Search

This paper introduces a sophisticated approach for identifying and categorizing broken rotor bars in direct torque-controlled (DTC) induction motors. DTC is implemented in industrial drive systems as a suitable control method to preserve torque control performance, which sometimes shows its impact on fault-representing frequencies. This is because of the DTC's closed-loop control nature, whichtriesto reduce speed and torque ripples by changing the voltage profile. The proposed model utilizes the modified Shapley Additive exPlanations (SHAP) technique in combination with gradient-boosting decision trees (GBDT) to detect and classify the abnormalities in BRBs at diverse (0%, 25%, 50%, 75%, and 100%) loading conditions. To prevent overfitting of the proposed model, we used the adaptive fold cross-validation (AF-CV) technique, which can dynamically adjust the number of folds during the optimization process. By employing extensive feature engineering in the original dataset and then applying Shapely Additive exPlanations(SHAP)-based feature selection, our methodology effectively identifies informative features from signals (three-phase current, three-phase voltage, torque, and speed) and motor characteristics. The gradient-boosting decision tree (GBDT) classifier, trained using the given characteristics, extracts consistent and reliable classification performance under different loading circumstances and enables precise and accurate detection and classification of broken rotor bars. The proposed approach (SHAP-Fusion GBDT with AF-CV) is a major advancement in the field of machine learning in detecting motor anomalies at varying loading conditions and proved to be an effective mechanism for preventative maintenance and preventing faults in DTC-controlled induction motors byattaining an accuracy rate of 99% for all loading conditions. [ABSTRACT FROM AUTHOR]

This study introduces an innovative approach to diagnostics, employing a unique combination of techniques including a stratified group K-fold cross-validation method and a sparse stacked autoencoder (SSAE) alongside LightGBM. By examining signatures derived from motor current, voltage, speed, and torque, the framework aims to effectively detect and classify broken rotor bars (BRBs) within inverter-fed induction machines. In this kind of cross-validation method, class labels and grouping factors are spread out across folds by distributing motor operational data attributes equally over target label stratification and extra grouping information. By integrating SSAE and LightGBM, a gradient-boosting framework, we elevate the precision and efficacy of defect diagnosis. The SSAE feature extraction algorithm proves to be particularly effective in identifying small BRB signatures within motor operational data. Our approach relies on comprehensive datasets collected from motor systems operating under diverse loading conditions, ranging from 0% to 100%. Using a sparse stacked autoencoder, the model lowers the dimensionality and noise of the motor fault data. It then sends the cleaned data to the LightGBM network for fault diagnosis. LightGBM leverages the attributes of the sparse stacked autoencoder to showcase the distinctive qualities associated with BRBs. This integration offers the potential to improve defect identification by furnishing input representations that are both more precise and more concise. The proposed model (SSAE with LightGBM) was trained using 80% of the data, while the remaining 20% was used for testing. To validate the proposed architecture, we evaluate the accuracy, precision, recall, and F1-scores of the results using motor global signals, with the help of confusion matrices with receiver operating characteristic (ROC) curves. Following the training of a new LightGBM model with refined hyperparameters through Bayesian optimization, we proceed to conduct the final classification utilizing the optimal feature subset. Evaluation of the test dataset indicates that the BRBs diagnostic framework facilitates the detection and classification of issues with induction motor BRBs, achieving accuracy rates of up to 99% across all loading conditions. [ABSTRACT FROM AUTHOR]

This paper presents a supervised machine learning approach using eight popular classifiers for fault classification in power transmission lines. The classification of faults, indicated by the behavior of the electrical signals associated with them, plays a pivotal role in maintaining the reliability, stability, and security of electrical grids. However, most of the previous studies on fault analysis in power systems relied on proprietary data and lack of standardized benchmarks, hindering the comparison of algorithms and making performance more erratic. Moreover, the nonavailability of labeled data for all types of faults is the most problematic. This paper proposes to perform fault classification on a data set created from a real-time Simulink model to standardize performance and advance research in this area. A new strategy for non-intensive feature extraction is applied using relatively simpler techniques, eliminating computationally expensive techniques such as wavelets. Feature selection through dimensionality reduction techniques is used to improve model performance and more efficient use of computational resources. The performance of the learning algorithms (e.g. Decision Tree, Random Forest, etc.) has been analyzed with various preprocessing techniques (e.g. data scaling, transformation, etc.) and tuning of parameters, focusing on the accuracy and computational time ( $T_{c}$ ), for performance generalization and efficiency. Performing specific operations on data in sequence of steps provided flexibility and adaptability in processing the data, making it easy to train, evaluate, and validate the learning algorithms. The results demonstrated that the proposed scheme can be effectively used for fault classification with high accuracy and significant reductions in $T_{c}$ under various operating conditions. The study also determined the best estimator for each classifier when building and training the classifier models, offering a variety of options. Logistic Regression, Random Forest and Support Vector Machine were the outperforming classifiers and proved their potential for classifying faults in electric power transmission lines.

To address the issue of diagnosing hard and soft faults in active power factor correction (APFC) power supply, this study analyzes failure modes resulting from aging and malfunction of various sensitive components. The power fault waveform patterns are initially analyzed based on the circuit's THD, current ripple value, and RMS value. The inductor current signals in different fault modes are then utilized to extract and construct time–frequency fusion fault features of the APFC power supply. Finally, these feature quantities are downscaled and optimized using the RF algorithm. The SOA‐KELM model of the APFC converter is proposed, and the feature vectors under different fault modes are used to classify and diagnose faults, achieving hard and soft fault detection of the converter. The experiments show that the method achieves 100% accuracy for hard fault diagnosis and 96.36% accuracy for soft fault diagnosis of the converter, demonstrating high diagnostic accuracy. [ABSTRACT FROM AUTHOR]

