Search

Induction motors are important equipment in modern industry. However, the occurrence of fatigue failure following an extended period of operation invariably results in a catastrophic failure. As a result, monitoring and diagnosing induction motors is critical to avoiding unplanned shutdowns caused by premature failures. This article aims to develop an effective method for motor fault detection using time–frequency contents of vibration signals and an attention-based convolutional neural network model. First, the vibration signals are collected and labeled into five different categories: normal condition, outer ring fault, inner ring fault, misalignment condition, and broken rotor bar. Then, using the Morlet function, continuous wavelet transform (CWT) converts the vibratory time-series signals to the scalogram feature images. The time–frequency feature images are created after downsampling and converting the measured vibration signals to the frequency domain. These images are then resized and fed into the proposed convolutional attention neural network (CANN) to identify various induction motor failures. The experimental results demonstrate that the suggested model can provide an excellent diagnosis accuracy of 99.43%, significantly better than the state-of-the-art deep learning approaches for fault diagnosis. Moreover, the developed model’s robustness is validated against adversarial attacks based on the fast gradient sign method (FGSM) by including white Gaussian noise. [ABSTRACT FROM AUTHOR]

A transfer learning-based deep convolutional neural network (DCNN) is employed in this article to identify several typical electrical substation equipment incipient faults. Image dataset contains 11 000 captured electrical substation equipment images, and each image inside the dataset was labeled as normal condition or typical incipient faults (insulation oil leakage, insulator contamination, rusting, and paint off). High-dimension features of electrical substation equipment images were first extracted by classical pre-trained DCNN architectures through the transfer learning method, and different incipient faults were then classified by the fully-connected (FC) neural network using the SoftMax activation function. Remarkable fault classification accuracy was obtained on the validation image, which verifies the effectiveness of the purposed method. Performances of various pre-trained classic DCNN architectures were explored and evaluated by t-distributed stochastic neighbor embedding (t-SNE) feature cluster maps, learning curves, and confusion matrixes. Results show that a model consists of MobileNetV2 DCNN architecture and two FC neural network layers could finalize fault classification tasks on 1000 images in 25 s with an accuracy of 98% which achieves better performance than traditional image classify methods. The proposed method could be useful for designing an electrical substation image monitoring system thus providing early prediction of incipient equipment faults. [ABSTRACT FROM AUTHOR]

Fault classification is vital in smart manufacturing, and convolutional neural network (CNN) has been widely applied in fault classification. But the performance of CNN heavily depends on its learning rate. As the default setting on learning rate cannot guarantee its performance, the learning rate tuning process becomes essential. However, the traditional learning rate tuning methods either cost much time consumption or rely on the experts’ experiences, so it is a considerable barrier for the users. To overcome this drawback, this article proposes a CNN with automatic learning rate scheduler (AutoLR-CNN) for fault classification. First, the long short-term memory (LSTM) is used to extract the features of the past loss of CNN. Then, an agent based on deep deterministic policy gradient (DDPG) is trained to automatically control the learning rate for CNN online. Third, the double CNN structure is developed to enhance the stability of the proposed method. The proposed AutoLR-CNN is tested on two famous bearing data sets and a practical bearing data set on wind turbine. The results of AutoLR-CNN are superior to six commonly used baseline learning rate schedulers in Tensorflow. AutoLR-CNN is also compared with other reported machine learning and deep learning methods. The results show that AutoLR-CNN has achieved the state-of-the-art performance in fault classification. [ABSTRACT FROM AUTHOR]

Stacked auto-encoder (SAE)-based deep learning has been introduced for fault classification in recent years, which has the potential to extract deep abstract features from the raw input data. However, SAE cannot ensure the relevance of deep features with the fault types due to its unsupervised self-reconstruction in the pretraining stage. To overcome this problem, a stacked supervised auto-encoder is proposed to pretrain the deep network and obtain deep fault-relevant features from raw input data. In each supervised auto-encoder, informative features are learned from the input data with the goal that they can largely distinguish different fault types. By stacking multiple supervised auto-encoders hierarchically, high-level fault-relevant features are gradually learned from raw input data, which can improve the classification accuracy of the classifiers. The proposed SSAE is tested on the Tennessee–Eastman (TE) benchmark process and a real industrial hydrocracking process. The results show the effectiveness and flexibility of SSAE. • A stacked supervised auto-encoder is proposed to pretrain deep network and obtain deep fault-relevant features. • In each supervised auto-encoder, informative features are learned from the input data with the goal that they can largely distinguish different fault types. • High-level fault-relevant features are gradually learned from raw input data by hierarchically stacking multiple supervised auto-encoders. • High classification performance of the proposed method is validated on TE process and an industrial hydrocracking process. [ABSTRACT FROM AUTHOR]

