Fault Diagnosis of Rotating Machinery Based on Deep Reinforcement Learning and Reciprocal of Smoothness Index

Rotating machinery are widely used in industry, and vibration analysis is one of the most common methods to monitor health condition of rotating machinery. However, due to the presence of outliers and interference, vibration signal becomes very complicated in reality, and it is important to reduce the influence of outliers and interference. Since a bandpass filter can eliminate a lot of above influence, it is usually selected to process vibration signal in classic fault diagnosis. The selection of the lower and upper cutoff frequencies of the bandpass filter is very critical. In order to extract fault characteristics from vibration signal, this paper proposes a new method which uses deep reinforcement learning algorithm and the reciprocal of smoothness index to control the bandpass filter to select a frequency band with the highest signal-to-noise ratio. Then, envelope demodulation is performed on the filtered signal so as to diagnose the faults of rotating machinery. Two sets of data collected from the test rig are used to validate the effectiveness of the proposed method. The comparisons with fast kurtogram and GiniIndexgram show the superiority of the proposed method. It also suggests that reinforcement learning has a great potential in the field of mechanical fault diagnosis.