A computer-based algorithm for machine fault diagnosis

作者：Li Li^*; Zeng Dexue; Wang Hongmei; Chen Baojia

来源：International Review on Computers and Software, 2012, 7(7): 3554-3558.

• Hidden Markov model(HMM)
• Machine fault diagnosis
• Maximal likelihood probability (MLP)
• Mechanical vibration signals
• Parametric models
• Rolling bearing diagnosis
• Wavelet packet energy spectrum
• Wavelet packet method

Since mechanical vibration signal contains a lot of noise, it is an important problem in machinery diagnosis to extract the useful information from measured signals. The paper presents a method based on parametric model -- hidden Markov model (HMM) to diagnose mechanical faults. First, wavelet packet method is used to decompose a measured signal into a set of sub-band signals. The energy values of the sub-band signals based on wavelet packet energy spectrum (WPES) are extracted as the feature values, which are normalized and quantification coded to be used as the input of HMM model. Then, HMM is trained to make the output of N dimensional model (represented by N types of faults) achieve maximum likelihood probability (MLP). The fault classifier based on HMM is obtained. Finally, when a detected signal input to the classifier, the type corresponded to the MLP is the fault type to be diagnosed. Applied the method to rolling bearing diagnosis, different faults in the rolling ball are identified. The results show that this method is simple and reliable, and with good diagnosis effects.

• 出版日期2012