Artificial Neural Network And Its Application In The Turbine Generator Vibration Fault Prediction

In recent years, research on Artificial neural networks(ANN) has made great progress and research on the application of ANN has been conducted in almost every field. This dissertation makes an intensive study on the application of networks to steam turbine.In chapter one the history, status and development of the fault predicting for turbogenerator units are described. It also deals with the contents of the research work of the project.In chapter two ANN and BP algorithm are introduced in brief. An improved algorithm is proposed , which can help accelerate convergence and avoid sinking into the local minimum.The chapter three is my paper emphasis. The shortcomings of BP networks when it is used to predict the times series with increasing tendency are analyzed. Then an. improved BP networks, the combined BP networks(CBP), and its training algorithm are proposed, and the predictive performance of CBP is compared with the original BP networks and other predictive models. Finally, by means of using simulated data of the faults of steam turbines ,the predictive performance of CBP is checked.In chapter four the function , structure and concrete realization course of the software for predicting system are introduced.In chapter five the predictive performance of CBP is checked by means of using the operation data.In the last chapter of this dissertation, the conclusion and outlook of the research work are given, some problems which need further research are pointed out.