This paper proposes a fast protection scheme of a hybrid AC/DC distribution network based on features extraction of the decomposed signals. The signal decomposition is achieved using discrete wavelets transform (DWT). The proposed scheme offers three main functions, fast fault detection for both network sides, AC fault classification with the ability to identify the faulted phase/ phases, and protection coordination between AC and DC sides. In addition, the scheme is based on local measurement so no need for extra-cost for communication infrastructure. The absence of communication reduces the overall cost and, as a most important feature, it reduces the fault clearing time. The system is thoroughly analyzed under many types of faults in both AC and DC sides. Also, different disturbance conditions are considered which used in threshold determination. The simulation results show that the proposed protection scheme was not only able to protect the system but also it offers a generalized fast acting communication-less approach. [ABSTRACT FROM AUTHOR]

This paper introduces a sophisticated approach for identifying and categorizing broken rotor bars in direct torque-controlled (DTC) induction motors. DTC is implemented in industrial drive systems as a suitable control method to preserve torque control performance, which sometimes shows its impact on fault-representing frequencies. This is because of the DTC’s closed-loop control nature, whichtriesto reduce speed and torque ripples by changing the voltage profile. The proposed model utilizes the modified Shapley Additive exPlanations (SHAP) technique in combination with gradient-boosting decision trees (GBDT) to detect and classify the abnormalities in BRBs at diverse (0%, 25%, 50%, 75%, and 100%) loading conditions. To prevent overfitting of the proposed model, we used the adaptive fold cross-validation (AF-CV) technique, which can dynamically adjust the number of folds during the optimization process. By employing extensive feature engineering in the original dataset and then applying Shapely Additive exPlanations(SHAP)-based feature selection, our methodology effectively identifies informative features from signals (three-phase current, three-phase voltage, torque, and speed) and motor characteristics. The gradient-boosting decision tree (GBDT) classifier, trained using the given characteristics, extracts consistent and reliable classification performance under different loading circumstances and enables precise and accurate detection and classification of broken rotor bars. The proposed approach (SHAP-Fusion GBDT with AF-CV) is a major advancement in the field of machine learning in detecting motor anomalies at varying loading conditions and proved to be an effective mechanism for preventative maintenance and preventing faults in DTC-controlled induction motors byattaining an accuracy rate of 99% for all loading conditions.

As the core components of electric vehicles, the safety of the electric system, including motors, batteries, and electronic control systems, has always been of great concern. To provide early warning of electric-system failure and troubleshoot the problem in time, this study proposes a novel energy-vehicle electric-system failure-classification method, which is named Pearson-ShuffleDarkNet37-SE-Fully Connected-Net (PSDSEF). Firstly, the raw data were preprocessed and dimensionality reduction was performed after the Pearson correlation coefficient; then, data features were extracted utilizing ShuffleNet and an improved DarkNet37-SE network based on DarkNet53; secondly, the inserted squeeze-and-excitation networks (SE-Net) channel attention were able to obtain more fault-related target information; finally, the prediction results of the ShuffleNet and DarkNet37-SE networks were aggregated with a fully connected neural network to output the classification results. The experimental results showed that the proposed PSDSEF-based electric vehicles electric-system fault-classification method achieved an accuracy of 97.22%, which is better than other classical convolutional neural networks with the highest accuracy of 92.19% (ResNet101); the training time is faster than the average training time of the comparative networks. The proposed PSDSEF has the advantage of high classification accuracy and small number of parameters. [ABSTRACT FROM AUTHOR]

To solve the problems of low accuracy, difficult diagnosis and long time consuming of the current fan bearing fault diagnosis, an improved grey wolf optimization (IGWO) algorithm and a support vector machine (SVM) fault diagnosis method are proposed. In order to accurately extract the fault features, the wavelet packet decomposition method in the time-frequency domain analysis is used to extract the fault vibration signal. Take the wavelet packet decomposition energy of the eight frequency bands as the fault feature， the eigenvectors are constructed. Then, the fault model of SVM is established and the parameters of the SVM model are optimized by the IGWO algorithm, which avoids the defects of local optimum and slow convergence. According to the experimental analysis result, the average fault recognition rate of the IGWO algorithm is up to 99.3%, and it can identify fault types more quickly, more efficiently and more accurately, which provides a good support for the development of fault diagnosis.

The presented paper introduces an accurate approach for detecting and classifying parametric or soft faults that affect analog integrated circuits. This technique is based on the use of machine learning algorithm to improve the accuracy and the performance of fault classification process. To achieve this, the real and imaginary frequency responses of output voltage and supply current of the circuits under test (CUT) are used to extract features for both normal and faulty cases. These features are then exploited to train machine learning classifiers, from which the selected one among its equivalents is the quadratic discriminant classifier since it allowed the highest average accuracy score. The faults to be investigated are parametric ones affecting resistors and capacitors values. The proposed approach is validated using three filters circuits that are Sallen-Key band-pass filter, four op-amp biquad high-pass filter, and a leapfrog filter circuit. Obtained results indicate a high classification average accuracy for all circuits that are undergone testing. The proposed approach has provided a highest classification accuracy level comparing to other research works. [ABSTRACT FROM AUTHOR]

This paper presents an efficient approach to estimate the failures of various components in an induction motor using motor current signature analysis. Conventional sensor-based fault detection methods lead to huge manpower and require greater number of sensors. To overcome these drawbacks, current signature base fault detection is proposed. An advanced spectral analysis, namely discrete wavelet transform (DWT), is used for frequency domain analysis. This paper also presents fault severity estimation using feature extraction-based evaluation of DWT coefficients. As the DWT gives many coefficients at higher level decomposition which is essential for high resolution, fault classification and severity index become challenging. To address this issue, artificial neural network (ANN) algorithm is used after DWT decomposition. The fault severity is predicted by proposed fault indexing parameter of various features like energy, standard deviation, skewness, variance, RMS values. Conventional algorithms like support vector machine, k‐nearest neighbour, local mean decomposition-singular value decomposition and extreme learning machine have given maximum of 98–99% accuracy, Whereas the proposed DWT-based ANN has given 100% accuracy with tanh function. Moreover, the testing loss with this function is also very less. Experimental results have affirmed the accuracy of proposed fault detection of various faults in induction motor of rating 3—Phase, 1. 5KW, 440 V and 50 Hz. [ABSTRACT FROM AUTHOR]

