Research On Time-Frequency Domain Fault Diagnosis Method Based On One-Dimensional Signal

Fault diagnosis and signal detection technology has always been an important research topic,especially with the development of science and technology,the emergence of complex giant systems,it is even more demanding for fault diagnosis methods.In the current popularity of myoelectric prostheses and wearable devices,more requirements are placed on the accuracy of signal detection techniques.Based on the time-frequency domain analysis method of one-dimensional sampling signals,the detection of EMG signals,the fault diagnosis of rotating machinery and the fault diagnosis of equipment power system are studied in this thesis,and different methods are proposed and introduced:(1)Aiming at the detection of EMG signals,based on the double threshold detection algorithm,a multi-threshold detection algorithm is proposed.By setting the width threshold,the useful signals that are not recognized by the double-threshold detection method are identified,and the pulse interference that generates false alarms is detected.Filtration was carried out,and the idea of antagonism was introduced therein,which improved the detection accuracy.(2)Aiming at the fault of rotating machinery,a fault classification method based on time domain vibration signal in multi-dimensional space is proposed.Firstly,based on the periodic characteristics of rotating signal sampling,the data collected by each sensor in one rotation period is expressed as In the form of a multidimensional vector,the data collected in different periods is composed into a matrix,and the eigenvalues of the covariance matrix are obtained on this basis.Then,the eigenvalues obtained by different sensors are combined to obtain the eigenvalue vector after the original data is expanded.In order to distinguish different types of faults,the dimensioned vector of the dimension is regarded as a point in the multidimensional space,the different points represent different data sets,and the points of the data of the same fault are represented in the multidimensional space.Recently,different types of fault data are distinguished by this property.Finally,the method was tested using data collected from the laboratory.(3)A fault diagnosis method combining fast Fourier transform and neural network.Firstly,the one-dimensional time domain signal obtained by sampling is processed by the fast Fourier transform algorithm to obtain the frequency domain signal,and then the frequency domain signal is rewritten into a matrix form,and the eigenvalue of the covariance matrix of the matrix is obtained as the original signal.Characteristics.The eigenvalues of the obtained signals are used as the input of the neural network to detect and classify the signals,and the methods of extracting the feature information in the time domain and then classifying the neural networks are compared and analyzed.