Intelligent Diagnosis method for Multi-flaws of Roller Bearing by Time-Frequency Waveform Distribution and Extreme Learning Machine

摘要

This paper proposed a method for intelligent condition diagnosis of rotating machinery using the time-frequency waveform distribution (TFWD) and extremelearning machine (ELM). The precisely diagnosing method of single flaw and multi-flaws in a roller bearing are proposed as follows: First, the measureddata is processed by time-frequency analysis. Second, in order to extraction the feature of signals, waveform distributions in the time domain and frequencydomain are calculated separately. Third, extreme learning machine is introduced to distinguish the signal state by the time-frequency waveform distributionsin frequency domain. Moreover, if the state of single is abnormal, the time-frequency waveform distributions in time domain are used to identify the flaw issignal flaw or multi-flaws. The efficacy of these novel methods are confirmed by the results of the condition diagnosis for roller bearings with single flawand multi-flaws at the speed of 600 rpm, 900 rpm and 1200 rpm.