Research Of Blast Furnace Fault Diagnosis Based On Multi-Agent

Steel industry is the basic industry that related to the national economy and the people’s livelihood. As the key equipment for iron-making, blast furnace normal smelting is significant for iron quality and enterprise benefit. Along with the continued strong demand for steel in recent years, the development of blast furnace toward large-scale and information. The traditional and foreign blast furnace fault diagnosis technology showed great limitations in the application of new equipment due to the diffiereces of detection equipment, mineral raw material and operation modes. In recent years, as a branch of distributed artificial intelligence, the research of multi-agent systems gets more and more attention. Multi-agent research achievement continues to emerge provides a new thinking for large-scale complex system.This thesis analysis the common fault of blast furnace and detailed study system design and diagnostic method in fault diagnosis of the large-scale complex system, then used these methods for simulation running in blast furnace fault diagnosis.The main work is as follows:The system analyse the symptom and the cause of the common faults of blast furnace. Aim at complexity, oncurrency and nonlinear characteristics in blast furnace fault, multi-agent system consists of distributed, collaborative, parallelism, etc. Multi-agent fault diagnosis technology is suitable for blast furnace fault diagnosis system. The concept and structure of agent and multi-agent is detailed analysised and take full advantage of the preponderance that multi-agent method applied in large-scale complex system.According to the principle and the ways of multi-agent task decomposition, blast furnace fault object is decomposed into several sub object.Diagnostic agent is the core of multi-agent fault diagnosis system, it relevance to the property and accuracy of diagnosis system. Intelligent diagnosis method is chosen according to the diagnosis object feature in multi-agent system. In order to improve accuracy rate of the diagnosis, one fault even need two or more methods for diagnosis. Summarizing the experience of the existing diagnosis methods, this paper used PCA-BP neural network diagnosis agent and fuzzy expert system diagnosis agent to respectively diagnosis the blast furnace status fault. Both methods make up the dimension disasters and the local minimum in the traditional single neural network and the authority deficiency in single fuzzy logic and the symbolic representation and deduction in single expert system. Then diagnosis results of the two diagnosis method are fuzzed by using weighted average method to improve diagnosis accuracy.According to the field design structure in blast furnace diagnosis system, this design constructs simulation platform under the lab environment, run and test data acquisition and communication constructed by siemens PLC, it also compiled the monitoring and diagnosis interface using by Visual Basic 6.0 and Matlab 2012b. Both can be used to verify the effective and feasible of the system.