A Novel Nonlinear Dynamic Measure for Early Detection of Bearing Fault Using Weighted Squared Envelope-Based Symbolic Lempel-Ziv Complexity

Lempel-Ziv complexity (LZC) stands out as a promising nonlinear measure for monitoring bearing health status. However, during the early stage of fault inception in real-life applications, small amplitude-bearing fault impulses get submerged under unwanted environmental noise. Because of this reason, the original LZC algorithm fails to detect a bearing fault at the earliest point of inception. Aiming to overcome this limitation of LZC, in this research, first, transient impulses corresponding to a bearing fault are extracted by weighting the squared envelope of the collected vibration signal. Then, in order to quantify the complexity of the weighted squared envelope (WSE), the symbolization process is incorporated into the original LZC algorithm to enhance its performance by changing the binary symbolization to a multisymbolization phenomenon. In this context, the proposed measure is termed WSE-based symbolic Lempel-Ziv (WSE-SLZ) complexity. One synthetic simulation model and two different runs to failure-bearing data have been used for performance verification of the proposed WSE-SLZ method. Results show that the proposed WSE-SLZ not only detects bearing fault at the earliest point of inception but also performs better than conventional measure Mahalanobis distance (MD) and permutation entropy (PE), as well as advanced complexity calculation method, namely, symbolic LZC (SLZC).