A Fault Diagnosis Method Based on Improved Adaptive Filtering and Joint Distribution Adaptation

Abstract

In the real environment of industrial equipment, the vibration signals of essential components show deviations due to the fault and noise. Notably, the noise in the signal will interfere with the diagnosis process of the signal and reduce the accuracy of fault diagnosis. Based on the above problem, adaptive filtering (AF) is used as an excellent method to attenuate noise without specifying the noise type. However, how to define the most appropriate length and type of morphological filter element is the most inherent problem which needs to be solved first. This paper proposed a cooperative diagnosis method of rolling bearings vibration signal based on improved adaptive filtering and joint distribution adaptation (JDA). First, the kurtosis under different element types and lengths is calculated as an index. The structural element corresponding to maximum kurtosis is selected as the most suitable morphological filter element because the different morphological filter elements reflect the effect of feature extraction. Then, JDA aims to improve both the marginal distribution and the conditional distribution to solve the chaotic distribution of time-domain features under variable working conditions. Finally, the improved least squares support vector machine (LSSVM) verified the effectiveness and improvement of the proposed method under bearing acceleration signal. At the same time, the comparative experiment proved that the proposed method not only directly corrects the most appropriate elements greatly optimizes the feature structure, but also enhances the accuracy of fault diagnosis. INDEX TERMS adaptive filters, morphological operation, time domain features, transfer learning, joint distribution adaptation, LSSVM, fault diagnosis.

Deep Transfer Network With Multi-Kernel Dynamic Distribution Adaptation for Cross-Machine Fault Diagnosis

“…We compare the transfer results and transfer performances of the presented method with those state-of-the-art machine learning methods including CNN, transfer component analysis (TCA), joint distribution adaptation (JDA) [31], geodesic flow kernel (GFK) [32], deep domain confusion (DDC), deep adaptation network (DAN) and DTN with DDA. For comparison, the structural parameters of CNN are the same as DTN.…”

Research on a Rolling Bearing Fault Detection Method With Wavelet Convolution Deep Transfer Learning

“…The training and transfer methods are consistent with the proposed method. Two deep transfer learning methods are selected: the Transfer Component Analysis (TCA) [32] and the Joint Distribution Adaptation (JDA) [33]. To further demonstrate the effectiveness of classification, the dataset is used to test all models.…”

A Novel Domain Adaptation-Based Intelligent Fault Diagnosis Model to Handle Sample Class Imbalanced Problem

“…Han et al [21] introduced the joint distribution adaptation (JDA) into a deep transfer network (DTN) to avoid negative adaptation and presented smooth convergence for fault diagnosis in industry applications. The discussion in [12] showed that joint distribution adaptation may obtain better performance in fault diagnosis by reweighting the source instance on the basis of its correlation with the target instance to reduce the cross-domain feature distribution discrepancy [22,23]. Chen et al [23] developed an unsupervised domain adaptation approach to reduce the domain shifts between the data gathered from the experimental platform and the operating platform of the rotating machine by aligning the features extracted from the two data domains.…”

“…Many existing DA approaches usually aim to reduce the difference of cross-domain feature distributions, e.g., the distribution adaptation, the instance reweighting, or joint matching (join the distribution adaptation and instance reweighting). The distribution adaptation approaches [ 12 , 13 , 14 ] mainly include marginal adaptation (MDA) [ 15 , 16 , 17 , 18 , 19 ], conditional adaptation (CDA) [ 20 ], or both [ 12 , 21 ] and are applied for most distribution adaptation approaches. Lu et al [ 15 ] adapted the marginal distribution by MMD to minimize the distribution discrepancy across domains and introduced MMD into a deep neural network (DNN).…”

“…The class imbalance will lead to a substantially large cross-domain distribution difference and usually exists in many domain adaptation scenarios. Unfortunately, the class imbalance is usually ignored for most DA approaches [ 12 , 14 ]. They usually assume the sample classes are balanced or tackle the sample bias for one domain, which decreases the validity of DA.…”