Application of Adaptive Lasso-Based Minimum Entropy Deconvolution for Bearing Fault Detection Based on Vibration Signal

Minimum entropy deconvolution (MED) and its related methods have been applied to bearing fault detection extensively due to their good performance in fault feature enhancement. However, their performance is also constrained largely by their parameter setup. In order to improve the performance, an adaptive least absolute shrinkage and selection operator (Lasso)-based MED (AdaLMED) is proposed in this article. Different from previous MED and its related methods, AdaLMED utilizes their respective advantage of Lasso and MED to strengthen its performance on fault detection. Moreover, a k-sparsity strategy is introduced to Lasso for setting its regularization parameter adaptively in AdaLMED, which improves the performance of Lasso effectively without losing fault-related information. Based on the fault feature extraction ability and fault feature enhancement ability of Lasso and MED, the proposed AdaLMED has a better performance than the previous MED and MED-related methods. For verifying the performance of the proposed AdaLMED, AdaLMED and MED-related methods are performed on the simulation and practical fault signal, respectively. The results indicate that AdaLMED has the excellent robust stability to parameter setup and better applicability in practical industrial application compared with MED and other MED-related methods.