Research On Multi-Agent Theory And Its Application In Gauge Control System

In recent years, with the continuous improvement of the control requirements and the complexity of the industrial process control systems, traditional control strategies and simple control methods are more and more difficult to obtain satisfactory quality in complex process control, thus the support of the new theory is urgently needed. The powerful expression of multi-agent theory about complex systems and its open, intelligent features, making it become the best theory to support modern industrial control systems. Continuous rolling process is a typical complex industrial process, and the strip thickness accuracy is an important performance indicator. Therefore, the multi-agent theory is applied to the automatic gauge control systems, the control system is divided by functions, tasks and knowledge according to the multi-agent theory, then the model structure of the agent, cooperation and coordination mechanisms between the agents, design and implementation of multi-agent control system are studied.Firstly, the thesis starts with the basic concepts of thickness control. On the basis of analyzing the reason of strip thickness fluctuation and the basic law of thickness variation, we introduce the basic form of typical automatic gauge control and their features. At the same time, the thesis studies the basic theory of multi-agent system, then a detailed description about the basic concepts and architectures of agent and multi-agent system are carries out. In addition, we research the cooperation and coordination methods, the communication mechanism between agents and so on.Secondly, on the basis of the classical automatic gauge control system and advanced multi-agent theory and technology, the design of automatic gauge control system based on multi-agent theory is completed. The multi-agent automatic gauge control system is mainly composed of AGC agent, prediction agent, decision agent and management agent. The AGC agent contains two control methods which are GM-AGC and Tension-AGC, and realizes the basic control tasks through complementary advantages of both; The purpose of prediction agent is to predict thickness and the plasticity coefficient quickly and accurately, and then enhance the thickness control accuracy and efficiency of rolling mill; Cooperation mechanism between the AGC agents needs the support of the decision agent; Management agent take the security and stability of the system as main indicators, then supervise, manage and coordinate each agent and information.Finally, the thesis studies the improvement programs and algorithms to realize the agents’ function. The prediction agent use BP neural network based on GA-AdaBoost to enhance parameter prediction ability. Design a neural-fuzzy inference system based decision agent to improve the coordination mechanism between the AGC agents. The simulation results verify the good performance of multi-agent automatic gauge control system.