Sound signal analysis method overloaded freight train rolling bearing fault diagnosis
|School||Central South University|
|Keywords||Acoustic signal Rolling bearings Wavelet Analysis Neural Networks Fault Diagnosis|
The freight train rolling bearing fault diagnosis technology is a close connection with the actual science of engineering, the production of the actual need is the root cause of it is the development of the study simple diagnostic method has broad practical value. Rolling noise from freight trains and dissemination of theoretical analysis, considering the impact of freight trains Rolling sound attenuation and background noise, non-stationary acoustic signal for freight trains Rolling this characteristic, the use of modern signal analysis techniques on the acoustic signal processing and fault identification, in order to improve the effectiveness and reliability of the identification of its diagnosis. This paper studies the following aspects: for freight trains Rolling sound non-stationary signals using multi-resolution nature of wavelet transform to analyze the acoustic signal, found that the impact of the ingredients in the details of the wavelet decomposition signal is amplified, compared to the frequency and formed under various fault ground fault frequency to find the cause of the failure, in order to achieve the effective recognition of the signal waveform. This paper presents a nonlinear wavelet transform denoising methods - hierarchical threshold denoising algorithm. The simulation results show that the method can significantly improve the precision of the filter can effectively remove the noise while the main details of the signal reserved. And then from the power spectral density of the signal after wavelet transform analysis, simulation results show that the self-power spectral density analysis based on wavelet transform can effectively extract the characteristic frequency of the sound signals of freight trains rolling bearing fault, suitable for acoustic signals such nonstationary signal analysis and research. Feature extraction, this paper proposed a new method - based on the the interval wavelet packet characteristics of the frequency band of local energy extracted, it can be based on the need to subdivide each band. Proven characteristic factor can be a good work on behalf of the rolling bearing. The application of neural networks in the freight train rolling bearings intelligent diagnosis. Extracted fault feature based on the sound signal analysis in many ways, but each method only reflected in a failure characteristics alone diagnostic effect is not very good. This article is by contrast the fault feature a combination of different methods to extract as the input of the neural network, and ultimately determine freight train rolling bearing fault diagnosis using wavelet analysis and neural network combining. Rolling bearing fault diagnosis, neural network can reduce the operator's expertise requirements, fault diagnosis from the traditional method to artificial intelligence direction. At the same time, the intelligent diagnostic system technology can greatly reduce the maintenance personnel working pressure.