Study On Key Technique In Fault Diagnosis System For Turbogenerator

This dissertation contains seven chapters concerned mainly with the theory and methodology of fault diagnosis for rotating machinery such as turbogenerator. A fault diagnosis system used in power plant has been develpoed. Several key technique in fault diagnosis system have been further studied.The dissertation covers the following major problems:1. Study on automatic diagnosis method with the features of fast calculating and high belief degree and build a on-line diagnosis system based on this method.2. Study on knowledge-based fault diagnosis system systemeticly to find the causes of faults, maily include the following aspects: representation of symptoms and experience knowledge; measurement of fuzzy symptoms; inference algorithm of the inaccuracy diagnosis; interface of acquisition of evidence and management of knowledgebase.3. Research on synthetic methods for fault diagnosis system.The main ideas of this dissertation are as follows:Chapter one describes the history, status and development of fault diagnosis technique for rotating machinery, and analyses the advantages and disadvantages of some fault diagnosis mehtods, such as nueral network, fuzzy thoery and knowledge-based fault diagnostic mehtods. This chapter also deals with the background and importance of the dissertation. At the end of the chapter, the research contents are gathered up.The second chapter is about the fault symptoms used as the basic knowledge in fault diagnosis field. A new concept of diagnosis parameter is defined in this chapter. The use of parameters is to make the sympotom description standard. The fuzzy degree of those fuzzy symoptoms is measured by fuzzy membership function.Chapter three discusses the vibration information based automatic diagnosis mehtod. The acquisition of spectrum feature of fault pattern class by the fuzzyclustering algorithm is the bisis of pattern recognition in the level of fault class. The result acquired by fuzzy relationship between vibration symptom and fault shows the feature of fuzzy-relation-based method. A synthetic method based on-line diagnosis system has been built for mornitoring and diagnosing multi-points of multi-sets.The aim is to develop knowledge-based fault diagnosis expert system in chapter four. It covers some important aspects of expert system. The first aspect in research scopes is about the representation and using of inaccuracy knowledge; The second is to solve the inaccuracy inference by using inference algorithm based on weighted fuzzy rule. The input interface of evidence is designed to use fuzzy theory, so it can avoid the difficulties in traditional method when inputing belief by human The application of database technique in knowledge-based building is discussed in chapter four too. At the end of the chapter, a multi-task fault diagnosis expert system is built and it runs under the control of mixed inference strategy.Chapter five introduces the task part strategy following the idea of "part-fusion". Two kinds of synthetic models of same signal-based multi-feature information and muilt-signal-based information are discussed. At the same time, the multi-dimension knowledge representation model is built by the analysis of the konwledge principle insynthetic diagnosis system. A synthetic diagnosis system architecture is constructed for turbogenerator at last.The whole architecture of condition detection and fault diagnosis system is demonstrated in chapter six, which includes the hardware and software architecture, then the author analyses the feature of this application software for fault diagnosis .A few of main functions of system are illustrated by examples.In the last chapter of this dissertation, the reserch work is summarized and some important conclusions follow the next. The author explicates some developing aspects in the field of fault diagnosis for the coming research work.