Differential Extraction and Experimental Validation of Essential Characteristics in Gear Faults

The analysis and extraction of vibration characteristics exclusively associated with gear faults are crucial for improving the reliability of fault diagnosis. The vibration feature specifically generated by gear faults is considered as the essential characteristic (EC) in this article. Firstly, this article takes typical surface pitting and root cracks as examples, the fault EC is extracted through a differential extraction operation in dynamic, followed by analyses conducted in both the time-domain (TEC) and frequency-domain (FEC). Subsequently, a novel differential extraction method for real fault signals is proposed, encompassing the exponential decay dictionary (EDD) construction, the frequency correction and differential operation. Finally, the accuracy of fault EC theory and the effectiveness of EC extraction method are validated through two experimental cases. This article provides a new perspective for fault dynamics, and the proposed method effectively removes interference components within the modulation band, enabling the extraction of EC. Furthermore, it holds potential applications in data mining or machine learning models.