A MP-based Method for Periodic Fault Impulses Detection in Rotating Machinery

As the core component of contemporary industry, rotating machinery has been applied in various industries. However, rotating machinery typically serves in harsh environments, which makes it very prone to performance degradation or even failure. And failure or decline in performance will cause significant economic losses and even casualties. Thus, the investigation of rotating machinery condition monitoring is critical to guarantee the safe running of machinery and the reduction of maintenance costs. However, condition monitoring focus on traditional signal processing technologies relies too heavily on expert experience and imbalanced data limits the application of intelligent diagnostic methods in practice. In view of the limitation, a novel approach based on matrix profile (MP) is presented for periodic impulses detection in rotating machinery condition monitoring. In this work, the local matrix profile (LMP) is first used to detect periodic impulses which are caused by the local damage of rotating machinery. Then, in order to improve the impulsive feature, an adaptive strategy for parameter selection based on L-kurtosis index is introduced. Furthermore, to examine the performance of the presented method, the simulated signal and actual data are analyzed. The findings indicate that the approach can successfully determine the state of rotating machinery.