은닉 마르코프 모델과 인공신경망을 이용한 기계시스템의 결함 진단

Abstract

For the fault diagnosis of a mechanical system, pattern recognition methods have been widely used in recent research. A pattern recognition method determines the soundness of a mechanical system by detecting the variation of the system vibration characteristics. The Hidden Markov Model (HMM) and Artificial Neural Network (ANN) are recently used as the pattern recognition method in various fields. Most of the previous related researches use only one pattern recognition tool to classify the signals. In this research, a hybrid method that combines the HMM and ANN for the fault diagnosis of a mechanical system is introduced in order to improve the diagnostic accuracy. As a target model of fault diagnosis, a rotating wind turbine blade having a crack is selected. By extracting the acceleration from rotating blade having a crack, conduct the diagnosis about identifying the presence of a crack and finding the location and depth of a crack. Because vibration characteristics of a rotating blade according to location and depth of a crack have differences, it is possible to identify the crack location and depth using pattern recognition. In this research, the Fast Fourier Transform (FFT) is employed to extract feature vector. Moreover, to improve the diagnostic accuracy of the method in spite of noise existence, a moment having a few specific frequencies is applied to the blade.