Fault diagnosis of bearing running status using mutual information

Fault diagnosis and the prediction of remaining useful life play a key role in the Prognostics and Health management (PHM). One of the most important challenges in modern PHM is how to diagnose the fault of mechanical equipment accurately. Aiming to improve the fault diagnosis precision of the rotating machinery, a novel method of fault diagnosis combines mutual information models and second generation wavelet packet decomposition is presented in this paper. The traditional approaches to fault diagnosis always focus on the signals of a certain time. This method is different from traditional models because fault can be diagnosed more accurately by comparing the conditions of two different periods. Firstly, vibration signal of different times is extracted from the working bearings. Secondly, each frequency band's energy is calculated through the second generation decomposition and the energy of joint probability distribution of two different periods of time as well. Finally, the mutual information of two different periods of time is gained by using their joint probability distribution. A life test of a bearing is used to validate the proposed methodology and the results demonstrate that the proposed methodology is an effective tool to improve the accuracy of fault diagnosis of bearings running status.