An Adaptive Weighted Multiscale Convolutional Neural Network for Rotating Machinery Fault Diagnosis Under Variable Operating Conditions

Abstract

Extracting robust fault sensitive features of vibration signals remains a challenge for rotating machinery fault diagnosis under variable operating conditions. Most existing fault diagnosis methods based on the convolutional neural network (CNN) can only extract single-scale features, which not only loss fault sensitive information on other scales, but also suffer from the domain shift problem. In this work, a novel end-to-end deep learning network named adaptive weighted multiscale convolutional neural network (AWMSCNN) is proposed to adaptively extract robust and discriminative multiscale fusion features from raw vibration signals. The AWMSCNN consists of three main components: the denoising layer, the adaptive weighted multiscale convolutional (AWMSC) block, and the multiscale feature fusion layer. The AWMSC block can learn rich and complementary features on multiple scales in parallel. Then, an adaptive weight vector is introduced to modulate multiscale features to emphasize fault sensitive features and suppress features that are sensitive to operating conditions. The train wheelset bearing dataset and the bearing dataset provided by Case Western Reserve University (CWRU) are used to verify the superiority of the proposed model over the basic CNN and other multiscale CNN models. The experiment results show that the proposed model has strong fault discriminative ability and domain adaptive ability against variable operating conditions. INDEX TERMS Adaptive weighted multiscale feature learning, convolutional neural network, deep learning, fault diagnosis, rotating machinery, variable operating conditions.

A Deep Multi-Label Learning Framework for the Intelligent Fault Diagnosis of Machines

“…The main purpose of intelligent fault diagnosis is to use artificial intelligent techniques to process the monitored signals of the machines quickly and recognize the mechanical health conditions automatically [7]. Since deep learning allows deep neural networks to accomplish the tasks of feature extraction and fault classification automatically [8][9][10], researchers recently introduce this concept into intelligent fault diagnosis of machines and use it for deal with different tasks such as feature learning [11,12], transfer learning [13,14], imbalanced classification [15,16], fewshot learning [17][18][19] and other tasks [20,21]. For example, Oh et al [22] used multi-class deep belief networks to learn features from the vibration-images of rotor systems and recognize their health conditions.…”

Enhanced Lightweight Multiscale Convolutional Neural Network for Rolling Bearing Fault Diagnosis

“…On the other hand, the impact segment containing pulse components in vibration signals reflect the fault behavior of rolling bearings. The frequency of the pulse components caused by faults of different positions and severity levels varies greatly, which result in the features that sensitive to different faults are distributed on different time scales of vibration signals [27,28]. Therefore, vibration signals usually exhibit multiscale characteristics and contain intricate patterns on multiple time scales [29][30][31].…”

“…In MK-ResCNN, convolutional kernels with different sizes were utilized to extract multiscale features from vibration signals in parallel, and identity mapping and residual mapping were introduced to overcome the degradation problem caused by deep networks. In [27], a multiscale feature extraction method similar to the MK-ResCNN was adopted. Besides, an adaptive weight vector was introduced to emphasize the scale feature sensitive to faults.…”

A Bi-Level Nested Sparse Optimization for Adaptive Mechanical Fault Feature Detection

“…A primary issue in the FDI procedure is to detect feature information from noisy measurements and many advanced signal processing methods have been developed [7], such as spectrum analysis [8], time-frequency analysis [9], wavelet transform (WT) [10], spectrum kurtosis(SK) [11], adaptive mode decomposition [12], cyclostationary descriptors [13], and deep learning [14], [15], etc. In recent years, sparsity representation based fault diagnosis (SRFD) techniques have been one of the hottest topics in the signal processing society and aroused extensive interests.…”