Research On Signal Processing Method Of Ultrasonic Flaw Detection For Composite Insulator

Composite insulators are an important part of transmission lines.Their performance is closely related to the safety and stability of power systems.However,due to various reasons,composite insulator has defects inside.Therefore,it is necessary to detect the internal defects of composite insulators.Ultrasonic detection is widely used for its the advantages of being harmless to the human body and high detection accuracy.When ultrasonic detection is used to detect internal defects of composite insulators,the processing of ultrasonic echo signals plays a crucial role in the detection results.In this thesis,the ultrasonic flaw detection signal processing method of composite insulators is mainly studied.Firstly,on the basis of the research status of ultrasonic detection signal processing methods for composite insulators at home and abroad,the types of internal defects of composite insulators and the mathematical model of ultrasonic echo signals are analyzed.Secondly,aiming at the problem of noise in ultrasonic signal,an improved wavelet threshold de-noising method based on empirical mode decomposition and permutation entropy is proposed.Simulation results show that the proposed method has a de-noising effect.Then,based on the analysis of the advantages and disadvantages of wavelet packet transform and empirical mode decomposition in feature extraction,the two methods are combined to extract feature vectors.And for the problem that the extracted feature vector group has too high dimension,the principal component analysis method is used to reduce the dimension of the extracted feature vector group,which lays a foundation for the subsequent defect type identification.Finally,the identification algorithm of ultrasonic echo signals is studied.In view of the low identification accuracy of BP neural network under small sample defects,the support vector machine is used to identify and classify ultrasonic echo signals.The simulation results show that:In the case of different training samples,the support vector machine has a longer running time than the BP neural network,but its identification accuracy is higher than that of the BP neural network.In this thesis,the research on the ultrasonic flaw detection signal processing method of composite insulators effectively avoids the noise interference in the ultrasonic flaw detection signal,and can accurately identify different defect types,which provide a theoretical basis for the performance evaluation of composite insulators.