Turbine State Monitoring And Fault Intelligent Diagnosis System

As for turbine, the importance of its fault diagnosis is introduced; the research status quo in this field all over the world and the difference from variable diagnosis method is presented. Hybrid model fault diagnosis method based on expert system and back-propagation algorithm is applied to solve the turbine fault diagnosis. Good effect is obtained. The main research content and result is followed:(1) Some existed fault diagnosis methods are summarized. Fault diagnosis method based on mathematical model, which is applied less, includes Fault diagnosis method based on state estimation and process parameters estimation. Fault diagnosis method based on wavelet is worthy of application. Fault diagnosis method based on knowledge is being used widely.(2) Signal processing means of acquired data is introduced. The technologies including nonlinear compensation, scales conversion, algorithms of signal preprocessing, FFT spectrum analyzing is emphasized.(3) Multiple neural networks based on layers diagnosis model, Back-propagation algorithm and its improved algorithm is analyzed in details. Training state of back propagation is explained according to turbine fault sample. Then improved method of using layers classification diagnosis model is put forward.(4) Faults diagnosis technologies based on expert system is analyzed. Basic structure, function, knowledge expression and acquirement, designing steps and development principles of expert system are introduced in details. Expert system based on rules is applied to turbine fault diagnosis.(5) Turbine fault intelligent diagnosis system is based on neural networks and expert system. Its framework is analyzed. It is made up of several low micro controllers system and one upper computer in the industry area. Realization of hardware and C51 software in the low micro controller is introduced. Diagnosis software with rich function is developed by Visual C++ 6.0 tool.The left problems and direction to further research is followed:(1) Wavelet technology should be applied in odd signal examining and separationbetween valued signal and noise.(2) If reforming neural networks structure, optimizing neural networks weights by genetic algorithm and program code, then perhaps diagnosis efficiency is enhanced to a great extent.