Research On Ore Grinding Process Intelligent Control System Based On Multi-agent

Ore Grinding is an important manufacturing process in mineral processing,which decides the energy consumption, economic and technical indexes of amineral processing plant. It makes Grinding process hard with long time delay,Multivariable coupling and production uncertainty. With the requirements oflowering manufacturing energy consumption and improving energy utility, it iseven necessary to improve Grinding process control technology. Predictive control,adaptive control, artificial intelligent control have replaced the traditional simpleautocontrol techniques gradually and artificial intelligent control has been focusedgreatly in Grinding processing.Based on analysis on Grinding process and integrative study on various oreGrinding intelligent controlling techniques, this paper proposes to construct anintelligent control system with Multi-Agent technique. In terms of Multi-Agentsystem application, an ore Grinding process is complicated, continuous anddynamic. There are various factors in controlling ore Grinding process and controland decision are made by experts with experience. Information in ore Grindingprocess exists in distributed form. These features are included in Multi-Agentsystem application, which makes it feasible to study Multi-Agent system.Differing from the other intelligent control systems, this paper makes localanalysis on the entire control process in ore Grinding and follows the controlstrategy combing local and overall， integrating trouble diagnosis, production control and operation status analysis, which makes the system an integrative andcomprehensive information platform. This concept embodies the flexibletechnique, coordinative organization and comprehensive decision of Multi-Agentsystem, which fulfils the capabilities of Multi-Agent system to solve complicatedproblems and constructs an intelligent control system model of Multi-AgentGrinding process. The system application shows the theory of Multi-Agent systemcan perform better information merging, adjusting production parameters andfulfill intelligent decision in overall process to improve productivity, lower energyconsumption and bring considerable economic benefit.The following studies have been conducted in the paper:1. The analysis and summary have been made in Grinding process controlboth home and abroad, and a special comparison study in artificial intelligentcontrol and application feasibility of Multi-Agent technology in Grinding processcontrol. Based on the study of application of Multi-Agent technique in Grindingprocess, this paper proposes that Multi-Agent technique performs well incoordinating and solving various problems, which satisfies the requirements ofinformation merging and integrative deciding in Grinding process control.Therefore, a Multi-Agent intelligent control system model had been proposed toconstruct.2. A deeper analysis has been made in Grinding process and Grinding controlstrategy established in terms of multi-task decomposition. First, divide Grindingprocess control into several individual sub-control tasks, such as feeding control,level control, pressure control and Grinding trouble diagnosis and control decision.Then, integrate sub-control tasks according to coordination rules. Finally, the control of entire Grinding process can be realized when sub-tasks are conducted.3. Modeling method of Multi-Agent system has been studied in this paperbased on task-oriented Grinding control strategy. An intelligent control systemmodel of Multi-Agent Grinding has been established and model structure, functionmode and data flow model have been designed. The concept of establishing themodel: abstract the sub-tasks into individual Agent of Multi-Agent system fromGrinding process, namely feed control Agent, level control Agent, pressure controlAgent, trouble diagnosis Agent. An individual Agent performs its control taskindependently, and the system control tasks can be performed throughcoordination of individualAgents.4. This paper has conducted modeling construction and function realization ofan individual Agent in the frame of Multi-Agent Grinding intelligent controlsystem.Model study of trouble diagnosis: features of trouble in Grinding process areunder study in this paper. Considering Multi-source of trouble, uncertainty anddata distribution mode, a Grinding trouble diagnosis method based on datamerging has been put forward in this paper. The target and strategy of Grindingtrouble diagnosis have been made through data collection and experience summary.With the application of fuzzy expert system, a knowledge system of Grindingtrouble diagnosis has been established to perform reasoning of uncertainty oftroubles based on diagnosis rules and feature data. A model for Grinding troublediagnosis Agent has been constructed to perform Jess-based reasoning of troublediagnosis.Model study of task processing Agent: A control model based on adaptiveneural fuzzy network has been put forward, a clustering algorithm of data sample improved and similarity fusion algorithm introduced to perform clusteridentification. This model, based on fuzzy rules and data modeling, can decide andoptimize parameters in system identification and the best input and outputvariables can be obtained. A simulation test and comparison test have beenconducted in this paper, which shows the model has better adaptability andaccuracy.5. This paper has performed a study on coordinating and consulting processof Multi-Agent system, which compares centralized and distributed coordinatingmethods and makes a task-sharing coordination strategy of data transfer, resultsharing and information sharing based on blackboard. In blackboard model, dataservice Agent, blackboard control Agent and control decision Agent areestablished; by blackboard Agent centralized coordination, sub-task Agents makecommunication and data demands through data service Agent and perform dataand result exchange through blackboard sharing database, and control decisionAgent performs control decision and conflict resolution on the operation results ofsub-tasks. The test and experiment on functions of blackboard system modelshows the coordination mechanism proposed in this paper can exerciseindependent executive ability of an individual Agent and avoid resourcecompetition and result conflict, which satisfies the function need of Grindingproduction, improves performance efficiency of individual Agents and the abilityof overall control decision.6. A model of Multi-Agent Grinding control system based on JADEMulti-Agent system development technology has been established. Modules offeed control Agent, pressure control Agent, level control Agent, blackboard controlAgent, data service Agent, control decision Agent and human-machine interactionAgent have been constructed in this paper and the functions of the system can be fulfilled. Production data in Grinding process is applied to perform comprehensivetest on the system and the test result shows: the system has satisfied the designrequirements of the mineral processing plant. In the same operating condition,Grinding production under Multi-Agent intelligent control system becomes steadyto regulate feeding, pressure and level values, increase the processing amount ofthe ball grinder and improve production efficiency and product quality with betterstability.Based on the study and practice conducted in this paper, followingimprovement can be made in the future:1. Multi-Agent Grinding intelligent control system has a better applicationprospect. Further study should be done in Grinding production process and moreintelligent control in production process conducted.2. Multi-Agent Grinding intelligent control system, based on Multi-Agentsystem organization, has better extendibility and system function extension can berealized through increasing the number ofAgents.3. Further study should be conducted on trouble diagnosis, adding mechanicalautomation knowledge and accumulating more operation experience and reducetroubles in production process.4. Further optimize Grinding control results to satisfy the higher demand ofGrinding production target in future mineral processing plants in order to bringmore economic benefit.5. Further study in Agent theory and Multi-Agent technology to improve theMulti-Agent system performance and reasoning, planning and learning abilities.The study of Multi-Agent control technology enriches the means of Grindingintelligent control, which is an application breakthrough in the field. Besides, the Multi-Agent Grinding intelligent control system changes the traditional concept ofGrinding control, which is a comprehensive technology integrating computertechnology, automation technology and intelligent control and plays an active rolein developing technical innovation in Grinding production with a wide applicationprospect. Science and technology is the first productivity and future mineralprocessing and mining and metallurgy are destined to step into a new era ofintelligent control.