Development Of Multi-Agent Based Fault Diagnosis System For The Rotating Machinery

The multi-agent technique is one of the key research topics in the machinery fault diagnosis. During the developing of the machinery fault diagnosis system for the rotating machinery, some key techniques are deeply studied in the paper.Firstly, the three layers distributed network structure is analyzed, and based on the CORBA and agent technique, the multi-agent frame for the system is offered in the paper, and the general structure of the agent in the system is designed also. Futhermore, applying the architecture of middle application program layer, ie the diagnosis control agent, the diagnosis control center is realized in the system; through reconstructing the traditional neural network and fuzzy diagnosis means respectively, BP artifical neural network diagnosis agent and the fuzzy theory diagnosis agent are accomplished in the system. Secondly, the dynamical coalition method and strategy based on the Dempster-Shafer evidence theory are presented in the paper. Some improvements of Dempster-Shafer theory, and multi-factor, layered mechanism are realized in the paper. And the method, accumulating the evidence of experience and self-learning, is provided in the paper.Finally, with JAVA (JSP) and CORBA, the rotating machinery fault multi-agent diagnosis system is designed out in the paper, the emulated result and the example show that the system is effective and feasible..