Fault detection and classification is a crucial issue in modern industrial processes for ensuring steady operation and high product quality. The process data collected and stored fully reflect the equipment running state and the production process. Moreover, the extracted nonlinear features can directly affect the effectiveness of the data‐driven fault classification model. In this paper, a novel fault classification method based on nonlinear feature extraction using reconstructed distance‐based discriminant locality preserving projection (RD‐DLPP) is proposed. First, a hypersphere model for each class of data is developed according to the spatial structures and classes information in high‐dimensional space. The hyperspheres are used as indicators to evaluate the discriminatory difficulty of samples. Second, the constraints of the correlations between the k‐nearest neighbour points of the sample and the hypersphere are introduced, which can efficiently reconstruct new measure metrics between the sample and its k‐nearest neighbour points. Finally, an improved fault classification model based on RD‐DLPP is established for the construction of the highly discriminant subspace. The Bayesian decision is then used to classify the samples. The feasibility and efficiency of the proposed method are verified by the Tennessee Eastman process as a case study. [ABSTRACT FROM AUTHOR]

Fault-tolerant control of a three-phase inverter can be achieved by performing a hardware reconfiguration of the six-switch and three-phase (6S3P) topology to the four-switch and three-phase (4S3P) topology after detection and localisation of the faulty phase. Together with hardware reconfiguration, the SVPWM algorithm must be appropriately modified to handle the new 4S3P topology. The presented study focuses on diagnosing three-phase faults in two steps: fault detection and localisation. Fault detection is needed to recognise the healthy or unhealthy state of the inverter. The binary state recognition problem can be solved by preparing a feature vector that is calculated from phase currents (ia, ib, and ic) in the time and frequency domains. After the fault diagnosis system recognises the unhealthy state, it investigates the signals to localise which phase of the inverter is faulty. The multiclass classification was solved by a transformation of the three-phase currents into a single RGB image and by training a convolutional neural network. The proposed methodology for the diagnosis of three-phase inverters was tested based on a simulation model representing a laboratory test bench. After the learning process, fault detection was possible based on a 128-sample window (corresponding to a time of 0.64 ms) with an accuracy of 99 percent. In the next step, the localisation of selected individual faults was performed on the basis of a 256-sample window (corresponding to a time of 1.28 ms) with an accuracy of 100 percent. [ABSTRACT FROM AUTHOR]

In condition monitoring, accurate fault identification is an essential task for designing a proper maintenance strategy. Misalignment is one of the main faults in rotary machinery, because 70% of the failure occurs due to misalignment. Conventionally, the diagnosis of misalignment is carried out through vibration measurements. Especially, the presence of strong 2x vibration peak is generally accepted. Both angular and parallel misalignment shows peak at 2x, therefore, distinguishing misalignment type by using vibration signals alone is a difficult activity. This paper discusses classification of misalignment i.e., angular and parallel by using a diagnostic medium such as the acoustic emission and the rotor vibration signal. Vibro-acoustic sensors are used to collect data from the misaligned rotor system at two different loading, three different speed and three defect severity conditions. Time domain features are extracted and graded according to their significance using t test (One-way ANOVA) technique. Extracted features are used to train different algorithms. The outcome obtained using support vector machine (SVM) is 100% accurate. Vibro-acoustic sensor data fusion technique is employed to classify various forms of misalignment under different operating conditions. This work also intended to explore using a small amount of training data using different algorithms. The proposed method outperforms fault classification using vibration signal and acoustic signal separately. [ABSTRACT FROM AUTHOR]

The smart grid is an intelligent power system network that should be reliable and resilient for sustainable operation. Wide-Area Measurement Systems (WAMS) are deployed in the power grid to provide real-time situational awareness to the power grid operators. An excellent strategy for exploiting the WAMS data effectively is to extract relevant insights from the increasing volume of data collected. Feature extraction techniques are pivotal in developing data-driven models for power systems. This paper proposes an ensemble feature extraction method for developing intelligent data-driven models for transmission line fault detection and classification. A comparative efficacy analysis of the proposed ensemble feature extraction method is carried out with state-of-the-art feature extraction methods. The models developed and evaluated with the feature data derived with the proposed method give an accuracy of 100 % for fault detection and 99.78 % for fault classification. This method also has the advantage of significantly reducing training and testing time. Features are extracted from the WAMS data collected by simulating an IEEE 39 bus test system in the Power World simulator. [ABSTRACT FROM AUTHOR]

Recently, the concept of the microgrid (MG) has been developed to assist the penetration of large numbers of distributed energy resources (DERs) into distribution networks. However, the integration of DERs in the form of MGs disturbs the operating codes of traditional distribution networks. Consequently, traditional protection strategies cannot be applied to MG against short-circuit faults. This paper presents a novel intelligent protection strategy (NIPS) for MGs based on empirical wavelet transform (EWT) and long short-term memory (LSTM) networks. In the proposed NIPS, firstly, the three-phase current signals measured by protective relays are decomposed into empirical modes (EMs). Then, various statistical features are extracted from the obtained EMs. Afterwards, the extracted features along with the three-phase current measurement are input to three different LSTM network to obtain exact fault type, phase, and location information. Finally, a trip signal based on the obtained fault information is generated to disconnect the faulty portion from the rest of the MG. The significant feature of the proposed NIPS is that it does not need adaptive relaying and communication networks. Moreover, it is independent of the operating scenario and hence fault current magnitude. To evaluate the efficacy of the proposed NIPS, exhaustive simulations are performed on an international electro-technical commission (IEC) MG. The simulation results confirm the efficiency of the proposed NIPs in terms of accuracy, dependability, and security. Moreover, comparisons with existing intelligent protection schemes validate that the proposed NIPS is highly accurate, secure, and dependable. [ABSTRACT FROM AUTHOR]

