Multi-Scale Recursive Semi-Supervised Deep Learning Fault Diagnosis Method with Attention Gate

Abstract

The efficiency of deep learning-based fault diagnosis methods for bearings is affected by the sample size of the labeled data, which might be insufficient in the engineering field. Self-training is a commonly used semi-supervised method, which is usually limited by the accuracy of features for unlabeled data screening. It is significant to design an efficient training mechanism to extract accurate features and a novel feature fusion mechanism to ensure that the fused feature is capable of screening. A novel training mechanism of multi-scale recursion (MRAE) is designed for Autoencoder in this article, which can be used for accurate feature extraction with a small amount of labeled data. An attention gate-based fusion mechanism was constructed to make full use of all useful features in the sense that it can incorporate distinguishing features on different scales. Utilizing large numbers of unlabeled data, the proposed multi-scale recursive semi-supervised deep learning fault diagnosis method with attention gate (MRAE-AG) can efficiently improve the fault diagnosis performance of DNNs trained by a small number of labeled data. A benchmark dataset from the Case Western Reserve University bearing data center was used to validate this novel method which shows that 7.76% accuracy improvement can be achieved in the case when only 10 labeled samples was available for supervised training of the DNN-based fault diagnosis model.

Deep Learning based Approaches for Intelligent Industrial Machinery Health Management & Fault Diagnosis in Resource-Constrained Environments

“…Neural Networks [104], [53], [105], [106], [107], [108], [109], [110], [111], [112], [113], [114], [115], [116], [71] 2023 Convolution Neural Networks [117], [105], [54], [118], [106], [107], [119], [120], [121], [122], [108], [110], [111], [112], [123], [115], [116] 2023 Recurrent Neural Networks [124], [125], [126], [127], [128], [129], [130], [131] 2023 LSTM Methods [132], [133], [134], [135] 2023 Transformer-based Methods [136], [137], [138], [139] 2023 Auto Encoders [140],…”

“…Zhiqiang et al presented a novel method for fault diagnosis of rolling bearings under multiple working conditions using a deep neural network (DNN) with an attention gate and a multiscale recursive fusion strategy. The method aims to overcome the information loss problem in DNN and to extract the potential features related to working condition changes [127].…”

A Parallel Ensemble Learning Model for Fault Detection and Diagnosis of Industrial Machinery

“…We conducted two additional empirical studies to further evaluate the effectiveness of EMDL-FDD as compared with other state-of-the-art methods for FDD. Table 9(a) presents a comprehensive comparison of EMDL-FDD with six state-of-the-art techniques on the CWRU data set, namely a stacked denoising autoencoder (SDAE) [40], hierarchical CNN (HCNN) [41], multi-scale recursive semi-supervised DL method (MRAE-AG) [16], hierarchical adaptive CNN (HACNN) [42], ensemble DBN (EDBN) [43], and predictive maintenance with CNN (PdM-CNN) [9]. Among these six methods, HACNN produces the highest accuracy rate of 97.9%, as shown in Table 9(a).…”

“…Method Accuracy Method Accuracy SDAE [40] 91.79% FANN [44] 93.5% HCNN [41] 92.60% KE-ANN [45] 95.8% MRAE-AG [16] 94.38% SBM [34] 96.4% HACNN [42] 97.90% SMOTE-DNN [46] 97.0% EDBN [43] 96.95% ISBM [47] 98.5% PdM-CNN [9] 97.3% PdM-CNN [9] 99.58% EMDL-FDD 98.45% EMDL-FDD 99.79%…”

“…The use of ML/DL for identifying and diagnosing failures in industrial motors has become more prevalent in recent years. In this regard, DL models have become widely used in FDD of machinery [15][16][17], as they can address the limitations of traditional ML methods in performing automated feature extraction. On the other hand, ensemble methods, which combine the strengths of multiple techniques to overcome their individual limitations, have been adopted for solving FDD problems [17,18].…”