A new fault diagnosis approach for bearing based on multi-scale entropy of the optimized VMD

It is well known that the parameter K of variational mode decomposition (VMD) needs to be preset according to prior knowledge without theoretical support for optimal setting. Thus, for the existing fault diagnosis methods for bearing based on VMD, the correctness of characteristic extraction and accuracy of fault diagnosis are extremely difficult to be guaranteed. To solve this problem, a novel collaborative diagnosis approach of petrochemical equipment for bearing based on optimal VMD and multiscale entropy (MSE) is proposed. Firstly, because optimizing the decomposition parameter K for VMD is difficult, an effective estimation model of K is constructed according to the frequency distribution characteristics of the decomposition components of local mean decomposition (LMD). Then, a novel characteristic extraction technique collaborating MSE and linear discriminant analysis (LDA) is proposed to establish characteristic samples. Furthermore, aiming at the fault characteristic of small samples for bearing, support vector machine (SVM) is introduced to identify the fault characteristics. Finally, the bearing fault data collected from the simulation platform of the petrochemical equipment laboratory is used to verify the effectiveness and engineering practicability of the proposed approach. The comparative analysises show that the proposed algorithm can effectively diagnose faults of the bearing with good engineering operability and scalability. © 2020, Editorial Office of Control and Decision. All right reserved.