Background: Misalignment is one of the major reasons for rotating machinery breakdown. Conventionally, misalignment diagnosis is done by the occurrence of a strong peak at 2x running speed is widely accepted. Purpose: Both angular and parallel misalignment type display peak at 2x running speed, so it is difficult to discern the form of misalignment by using vibrational signal alone. This study addresses classification of misalignment faults (parallel, angular, and combined) by using a diagnostic medium by vibration signal and acoustic emission. Methods: The vibro-acoustic sensors are utilized to capture data with different speed and defect intensity conditions for misaligned rotor system. Time domain features are extracted, and features are ranked using t test (One-way ANOVA). Ranked features are used to train different machine learning algorithms. Results: The investigation shows the vibro-acoustic sensor data fusion as an efficacious tool for misalignment fault classification. This work also intended to carry out fault classification using a small amount of training data by employing different algorithms. Conclusion: The proposed method of misalignment fault classification provides better and reliable results than fault segregation using vibration and acoustic signals separately. [ABSTRACT FROM AUTHOR]

Machine learning and other data-driven methods have developed at a prolific rate for industrial applications due to the advent of industrial big data. However, industrial datasets may not be especially well-suited to supervised learning approaches that require extensive domain knowledge in the complete and accurate labeling of datasets. To address these challenges, a semi-supervised learning approach is proposed that makes use of partially labeled subsets. The proposed methodology is applied to high-dimensional in-process measurement data, utilizing a convolutional autoencoder (CAE) for unsupervised feature extraction. A multiclass extension for semi-supervised anomaly diagnosis is proposed that utilizes principal component analysis (PCA) as the basis for anomaly scoring, and the proposed approach intersects the results of targeted one-against-all phases on partially labeled sets to classify faults. Experiments in a case study on semiconductor manufacturing measurement data are performed to explore the relationship between latent features extracted and anomaly detection performance. The application of the proposed algorithm achieves a true positive detection rate of over 90% with false positive rate under 9% for both local and global anomaly types, with these results accomplished while reducing over 99% of the original input data dimensions. In addition, the approach also allows for positive samples to be identified that were previously undetected by human experts. These results are promising for the application of the proposed semi-supervised methodology in real industrial settings. [ABSTRACT FROM AUTHOR]

Rotor bars are one of the most failure-critical components in induction machines. We present an approach for developing a rotor bar fault identification classifier for induction machines. The developed machine learning-based models are based on simulated electrical current and vibration velocity data and measured vibration acceleration data. We introduce an approach that combines sequential model-based optimization and the nested cross-validation procedure to provide a reliable estimation of the classifiers' generalization performance. These methods have not been combined earlier in this context. Automation of selected parts of the modeling procedure is studied with the measured data. We compare the performance of logistic regression and CatBoost models using the fast Fourier-transformed signals or their extracted statistical features as the input data. We develop a technique to use domain knowledge to extract features from specific frequency ranges of the fast Fourier-transformed signals. While both approaches resulted in similar accuracy with simulated current and measured vibration acceleration data, the feature-based models were faster to develop and run. With measured vibration acceleration data, better accuracy was obtained with the raw fast Fourier-transformed signals. The results demonstrate that an accurate and fast broken rotor bar detection model can be developed with the presented approach. [ABSTRACT FROM AUTHOR]

Overhead energy transmission lines are highly susceptible to failure. To deal with this problem, some researchers have proposed different preprocessing stages, which comprise mainly feature extraction, selection, and dimension reduction for fault classification in transmission lines. The common techniques applied in the preprocessing stage are the wavelet and Fourier transforms. For the classification stage, the most used method is artificial neural network. This work aims to show the use of random forest method with a simple preprocessing step based on notch filter to classify faults in transmission lines. The performance of the model was compared with that obtained by a neural network to show its efficiency. Using k-fold cross-validation to train, test, and compare the models, it was obtained the mean accuracy of 89.59% for the neural network and 91.96% for the random forest for testing data. In the validation process, it was obtained accuracy of 96.49% and 91.49% for neural network and random forest models, respectively. Although the neural network model has shown better generalization capacity, the random forest model performed about eight times faster than the neural network. [ABSTRACT FROM AUTHOR]

This study develops a hybrid method to identify, classify, and locate electrical faults on transmission lines based on Machine Learning (ML) methods. Firstly, Wavelet Transform (WT) technique is applied to extract features from the current or voltage signals. The extracted signals are decomposed into eleven coefficients. These coefficients are calculated to the energy level, and the data of teen fault types are converted to the RGB image. Secondly, GoogLeNet model is applied to classify the fault, and Convolutional Neural Network (CNN) method is proposed to locate the fault. The proposed method is tested on the four-bus power system with the 220 kV transmission line via time-domain simulation using Matlab software. The conditions of the fault resistance random values and the pre-fault load changes are considered. The simulation results show that the proposed method has high accuracy and fast processing time, and is a useful tool for analyzing the system stability in the field of electricity. [ABSTRACT FROM AUTHOR]

DC-DC converter is the core component of power conversion module of integrated modular avionics. Condition monitoring and fault diagnosis of DC-DC converter can effectively improve the reliability of avionics equipment, reduce the maintenance cost and greatly improve the use efficiency of aircraft. In this paper, firstly, a typical DC-DC converter model based on SEPIC topology is designed in PSPICE environment, and the failure modes of DC-DC converter are analyzed. Secondly, the typical fault types of DC-DC converter are simulated, and the corresponding original data are obtained through simulation. Finally, the processing framework including data preprocessing, feature extraction and selection, and multi model fusion is used to do fault classification of the DC-DC converter. The fault diagnosis of the converter is simulated. Simulation results show the effectiveness of the proposed method.

