Gear Fault Diagnosis Using Dual Channel Data Fusion and EEMD Method

Abstract

When a gear fault is presented, the ensemble empirical decomposition (EEMD) method has been widely used in the detection of gear fault and has given good results. However, under signals from two vibration transducers which are installed in different directions, the vibration data from every sensor becomes necessary to obtain much more precise results. This paper suggests a new feature frequency extraction approach for gear malfunctions diagnosis using the full vector spectrum (FCS) and EEMD method. The approach embodies model building of multi-sensor data fusion and main intrinsic mode functions(IMFs) analysing of merged data from two vibration sensors. By using EEMD the data from single channel can be decomposed into a number of IMFs. To alleviate the diversity problem of EEMD decomposition results occurring in two channels, full vector spectrum (FVS) is introduced. With full vector spectrum, information obtained from each sensor can be merged a piece of vibration evidence in frequency domain, and then we use Hilbert transform to calculate the envelope spectrum of each IMF to evaluate the fault feature frequency. In this work, simulated signal and gear experiment were conducted to validate this strategy. The results have proven that the proposed method can enhance the effectiveness of feature frequency extraction in diagnosing gear faults.