Gearbox Fualt Diagnosis Based On Time Domain,Frequency Domain-Wavelet Analysis And Neural Network

With the development of science, Machinery has developed to the direction of high performance, high efficiency, high automation and high reliability. The gearbox is used in changing the rotational speed and the transmission power for fixation rate, high torque tight and structure virtue etc. The gear box is an important component of mechanical device and break down easily; its running status has the very tremendous influence to the complete machine operating performance. So the importance of the condition testing and fault diagnosis of its transmission system consists not only in cutting time and cost of repairing, but also in raising diagnosing accurate, which improves repairing quality and acquires economic benefits dramatically.Vibration signal of gearbox is very complex, including not only going information of gears and bearings but also much information of other related parts and structures. It is difficult if we only use time-domain and frequency-domain or wavelet analysis means analyzing vibration signal to find fault accurately. If time-domain, frequency-domain and wavelet analysis specialty are provided at the same time, diagnostic veracity and reliability will be greatly improved. So this paper put forward and studies a new fault diagnosis technique-neural network diagnosis based on time-domain, frequency-domain and wavelet analysis. If time-domain, frequency-domain and wavelet analysis specialty are integrated to identify faults, and then trained as input of neural network. Study results show that the method can be improved gearbox fault diagnosis accuracy and reliability.The major works of this thesis are listed as below:(1) Vibration mechanism of the failure gearbox is analyzed. This paper analyses the types of gears and bearings failures and common causes, and study of vibration mechanism of gears and bearings.(2) The feature extraction of gearbox fault by time domain analysis and frequency domain analysis. By analyzing the vibration signal of 5 typical gear faults from the experiment, the author extracts characteristics in both time domain and frequency domain.(3) The feature extraction of gearbox fault by wavelet analysis. The basic theory of wavelet analysis and wavelet packets analysis are introduced. This paper used a feature extraction method based on wavelet packet energy. The method is illustrated reliability by an example.(4) Introduced the algorithm principle of the BP neural network, and this paper puts forward an improved BP algorithm, which can adapt learning rate using error correction, and carries out simulation and testing in terms of the algorithm. The results show that the improved BP algorithm has shortened the study time, improved the study efficiency and avoided the problem of local minimum in study to some degree.(5) Experiments were carried out. This paper designs the experiment project on fault diagnosis of the gearbox and builds a gearbox testing system in laboratory, then simulates 5 modes of gearbox representative faults, and collects the signal from the gearbox. Analyzing the vibration signal by time domain analysis, frequency domain analysis and wavelet analysis, then extracts characteristics as the input of improved BP neural network, and establishes network model. Through processing and analysis of many data, the diagnosis result is satisfactory. It shows that the new method presented in this paper, can be improved gearbox fault diagnosis accuracy and reliability.