Fault-tolerant control of a three-phase inverter can be achieved by performing a hardware reconfiguration of the six-switch and three-phase (6S3P) topology to the four-switch and three-phase (4S3P) topology after detection and localisation of the faulty phase. Together with hardware reconfiguration, the SVPWM algorithm must be appropriately modified to handle the new 4S3P topology. The presented study focuses on diagnosing three-phase faults in two steps: fault detection and localisation. Fault detection is needed to recognise the healthy or unhealthy state of the inverter. The binary state recognition problem can be solved by preparing a feature vector that is calculated from phase currents (ia, ib, and ic) in the time and frequency domains. After the fault diagnosis system recognises the unhealthy state, it investigates the signals to localise which phase of the inverter is faulty. The multiclass classification was solved by a transformation of the three-phase currents into a single RGB image and by training a convolutional neural network. The proposed methodology for the diagnosis of three-phase inverters was tested based on a simulation model representing a laboratory test bench. After the learning process, fault detection was possible based on a 128-sample window (corresponding to a time of 0.64 ms) with an accuracy of 99 percent. In the next step, the localisation of selected individual faults was performed on the basis of a 256-sample window (corresponding to a time of 1.28 ms) with an accuracy of 100 percent.

The photovoltaic (PV) array is the most sensible element in PV plants, which is subject to different type of faults and defects. Thus, to keep these plants working efficiently they should be monitored and protected carefully. Some faults if they are not detected and isolated promptly they may lead to hazardous risks. The diagnosis of PV systems is widely addressed and recently machine learning (ML) and deep leaning (DL) methods drawn the attention of many researchers. Most applications of ML methods are based on the use of the I–V curves measurement, as enough information and features can be extracted from the curves, to detect and classify faults. These methods showed their capability to classify some faults, like line to line, degradation, disconnected PV modules, partial shading effect, and bypass diode faults. Another approach is based on the use of thermal or electroluminescence images of PV modules/arrays to detect and identify defects, such as hot spot, snails crack, and others. In this paper, different ML and ensemble learning (EL) methods are evaluated for fault diagnosis of PV arrays. The focus is mainly on the detection and classification of some complex faults that may affect the PV arrays, i.e., multiple faults, and faults with similar I–V curves, that are not evaluated before. The results showed the ability of the methods developed to detect faults with very good accuracy (classification rate = number of classified instances/total instances), within 99%, while the classification faults is done with an acceptable accuracy, within 81.73%. Through this study it is shown when really ML and EL methods should be used, and some recommendations, challenges and future directions in this topic are presented. [ABSTRACT FROM AUTHOR]

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]

Data-driven fault diagnosis has attracted increasing research interest with a recent trend of aiming at large-scale and complex systems. In this article, we propose a method under a probabilistic framework, named dynamic Bhattacharyya bound (DBB), to extract features for fault diagnosis. An information criterion is adopted to determine the order of dimensionality reduction and time lags when applying the proposed approach. Compared with conventional diagnostic approaches, the proposed DBB approach has several advantageous features. First, the DBB approach minimizes an upper bound of the Bayes error which is a direct manifestation of the misclassification rate. Second, pairwise Bhattacharyya bounds between different faults are summed up in the objective function, enabling it to address the fault diagnosis of multiple faults that may have large overlaps. The proposed method is validated through the Tennessee Eastman process and it shows advantageous performance than other methods such as Fisher discriminant analysis (FDA), dynamic FDA, and LP-DFDA. [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]

Rolling element bearings (REBs) are vital parts of rotating machinery across various industries. For preventing breakdowns and damages during operation, it is crucial to establish appropriate techniques for condition monitoring and fault diagnostics of these bearings. The development of machine learning (ML) brings a new way of diagnosing the fault of rolling element bearings. In the current work, ML models, namely, Support Vector Machine (SVM) and K-Nearest Neighbor (KNN), are used to classify the faults associated with different ball bearing elements. Using open-source Case Western Reserve University (CWRU) bearing data, machine learning classifiers are trained with extracted time-domain and frequency-domain features. The results show that frequency-domain features are more convincing for the training of ML models, and the KNN classifier has a high level of accuracy compared to SVM. [ABSTRACT FROM AUTHOR]

In this paper, a procedure aimed at the automatic extraction of the features from polymer electrolyte membrane fuel cell impedance spectra is proposed. An artificial neural network that is trained by exploiting the similarity learning concept has been used. The network learns the features of the impedance spectra and maps each of them into the embedding space by clustering them accurately and by emphasising differences among spectra corresponding to different faults. The siamese network structure is optimised and the quality of the learnt representation is evaluated by analysing the clusters obtained in the features space. The dataset of experimental spectra has been augmented in two different ways and the results are compared. The clustering quality of the proposed siamese network is compared with the one of other state of the art approaches. [Display omitted] • A Siamese Neural Network is used to analyse the impedance spectra of a PEM fuel cell. • The SNN is able to extract the features corresponding to each faulty condition. • Two different data augmentation approaches are used to train the SNN and their performance is analysed. • The approach is tested on a set of noisy impedance spectra. [ABSTRACT FROM AUTHOR]

Recently, graph embedding methods have been successfully used in process monitoring. To improve the discriminant power, a novel supervised graph embedding method, called uncorrelated discriminant graph embedding (UDGE), is proposed. Different from the unsupervised design of locality preserving projection (LPP), UDGE utilizes both the local geometrical structure and label information to construct the similarity between different data points. The "local geometrical structure" means that each data point can be represented as a combination of its neighbours. Due to add the uncorrelated constraint, the extracted features of UDGE are statistically uncorrelated. Uncorrelated attributes are essential for dimension reduction since they contain minimum redundancy. The application of UDGE is evaluated on the Tennessee Eastman process (TEP) benchmark. Experimental results show UDGE can better separate different types of faults and provide more promising fault diagnosis performance. The code of UDGE is released in https://github.com/htz-ecust/Uncorrelated-Discriminant-Graph-Embedding-for-Fault-Classification. [ABSTRACT FROM AUTHOR]

