Research On Rotating Machinery Fault Diagnostics Based On Time-Frequency Image Feature Extraction

More than 200 years since the industrial revolution, great changes have taken place in the history of human society due to machine, the machine is becoming more and more high precision, high efficiency, high intelligence, mechanical equipment running state and fault feedback is particularly important. It is related to people's personal safety, the company's production efficiency and the process of social development. Rotating machinery is most of the existing form of operation machine, the rotating parts is the core of the whole machine, bearings and gears are two commonly used components, their status affects the stability and sustainability of the whole system. State analysis method based on analysis of vibration signal of rotating machinery is the most effective method by far. In this paper, based on the analysis of vibration signal of rotating machinery, according to gears and bearings, by using the method of image processing, to feature extraction and fault judgement, the main contents are as follows:(1) Based on the traditional signal analysis method, studies the different time-frequency processing methods, focusing on the complex unsteady signal of machine, put forward the deficiency of traditional time-frequency analysis under the unsteady signal, focuses on the wavelet time-frequency analysis and wavelet time-frequency diagram, transform the time-frequency image to gray level and normalized it, according to the characteristics of the gray time-frequency image to find the proper feature extraction method.(2) Through the analysis of image texture feature extraction, choose effective texture statistical characteristic analysis method to time-frequency signal:gray level co-occurrence matrix. Choose the appropriate method to generate gray level co-occurrence matrix for time-frequency image, extracts the characteristic parameters, and analyzes its existence deficiency, based on gray level co-occurrence matrix feature parameters, propose the improved characteristic parameters, and verified on the test data of gear. The improved algorithm parameters obtain better effects.(3) Based on the bearings test, get different fault state and fault type of bearing datas, studies the mathematical morphology of image processing method, introduce the morphological spectrum computation to the analysis of the time-frequency diagram, to deal with data and compares their separation form, analysis of different types of data, found that the morphological feature extraction based on time-frequency image for different fault types have more obvious effect.(4) Based on the NI company programming platform Lab VIEW, to a bridge crane lifting platform translation gear box bearings,write a program to realize the on-line detection and alarm, using the time-frequency image feature extraction method to offline data analysis, validate the practicability and reliability of the system.