Application of pattern recognition in gear faults based on the matching pursuit of a characteristic waveform. (July 2017)

• Record Type: Journal Article
• Authors: Cui, Lingli; Yao, Tianchang; Zhang, Yu; Gong, Xiangyang; Kang, Chenhui
• Is Part Of: Measurement. Volume 104(2017)
• Page Start: 212
• Page End: 222
• Publication Date: 2017-07
• Subjects: Faulty gear -- Characteristic waveform extraction -- Pattern recognition; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2017.03.021
• Languages: English
• ISSNs: 0263-2241

Abstract

Abstract: Aiming at the issue of recognizing the damage pattern of gear signals, a new matching method based on the morphological characteristics of the vibration waveforms was proposed in this paper. The signals of different types of gear faults share the same modulated frequency; therefore, only whether a gear is damaged or not can be identified from the frequency domain. However, it is difficult to recognize the different fault types of the gears. There are significant differences among the signal waveforms in the time domain of different fault types; thus, research on the recognition of the types of gear-damaged faults is conducted using the repeated characteristic waveforms of the same fault type based on the morphological characteristics of the gear fault signal. The signal morphological characteristics of different gear fault types were extracted using a searching algorithm, and then the matching effect was achieved from the basic atom library, which was constructed by the multi-stage characteristic waveform after an optimized process by the method optimal direction (MOD). Finally, the simulation signals of four types of different gear fault types and real engineering signals were processed using this method, and the effectiveness of the method was proved.