It is always a serious challenge to deal with the influence of noise for fault detection of rotating electrical machines under varying operating environments. To extract the most discriminative features for machine fault classification, a novel dimensionality reduction algorithm called joint sparsity and collaboration preserving projections (JSCPP) is proposed in this article, based on the platform of graph embedding framework. A joint combination of sparse representation (SR) and collaborative representation (CR) is used to construct the intrinsic and penalty graphs. SR emphasizes only on the dictionary atoms highly correlated to the query sample for local optimum while CR takes every atom into consideration for global optimum, thus applying a joint combination would make it benefit from the merits of both for graph construction. After JSCPP seeks the optimal projection directions by minimizing the intra-class compactness and maximizing the inter-class separability, a nearest neighbor classifier is then used. Through two sets of vibration data with white Gaussian noise interfered, the capability of JSCPP is demonstrated by effectively classifying different defect types and severity degrees of faulty bearings. The experimental results have also proved that the classification performance of the proposed method is much superior to the other compared algorithms in both accuracy rates and reliability.

Bearings are vital automation machine elements that are used quite frequently for power transmission and shaft bearing in rotating machines. The healthy operation of the bearings directly affects the performance of the rotating machines. Bearing faults may cause more vibration than normal in rotating machines, which wastes power. However, further bearing failures can cause vital damage to rotating machines. In this study, bearing vibration values are obtained through a special test setup. Different types and different sizes of artificial faults have been created in the bearings for the testing process. Data on these bearings are collected at different speeds. The purpose of the study is to diagnose faults in the bearings. In this context, a new approach is proposed. First, the one-dimensional local binary pattern (1D-LBP) method is applied to vibration signals, and all signal data are carried to the 1D-LBP plane. Statistical features are obtained from the signals in the 1D-LBP plane by using these features, and then the vibrational signals are classified by the gray relational analysis (GRA) model. Four different data sets are organized to test the proposed approach. The results of the test process with this proposed model have an accuracy of 99.044% for Dataset1 (different speed -300 rpm intervals), 94.224% for Dataset2 (different speed -60 rpm intervals), and 99.584% for Dataset3 (fault size (mm)); a 100% average success rate is observed for Dataset4 (fault type - error free bearing (EFB), inner ring fault (IRF), outer ring fault (ORF), and ball fault (BF)).

This article proposes the idea of fault classification-based control for the steady-state operation of unmanned surface vehicles (USVs). The idea of fault classification is achieved with the help of wavelet transforms and support vector machines, and the control is performed using a wavelet fuzzy controller. Initially, a brief idea of faults that affect the stable operation of USVs is identified. Furthermore, the surge and sway translational motion of USVs are realized with the help of a ball balancer setup. The fault data are measured in terms of plate angle, ball position, and motor operating voltage for developing the fault classifier. The proposed algorithm depicted improved classification accuracy when compared with conventional methods. To accommodate the operation of the system as per the operating state, a wavelet-based fuzzy controller is proposed. The proposed controller solves the problem of position tracking and balancing for ball and plate system with high precision, hence achieving the stable operation. [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]

• A principal component analysis (PCA)-based supervised machine learning (SML) method is developed. • PCA-based SML is proposed to enhance fault detection and diagnosis (FDD) of photovoltaic (PV) systems. • The developed FDD approach uses feature extraction and selection, and fault classification tools. • The detection performance is studied using several PV system faults. • The results show good diagnosis efficiency and higher classification accuracy in PV systems. Fault detection and diagnosis (FDD) in the photovoltaic (PV) array has become a challenge due to the magnitudes of the faults, the presence of maximum power point trackers, non-linear PV characteristics, and the dependence on isolation efficiency. Thus, the aim of this paper is to develop an improved FDD technique of PV systems faults. The common FDD technique generally has two main steps: feature extraction and selection, and fault classification. Multivariate feature extraction and selection is very important for multivariate statistical systems monitoring. It can reduce the dimension of modeling data and improve the monitoring accuracy. Therefore, in the proposed FDD approach, the principal component analysis (PCA) technique is used for extracting and selecting the most relevant multivariate features and the supervised machine learning (SML) classifiers are applied for faults diagnosis. The FDD performance is established via different metrics using data extracted from different operating conditions of the grid-connected photovoltaic (GCPV) system. The obtained results confirm the feasibility and effectiveness of the proposed approaches for fault detection and diagnosis. [ABSTRACT FROM AUTHOR]

Rotor bars are one of the most failure-critical components in induction machines. We present an approach for developing a rotor bar fault identification classifier for induction machines. The developed machine learning-based models are based on simulated electrical current and vibration velocity data and measured vibration acceleration data. We introduce an approach that combines sequential model-based optimization and the nested cross-validation procedure to provide a reliable estimation of the classifiers’ generalization performance. These methods have not been combined earlier in this context. Automation of selected parts of the modeling procedure is studied with the measured data. We compare the performance of logistic regression and CatBoost models using the fast Fourier-transformed signals or their extracted statistical features as the input data. We develop a technique to use domain knowledge to extract features from specific frequency ranges of the fast Fourier-transformed signals. While both approaches resulted in similar accuracy with simulated current and measured vibration acceleration data, the feature-based models were faster to develop and run. With measured vibration acceleration data, better accuracy was obtained with the raw fast Fourier-transformed signals. The results demonstrate that an accurate and fast broken rotor bar detection model can be developed with the presented approach.

