Multiple Time-Frequency Curve Classification for Tacho-Less and Resampling-Less Compound Bearing Fault Detection Under Time-Varying Speed Conditions

Abstract

Rotating machinery usually runs at time-varying speeds, which often brings failures, especially compound faults. However, currently, there is still a lack of satisfactory coupled fault diagnosis methods at variable speeds. Several methods have been developed, which try to eliminate rotational speed influences by converting nonstationary fault characteristic frequencies into stationary ones. But it brings new problems of computational efficiency and accuracy, and requires an auxiliary tachometer to measure the rotational speed. Time-frequency analysis can solve the problems, and extract time-frequency curves for fault diagnosis through time-frequency representation (TFR). However, for signals containing multiple faults at variable speeds, uninterested components will be generated when extracting multiple time-frequency curve (MTFC), which makes it difficult for compound fault diagnosis. To solve these problems, a novel MTFC classification method for tacho-less and resampling-less compound bearing fault detection under time-varying speed conditions is proposed in this paper. Here the TFR characteristics of a multi-fault vibration signal is analyzed, and a compound fault diagnosis strategy based on MTFC classification is developed. Firstly, MTFC is extracted by a local peak search method. Then, according to the relationship between instantaneous shaft rotational frequency, instantaneous fault characteristics frequency and their respective harmonics, two classification criteria for curves are proposed and MTFC is classified into interested curves and uninterested curves. Finally, the average ratios between interested curves are matched to theoretical fault characteristic coefficients to determine the fault type. Case studies on rolling bearing experiments verified the effectiveness and superiority of the proposed method.