Adaptive compressive sensing method based on optimal classification for bearing vibration signals in axial piston pump

Bearing fault diagnosis is very important to improve the reliability and safety of axial piston pump. There are some major problems that the existing techniques require large amount of storage, transmission and processing time for vibration signal. In this paper, an adaptive compressive sensing method is developed for vibration signals reconstruction based on optimal classification algorithm. Firstly, vibration signals were divided into blocks according to their cutting size and an energy sequence was produced in accordance with the energy of each signal block. Secondly, the energy sequence was classified by the QPSO algorithm to obtain the biggest variance. Whereafter, using the K-SVD algorithm, the proposed adaptive sparse dictionary learning framework enhances reconstruction capability of vibration signals. Finally, bearing fault testing experiment validates the effectiveness of the adaptive compression sensing method method. The proposed method has higher reconstruction performance than the classical K-SVD dictionary training method under the same compression ratio. The reconstruction accuracy is further improved and the reconstruction time is shortened.