Fault classification is important for the fault cause analysis and faster power supply restoration. A deep-learning-based fault classification method in small current grounding power distribution systems is presented in this paper. The current and voltage signals are sampled at a substation when a fault occurred. The time-frequency energy matrix is constructed via applying Hilbert–Huang transform (HHT) band-pass filter to those sampled fault signals. Regarding the time-frequency energy matrix as the pixel matrix of digital image, a method for image similarity recognition based on convolution neural network (CNN) is used for fault classification. The presented method can extract the features of fault signals and accurately classify ten types of short-circuit faults, simultaneously. Two simulation models are established in the PSCAD/EMTDC and physical system environment, respectively. The performance of the presented method is studied in the MATLAB environment. Various kinds of fault conditions and factors including asynchronous sampling, different network structures, distribution generators access, and so on are considered to verify the adaptability of the presented method. The results of investigation show that the presented method has the characteristics of high accuracy and adaptability in fault classification of power distribution systems. [ABSTRACT FROM AUTHOR]

The paper presents a method using deep learning framework based on convolution neural network (CNN), for identification and localization of faults of transformer winding under impulse test. The results show that the proposed method outperforms the existing methods significantly. The present scheme eliminates the requirement of separate feature extraction and classification algorithms for the analysis of fault current patterns. A part of the proposed network performs feature learning and the other part classifies the features in a supervised manner. The method is computation intensive but capable of achieving very high degree of accuracy; on an average a margin of more than 7% compared to other published literature till date. [ABSTRACT FROM AUTHOR]

Since industrial process data often presents strong correlations, high complexity, and nonlinear patterns, a proficient deep learning model is required for the fault classification task. Recent researches have shown that deep learning models like stacked autoencoder (SAE) are able to learn deep abstract features from complex process data. Nevertheless, a traditional SAE cannot extract the informative fault-relevant features and data distribution features from industrial process data, which are necessary for effective fault classification in industrial processes. Thus, this study proposes a semi-supervised Inter-Relational Mahalanobis SAE (IRM-SAE) model to learn inter-relational distribution and fault-relevant dynamic features of process data for fault classification. First, the Inter-Relational Mahalanobis loss is introduced into the original objective function of SAE to learn meaningful inter-relational distribution features within the data. Then, an active time frame technique is developed to preprocess the input data to capture the dynamic features of the data. Furthermore, to fully utilize both labeled and unlabeled data in industrial processes, the semi-supervised strategy is introduced to learn fault-related features for better fault classification. To validate the performance of the proposed model, it is applied on the Tennessee–Eastman process and a real-world industrial hydrocracking process. The experimental results show that the proposed model has higher fault classification performance compared to other deep learning models. • A novel semi-supervised Inter-Relational Mahalanobis SAE (IRM-SAE) model is proposed to learn inter-relational distribution and fault-relevant dynamic features of process data. • An active time frame technique is developed to preprocess the process data to capture the dynamic features of the data. • To fully utilize both labeled and unlabeled data in industrial processes, the semi-supervised strategy is introduced to learn fault-related features for better fault classification. • The experimental results on two industrial processes show that the proposed model has higher fault classification performance compared to other deep learning models. [ABSTRACT FROM AUTHOR]

• A new deep learning method called Sparse Probabilistic Dynamic Network (SPDN) is proposed, which is based on Gaussian-Bernoulli Restricted Boltzmann Machine and the Long-Short Term Memory. • The proposed method takes into account the nonlinear and dynamic characteristics in process data, and deeply extracts the potential information in it. • The proposed method can obviously improve the classification accuracy in the case study. The dynamic and nonlinear characteristics of process data have become the major issue in data-driven process monitoring. Traditional data-driven methods are often only able to extract a single feature in process data. Therefore, how to effectively extract multi-dimensional features has become the focus of current research. Sparse probabilistic dynamic network (SPDN) is a deep learning model proposed in this paper for the purpose of fault classification. The method mainly takes the advantages of the sparse Gaussian-Bernoulli Restricted Boltzmann Machine (GRBM) and the recurrent neural network (RNN) with long-short term memory (LSTM) units. First, the sparse GRBM is used for nonlinear feature extraction in an unsupervised way. Then, LSTM is introduced to realize the modeling of sequence data which can effectively handle the dynamic feature of the data. In the Tennessee-Eastman benchmark process, the classification accuracies of the proposed method are proved to be far superior to MLP, RNN and PDN. Meanwhile, in order to prove the influence of the data dynamics and the internal parameters of the structure on the fault classification results, two additional experiments were carried out. [Display omitted] [ABSTRACT FROM AUTHOR]

In this paper, we explore learning methods to improve the performance of the open-circuit fault diagnosis of modular multilevel converters (MMCs). Two deep learning methods, namely, convolutional neural networks (CNN) and auto encoder based deep neural networks (AE-based DNN), as well as stand-alone SoftMax classifier are explored for the detection and classification of faults of MMC-based high voltage direct current converter (MMC-HVDC). Only AC-side three-phase current and the upper and lower bridges' currents of the MMCs are used directly in our proposed approaches without any explicit feature extraction or feature subset selection. The two-terminal MMC-HVDC system is implemented in Power Systems Computer-Aided Design/Electromagnetic Transients including DC (PSCAD/EMTDC) to verify and compare our methods. The simulation results indicate CNN, AE-based DNN, and SoftMax classifier can detect and classify faults with high detection accuracy and classification accuracy. Compared with CNN and AE-based DNN, the SoftMax classifier performed better in detection and classification accuracy as well as testing speed. The detection accuracy of AE-based DNN is a little better than CNN, while CNN needs less training time than the AE-based DNN and SoftMax classifier. [ABSTRACT FROM AUTHOR]