technology, 2004, 27(12 ) 1570-1573(In Chinese)[10] Ma F Q, Ding EJ, Jin n, et al. Intelligent control for jigging process. Journal of China University of Mining Technology2002,31(3):293-297.( In chinese)[11 Meng FQ, Wang Y C, Jiang JG et al. Study of multivariable fuzzy control for density and level of suspension in heavy media coal processing. Journal of China University of Mining Technology, 2005, 34(2): 252-255.(In Chinese[12] Kuang Y L, Deng J J, Liu H Y. Application of artificial intelligence in field of coal preparation. Journal of China Universityof Mining& Technology, 2001, 30 (6): 558-562.(In Chinese)[13] Jin M, Shen D Y, Zhou Y. An Ideal realization platform for real-time distributed multi-agent control theory-field bus controlstem. Information and Control, 2001, 30 (4): 322-327. (In Chinese[14] Hahndel S, Levi P. Distributed task planning method for autonomous agents in a FMS. In: Proc of IEEE/RSJ/GI int Conf onintelligent Robots and Systems (IROS 94). Munchen: IEEE Computer Society Press, 1994: 1285-1292[15] Han W, Jafari M A Component and agent based FMS modeling and controller synthesis. IEEE Trans Syst, Man and Cyber-netics-part C: Applications and Reviews, 2003, 33(2): 193-206[16] Ahmad MR, Shamsudin H M, Amin R M. Development of reactive control stratege for multi-agent robotics system. In: Procf Seventh International Conference on ControL, Automation, Robotics and Vision(ICARCVOZ). Munchen: IEEE Computer[17] Brennan R w, Fletcher M, Norrie D H. An agent-based approach to reconfiguration of real-time distributed control systemIEEE Trans On Robotics and Automation, 2002, 18(4): 444451中国煤化工CNMHG