| As the tank reactor being an important device, its operation and safety directly affect theproduction in the chemical industry. Studying the diagnosis technology of tank reactor is anecessary measure to protect the safety of the entire production system, reduce the accidents,promote the development of the whole national economy, and have a very important significance.Recently，the technology of fault diagnosis is developing very fast and there are many methodsfor it, but most of them are the simplex methods. Based on analyzing the currently domestic andforeign study, a fusion diagnostic system based on the Radial Basis Function (RBF) neuralnetwork and fuzzy-expert system is presented. The system is tested by the sample of the datacollected from different placements. The main work is as follows:(l) We detailed analyzed the typical fault reasons and the solving measures for reactor,which provides the theoretical basis for fault diagnosis based on information fusion.(2) Based on the in-depth analysis of the RBF neural network and Particle SwarmOptimization (PSO) theory, PSO is presented to train the RBF neural network and it proves thatPSO has powerful network generalization ability and stronger identification ability. At last, thereactor is diagnosed by trained neural network. The results which meet the fact are got.(3) We studied the fault diagnosis expert system principles and basic structure, thendiscusses the fuzzy-expert system has the advantages of traditional expert systems, andintroduced the reason of selecting fuzzy expert system. we focused on the establishment offuzzy-expert system. Knowledge base module based on fuzzy production rules, divided into factbase and rule base.(4) Finally, we describes the information fusion technology, the basic principles of faultdiagnosis, and the relationship between them, and then present a new fault diagnosis model andmethod based on information fusion technology. We get conclusions with deviation by differentdiagnosis methods, and then we further fusion these conclusions by decision level fusion. Thefinal result is more accurate than a single result. To the uncertainty of fault diagnosis, thedecision-making fusion based on D-S evidence theory is put forward. The basic conception of evidence theory is introduced, and combined with the example is analyzed. |
