Self-supervised feature extraction via time–frequency contrast for intelligent fault diagnosis of rotating machinery

Liu, Yang; Wen, Weigang; Bai, Yihao; Meng, Qingzhou

doi:10.1016/j.measurement.2023.112551

Cited by 13 publications

(5 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The calculation equation for the output 𝑂 𝑡 for attention module is shown in (16): where 𝑂 𝑡 = (𝑂 1 , 𝑂 2 , ⋯ , 𝑂 𝑇 ) output from the time attention mechanism and the historical target series Y = (𝑦 1 , 𝑦 2 , ⋯ , 𝑦 𝑇 ) are spliced together as the input of the decoder. The newly calculated time series 𝑌 ̃𝑡 = (𝑦 ̃1, 𝑦 ̃2, ⋯ , 𝑦 ̃𝑇) is obtained by (17), and the hidden layer state in the decoder is updated by (18). Finally, the GRUs connected in series are used to predict the time series at 𝑇 + 1, as shown in (19):…”

Section: Time Attention Mechanismmentioning

confidence: 99%

“…From the perspective of deep learning, compared with one-dimensional convolution, the two-dimensional convolution mode can make the network learn more information [17]. Liu et al [18] converted the original vibration signal into a twodimensional time-frequency image (TF-image) by a kind of differential continuous wavelet transform and used this image as the input of the network to realize the fault diagnosis of gears. Tong and Zhang et al [19], [20] converted the signal to GAF to perform fault diagnosis of rolling bearings to utilize the time domain information of the vibration signal fully.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fault Diagnosis Method for Imbalanced Data of Rotating Machinery Based on Time Domain Signal Prediction and SC-ResNeSt

et al. 2023

View full text Add to dashboard Cite

In an actual engineering environment, some rotating machines are usually in normal operation, but their time in a fault state is very short, which leads to a serious imbalance in the fault diagnosis datasets for rotating machinery, and gives the traditional network model the shortcomings of poor stability and low accuracy in practical engineering applications. To solve this problem, we propose a fault diagnosis method based on the combination of a new Dual-stage Attention-based Recurrent Neural Network (DA-RNN) and depth residual dispersion self-calibration convolution network (SC-ResNeSt). Firstly, a novel DA-RNN network with a gated cycle unit (GRU) as a coding-decoding unit was designed, and the network was used to predict and expand the scarce fault signals. Secondly, to make full use of the time domain information of vibration signals, a new image coding method, namely, Gram Angle Product Field (GAPF), was proposed. Then, because the traditional convolution layer lacks a dynamic receptive field to extract more representative features, self-calibrated convolution modules were introduced on the basis of the distraction network (ResNeSt), and a new network model, SC-ResNeSt, was established. Finally, the expanded vibration signal is converted into GAPF, which is used as the input for the SC-ResNeSt network to classify the fault types. To check the performance of the model, the Case Western Reserve University rolling bearing dataset and planetary gearbox dataset were used for testing. Ultimately, good results were obtained in a prediction experiment for bearing fault samples and a subsequent fault diagnosis experiment for bearings and gears, which verified the feasibility and practicability of the model.

show abstract

Section: Time Attention Mechanismmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Fault Diagnosis Method for Imbalanced Data of Rotating Machinery Based on Time Domain Signal Prediction and SC-ResNeSt

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Currently, SSL has been applied to mechanical fault diagnosis [21][22][23][24], and the key lies in data augmentation. However, the utilization of traditional data augmentation may disrupt the intrinsic structure of non-stationary 1D sequences, especially those characterized by cyclic angular vibration.…”

Section: Introductionmentioning

confidence: 99%

DA-VICReg: a data augmentation-free self-supervised learning approach for diesel engine fault diagnosis

Chen,

Xiang,

Wang

2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

Self-supervised learning aims to extract useful representations from unlabeled data by maximizing the agreement between positive pairs. However, traditional self-supervised learning relies on carefully designed data augmentation methods to generate positive pairs. When dealing with 1D vibration signals, data augmentation prone to potentially compromise the fault information in the original signals. Therefore, this paper proposes a data augmentation-free self-supervised learning framework for diesel engine fault diagnosis called Domain Adaptation Variance Invariance Covariance Regularization (DA-VICReg). The DA-VICReg uses cyclic angular vibrations collected within the same time period as positive pairs and extracts useful features from unlabeled data using a loss function composed of three terms: Variance, Invariance, and Covariance. We found that when positive pairs originate from different operating conditions, such as varying speeds and torques, the model can develop feature extraction capabilities that remain unaffected by changes in operating conditions. In addition, a spatial pyramid pooling layer and a trilinear attention module are used to extract vibration features at different scales and focus on critical spatial locations and channels. Finally, the proposed approach was validated through experiments on two types of diesel engines, and a comparison with prominent self-supervised learning methods confirms the superiority of the proposed approach. In engineering practice, this method can utilize a large amount of signals stored in different time periods for self-supervised training and learn useful features for downstream fault diagnosis tasks.

show abstract

“…Deep Learning (DL) models, as one of the mainstream methods in PHM technology, utilize multi-layered neural network structures to perform hierarchical abstraction of input data, automatically extracting complex features. This effectively compensates for the drawback of manual feature extraction methods, which lack the ability to adaptively learn features [4][5][6]. The mainstream DL models in PHM include deep belief network, autoencoder (AE), convolutional neural network (CNN), recurrent neural network, and generative adversarial network (GAN) [7].…”

Section: Introductionmentioning

confidence: 99%

Attention features selection oversampling technique (AFS-O) for rolling bearing fault diagnosis with class imbalance

Han,

Wang,

Shen

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

When using data-driven methods for fault diagnosis of mechanical rotating components such as gears and bearings, there is a problem of class imbalance in the lifecycle data collected by sensors. The most commonly used method to address this issue is the synthetic minority over-sampling technique, which synthesizes samples in the feature space, but its blind synthesis may lead to redundant features in the synthetic samples. To avoid this deficiency, this paper proposes a feature-weighted oversampling method called AFS-O (Attention Features Selection Oversampling Technique). First, time–frequency domain features are extracted from the full lifecycle data of bearings to construct an initial subset of features, which serves the input to AFS. Then, AFS is then used to obtain the distribution of feature selection patterns and generate feature weights to determine the inclusion or exclusion of each feature, thereby constructing an optimal subset of features. Finally, the optimal feature subset is synthetically oversampled to achieve class-balanced data, which is then fed into a classifier. AFS-O is applied to the rolling bearing accelerated lifetime dataset from Xi’an Jiaotong University. Experimental results demonstrate that AFS-O outperforms other state-of-the-art synthetic oversampling algorithms in terms of Gmean, F 2 score , and Recall, confirming the effectiveness of the proposed method.

show abstract

Self-supervised feature extraction via time–frequency contrast for intelligent fault diagnosis of rotating machinery

Cited by 13 publications

References 22 publications

Fault Diagnosis Method for Imbalanced Data of Rotating Machinery Based on Time Domain Signal Prediction and SC-ResNeSt

Fault Diagnosis Method for Imbalanced Data of Rotating Machinery Based on Time Domain Signal Prediction and SC-ResNeSt

DA-VICReg: a data augmentation-free self-supervised learning approach for diesel engine fault diagnosis

Attention features selection oversampling technique (AFS-O) for rolling bearing fault diagnosis with class imbalance

Contact Info

Product

Resources

About