Study On The Fault Diagnosis Of Turbine Set Based On The Second Generation Wavelet Analysis

Turbine set is the most important rotating machinery of the power system.Because of its high failure rates and difficulty of diagnosis,the research on its online monitoring and fault diagnosis has always been the hotspot in the industrial diagnosis field.Rolling bearing is one of the core components of the turbine.As the high fault rates,its working condition will directly influence the performances of the whole equipment,even more,will influence the safety of the entire power system.Therefore,this thesis' s objective is to study the status monitoring and diagnosis of the rolling bearings.A series of works is started from the extracting features of the vibration signal of bearings.The main contents are listed as follows:1.The vibration mechanisms of rotor and rolling bearing of turbine are explained.Also some modern theories include diagnosis information acquisition,fault features extraction and fault patterns recognition are introduced.2.Method based on the second generation wavelet analysis in fault diagnosis of rolling bearings is proposed.In signal decomposition,in order to avoid frequency derangement,The second generation wavelet energy feature extraction method based on scale transform is proposed.The Hilbert Vibration Decomposition(HVD)method is adopted to the faults' preliminary diagnosis according to the extracted energy features.The experimental results show that the method can extract the fault characteristic frequency accurately and locate it,and avoid frequency derangement.3 To improve the accuracy of fault classification,an improved BP and Elman neural networks method to diagnose the bearing fault is present.First the BP networks are adopted to train only limited in the sensitive features.As a result,the complexity of calculation is reduced.Then the Elman networks are used to predict the fault signals of bearing under different degrees of injury.Experiments results show that the improved BP and Elman neural networks method can be applied in fault diagnosis and early fault prediction of rolling bearings of turbine generator sets,and has a higher accuracy rate.