Recently, precise and deterministic feature extraction is one of the current research topics for bearing fault diagnosis. For this aim, an experimental bearing test setup was created in this study. In this setup, vibration signals were obtained from the bearings on which artificial faults were generated in specific sizes. A new feature extraction method based on co-occurrence matrices for bearing vibration signals was proposed instead of the conventional feature extraction methods, as in the literature. The One (1) Dimensional–Local Binary Patterns (1D-LBP) method was first applied to bearing vibration signals, and a new signal whose values ranged between 0–255 was obtained. Then, co-occurrence matrices were obtained from these signals. The correlation, energy, homogeneity, and contrast features were extracted from these matrices. Different machine learning methods were employed with these features to carry out the classification process. Three different data sets were used to test the proposed approach. As a result of analysing the signals with the proposed model, the success rate is 87.50% for dataset1 (different speed), 96.5% for dataset2 (fault size (mm)) and 99.30% for dataset3 (fault type – inner ring, outer ring, ball) was found, respectively. [ABSTRACT FROM AUTHOR]

The fault classification of medium-voltage transmission lines consisting of wind farms connected to the high-voltage (HV) transmission networks is carried out using the power-spectrum-based hyperbolic S-transform (HST) powerful technique for analysing non-linear and non-stationary fault signals through the non-intrusive monitoring systems. The HST technique extracts the useful features in the time-frequency domain from measuring fault current waveforms of the HV utility side to discriminate the fault types. Parseval's theorem is applied to each HST coefficient to quantify the energy distribution of various fault types for reducing the size of inputs for recognition algorithms. Next, multiclass support vector machines achieve identification. The results have proved that the proposed classification technique is independent of fault resistance, source impedance, and fault inception angles. Extensive simulations are conducted using the electromagnetic transients program to show that the recognition accuracy of the fault classification for all types is up to 96.84%. [ABSTRACT FROM AUTHOR]

Accurate fault classification is the premise of fault location and management study in a power distribution network. In most of the traditional fault classification methods used in power distribution network, the characteristic quantities are selected by experience, which will increase the uncertainty of fault classification results. A novel fault classification method based on deep belief networks (DBN) is proposed in this paper. Samples of fault current and voltage are preprocessed by min–max standardization and waveform splicing firstly, then they are used to train the DBN together with fault type label. Characteristic quantities of the current and voltage will be automatically extracted by the well‐trained DBN model, and the reliable fault type classification of distribution network can be realized. Simulation and experimental results show that the fault classification method is suitable for distribution network, and it has not only characteristics of obvious fault feature extraction and high fault classification accuracy, but also has good adaptability while the neutral grounding modes changing or used in power distribution network with distributed generator. © 2020 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC. [ABSTRACT FROM AUTHOR]

To improve power system reliability, a protection mechanism is highly needed. Early detection can be used to prevent failures in the power transmission line (TL). A classification system method is widely used to protect against false detection as well as assist the decision analysis. Each TL signal has a continuous pattern in which it can be detected and classified by the conventional methods, i.e., wavelet feature extraction and artificial neural network (ANN). However, the accuracy resulting from these mentioned models is relatively low. To overcome this issue, we propose a machine learning-based on Convolutional Neural Network (CNN) for the transmission line faults (TLFs) application. CNN is more suitable for pattern recognition compared to conventional ANN and ANN with Discrete Wavelet Transform (DWT) feature extraction. In this work, we first simulate our proposed model by using Simulink® and Matlab®. This simulation generates a fault signal dataset, which is divided into 45.738 data training and 4.752 data tests. Later, we design the number of machine learning classifiers. Each model classifier is trained by exposing it to the same dataset. The CNN design, with raw input, is determined as an optimal output model from the training process with 100% accuracy. [ABSTRACT FROM AUTHOR]

Detection and classification of any anomaly at its commencement are very crucial for optimal management of assets in power system grids. This paper presents a novel hybrid approach that combines S-transform (ST) and feedforward neural network (FFNN) for the detection and classification of distribution grid faults. In this proposed strategy, the measured three-phase current signals are processed through ST with a view to extracting useful statistical features. The extracted features are then fetched to FFNN in order to detect and classify different types of faults. The proposed approach is implemented in two different test distribution grids modeled and simulated in real-time digital simulator and MATLAB/SIMULINK. The obtained results justify the efficacy of the presented technique for both noise-free and noisy data. In addition, the developed technique is independent of fault resistance, inception angle, distance, and prefault loading condition. Besides, the comparative results confirm the superiority and competitiveness of the developed technique over the available techniques reported in the literature.

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 diagnosis of bearings using classification techniques plays an important role in industrial applications, and, hence, has received increasing attention. Recently, significant efforts have been made to develop various methods for bearing fault classification and the application of Case Western Reserve University (CWRU) data for validation has become a standard reference to test the fault classification algorithms. However, a systematic research for evaluating bearing fault classification performance using the CWRU data is still lacking. This paper aims to provide a comprehensive benchmark analysis of the CWRU data using various entropy and classification methods. The main contribution of this paper is applying entropy-based fault classification methods to establish a benchmark analysis of entire CWRU datasets, aiming to provide a proper assessment of any new classification methods. Recommendations are provided for the selection of the CWRU data to aid in testing new fault classification algorithms, which will enable the researches to develop and evaluate various diagnostic algorithms. In the end, the comparison results and discussion are reported as a useful baseline for future research. [ABSTRACT FROM AUTHOR]

In order to reduce the instability of manual extraction and screening features in pipeline fault diagnosis, a fault diagnosis method based on deep confidence network to reconstruct feature parameters and model is proposed. Under laboratory conditions, the acoustic emission signals of the pipeline under normal and different fault conditions are collected, the characteristic parameters are extracted, the characteristic parameters are reconstructed by the deep confidence network and the classification model is established, and the number of model nodes is adjusted according to the characteristics of the collected sample data, and the parameter optimization model is obtained after the final diagnosis. The studies have shown that: under the same conditions, using the classification model established after the reconstruction of the depth of belief networks characteristic parameters have better stability and higher accuracy. [ABSTRACT FROM AUTHOR]

