Gear Fault Feature Extraction Based On Improved Dynamic Time Warping And Blind Source Separation

The low-speed and heavy-load working environment often leads to serious wear,cracks,broken teeth and other faults of key components such as gearbox and gear,and then leads to the shutdown of the whole transmission system,resulting in huge losses.Fault feature extraction is a key part of the gear fault diagnosis technology.In order to extract effective fault features from mixed environmental noise and interference vibration signals,a gear feature extraction method based on improved dynamic time warping and improved independent component analysis is proposed in this paper.The main research contents are as follows:(1)First,the improved dynamic time warping algorithm is used to align the health signal and the fault signal of the gearbox,and then the original length of the signal is recovered by resampling to obtain the fault residual vector signal.In order to solve the problem that the first and last parts of the dynamic time warping algorithm cannot be aligned and the algorithm is inefficient,this paper proposes a fast derivative dynamic time warping method based on phase compensation(PCFDDTW)by combining the derivative dynamic time warping and the fast dynamic time warping.(2)In order to further extract fault feature components in residual signals,a singlechannel gear fault feature extraction method based on improved Independent component analysis(ICA)is designed.First,the collected empirical mode decomposition(EEMD)is used to reconstruct the multi-channel observation signals,and then the fault components are separated by improving the independent component analysis algorithm to highlight the fault characteristic information.In view of the shortcomings of ICA learning algorithm,such as high requirements for initial value selection,easy to fall into local extremum,and the need for formula derivation in advance,this paper introduces the roulette idea and proposes the independent component analysis method of adaptive inertial weight particle swarm optimization(AIWPSO-ICA),so as to optimize the learning algorithm,make it jump out of local extremum and improve the separation performance.(3)Finally,combining with PCFDDTW and AIWPSO-ICA,a complete gear fault feature extraction method is designed,and the gear root crack fault and partial broken tooth fault signals are collected and analyzed on the experimental platform,which verifies the effectiveness of the proposed method in this paper.