Research On The Multi Agent System For Controlling Cement Production Process

The cement production of new suspension preheater(NSP) kiln is complex, large scale and continuity. NSP is a system with variables, couples and multi dimensions. The cement production of our country occupies one in third of world's total amount, but the cement process control is still on the stage of handle operation at present. Therefore it is a main direction for our cement industry's development to raise the level of cement process control and to achieve optimal control of cement process, which purpose is to obtain good quality, high quantity and to reduce energy consumption.In this paper, the multi agent technology which is in the front of artificial technology was applied to develop the multi agent system for controlling cement process. This research earns important theoretical significance and practical value.The characteristics of cement process were analyzed. The control strategy of multi agent and the control scheme integrated with optimal control and diagnosis control were proposed. The multi agent system of cement process control with the organizational construction of distributed and multi level was developed.The fuzzy control model of cement process was built by the method of fuzzy control. The fuzzy controllers of raw material preheat, clinker calcine and cool were set up. The optimal control of cement process was achieved.Aim at the complicated abnormal status of cement process, the multi process and parallel mode of diagnosis and control was put forward, which combined with state judgment, trend analysis, situation evaluation and diagnosis control. The hybrid expression method of knowledge was adopted and the classify knowledge bases were constructed. The distributed, multi level and hybrid inference algorithm was set up, the approximate reasoning method was applied and the diagnosis control for the abnormal status of cement process was realized.The optimal control agent and the diagnosis control agent based on the BDI(Belief-Desire-Intention) model were constructed. The optimal control agents communicate and coordinate with diagnosis control agent by means of multiple level blackboards , which combined overall blackboard and local blackboards, and the extended KQML(Knowledge Query and Manipulation Language).The multi agent system software for controlling cement process was developed based on the above research, and was successfully applied in off-line control of the NSP cement plant with capacity of 5000t/d. The computer control level of cement industry is raised as a result of the research, development and application of the multi agent control system of cement process. Better economic profit and social benefit can be got and the prospect of the system's application is wide.