Bearing is one of the most critical parts used in rotary machines. Bearing faults break down the mechanism where it is located. Moreover, the faults may cause to malfunction by spreading to the entire system. Thus this may result in catastrophic failure eventually. Precise and decisive feature extraction from the raw vibration signal maintains to be one of the current topics explored for fault diagnosis in bearings. In this study, vibration signals are obtained from bearings which are formed with artificial faults of specific dimensions from a bearing test setup. Instead of employing traditional feature extraction methods found in the literature, a novel feature extraction method for bearing faults called one-dimensional ternary pattern (1D-TP) is applied. The proposed approach is a statistical method that uses patterns obtained from comparisons between neighbors of each value on vibration signals. The study aims to identify the size (mm) of the fault by determining the bearing part (inner ring, outer ring, ball) from which the faults in the bearings are caused. Several classification techniques were performed by using ternary patterns with RF (Random Forest), k-NN (k-nearest neighbor), SVM (Support Vector Machine), BayesNet, ANN (Artificial Neural Networks) models. As a result of analyzing the signals obtained from the experimental setup with the proposed model, 91.25% for dataset_1 (different speed), 100% for dataset_2 (fault type — inner ring, outer ring, ball) and 100% for dataset_3 (fault size (mm)) success rates are determined. • A novel approach is proposed to extraction of features from bearing signals. • One advantage is that this method uses all data points for feature extraction. • It is fast and can be use in real-time application. • High accuracies achieved for bearing fault classification. • Original data (Experimental setup of authors). [ABSTRACT FROM AUTHOR]

An imbalanced number of faulty and normal samples make the traditional supervised classification methods difficult to ensure their classification performance. Accordingly, semisupervised learning methods have recently become hotspots both in academic research and practical application domains. Different from previous schemes, this paper dedicates on the correlations, common features, and specific features among quality-related coupling faults in manufacturing industries. The main innovations are as follows: first, it is the first time to develop a robust semisupervised classification framework for quality-related coupling faults, which integrates semisupervised multitask feature selection and manifold learning; second, manifold structures and local discriminant information of unlabeled and limited labeled faulty samples are sufficiently explored to improve the classification performance; and third, correlations among quality-related coupling faults are accurately captured, which are crucial for understanding the uniqueness and relationships of them at the feature level. The proposed method is finally validated in a representative manufacturing industry, i.e., hot strip mill process, where detailed simulation processes are presented and better classification performance is shown compared with the existing approaches. [ABSTRACT FROM AUTHOR]

The inherent uncertainty of software gives a vague and imprecise solution when it is solved by human judgment. As the project expands, the issues of missing data values, outlier detection, feature subset selection and prediction of faultiness behaviour should be addressed. The feature selection process may lead to the production of high-dimensional data sets that may contribute to many irrelevant or redundant features. In this paper, we focussed on the optimal feature subset selection and fault prediction at the early stage of a project. We propose the novel approach of grey relational analysis (GRA) from grey system theory by optimizing the grey relational grade function using biogeography optimization referred to as B-GRA. The proposed algorithm gives resilience to users to select features for both continuous and categorical attributes. The issues such as feature subset selection, heterogeneity of data sets, outlier analysis and fault prediction are addressed, and then, B-GRA and GRA approaches on five publically available data sets are evaluated using statistical and machine learning techniques. Experimental results show significant results indicating that the proposed methodology can be used for the prediction of faults and produce conceivable results when compared with the GRA feature selection approach. [ABSTRACT FROM AUTHOR]

This paper focuses on data information preprocessing methods in the fault classification of rotating machinery. In order to avoid the information loss caused by the weighted fusion method, a merging fusion method is provided to obtain the final feature information. Furthermore, a direct fusion method that synchronizes the extraction and fusion of multi-sensor feature information is also proposed. The artificial neural network is used to test the three proposed information preprocessing methods and obtain rotary machinery fault classification methods. A final comparative experiment is given to compare the three methods proposed above in the fault classification of rotating machinery. [ABSTRACT FROM AUTHOR]

In this paper, machine learning technique is used to detect and classify all shunt faults in a UPFC compensated transmission line. A four-bus three-machine system with detailed modelling of UPFC has been used for fault simulation studies in MATLAB/Simulink. Instantaneous voltage and current signals obtained at local bus terminal are processed with DFT and statistical method for feature extraction. The input features of the ANN are minimised by using the statistical method. Generated features are used for training the ANN module. Trained ANN modules are used for testing different fault conditions in the time domain. Rigorous simulation studies have been performed with a wide variety of different possible fault situations. Simulation results bring out the superiority of the scheme. Moreover, the error introduced due to CT, CCVT and Dynamic behaviour of the UPFC has been considered for testing the trained ANNs by varying the different operating mode of UPFC, and different compensation levels, wherein all the cases, the performance is found reliable. [ABSTRACT FROM AUTHOR]

This paper proposes a new method for diagnosis of fault type and faulty phase of a series compensated transmission line. The standard deviation (SD) principle together with the fast discrete orthonormal s-transform (FDOST) and the decision tree (DT) is applied for the purpose of fault classification. The FDOST, as an efficient signal processing tool, is used for extracting the features from a half cycle window of voltage and current signals sampled from one end of the power system network. Finally, the SD of a half cycle post-fault samples of the FDOST coefficients is calculated to form the input feature vector for the DT-based classifier. The features are processed by the DT to classify faults. The practicability of the proposed method is validated by modified Western System Coordinating Council 3-machine 9-bus system simulated in the PSCAD/EMTDC software and field fault data captured from a real transmission network of Chhattisgarh state, India. The results confirm that the proposed method reliably classifies all types of faults with high efficacy. [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]