Transformer fault diagnosis method based on multilevel acoustic information

Dissolved gas analysis is one of the most effective methods for diagnosing transformer faults. The traditional method for oil-immersed transformer fault diagnosis can only recognize several types of defects and has a low accuracy rate. In order to improve the classification effectiveness, a refined classification method for transformer fault diagnosis is proposed. It can detect more types of faults with a higher accuracy rate. The proposed method is based on the probability-output relevance vector machine, and a three-layer four-classifier model is constructed to analyse the different diagnostic results of different kinds of input data. In this model, a binomial tree is used to transfer the multi-classification problem to four binary classification problems; each classifier is a binary classifier used to distinguish the transformer type between two types of error. The proposed method is employed for analysis of 100 DGA samples consisting of characteristic gas content. The experimental result shows that this method has a high diagnostic rate and can diagnose 11 kinds of operation state.

Current methods for processing acoustic signals for fault diagnosis are commonly based on the method of processing the vibration signals. It is difficult to obtain the information only existing in the acoustic signal. Therefore, considering the fault diagnosis based on the acoustic signal, a method based on the Logarithmic Mel (LM) spectrogram and a residual neural network (ResNet) is proposed. Firstly, to correspond with the auditory perception, the acoustic signal is processed and linearly represented in the LM spectrogram. Secondly, the ResNet model is introduced to extract the fault information from the LM spectrogram for the identification of the different types of faults. By training two datasets obtained from the gearbox platform, the different types of faults and the faults at different components of the gearbox are separately classified. The comparison results of different inputs with the LM spectrogram for training the ResNet model show that the LM spectrogram as the input could achieve the best classification accuracy. The comparison results of different classic deep learning models with the ResNet model demonstrate that the proposed methods can effectively extract features from acoustic signals in the LM spectrogram for the fault classification of the gearbox.

The expert system for fault diagnosis based on classified hierarchical model is addressed so as to improve the maintenance capability of engineering vehicles under complex battle environment. The analysis of fault feature, fault pattern and failure mechanism of the whole machine and its electrical system is accomplished. Furthermore, the structure of classified hierarchical diagnosis model and method of knowledge representation are implemented so that the knowledge base of the fault diagnosis system can be developed. The forward reasoning mechanism of the fault diagnosis expert system is bring forward based on neural network intelligent reasoning that can achieve fuzzy input and multiple fault diagnosis. This work effectively increases the fault maintenance, diagnosis and support capabilities of the military troops.

The respiratory sound classification paper focuses on analyzing sounds collected through the stethoscope at various chest positions to understand critical concerns of human breathing. Most doctors used a stethoscope for auscultation of lung sounds, along with heart sound, as a basic checkup and utilized later in predicting chest pathology. Using deep learning-based sound classification can build intelligent digital stethoscopes or understand overall health conditions later used along with other Lung organ analyses. The solution enables affordable and practical tools for remote parts of India, where doctors or specialists are not keen to work. Such solutions can be used as first analysis before further investigation by traveling to far-off specialist hospitals. The lung sound was prepared and hosted by a team from the University of Coimbra, Portugal; the University de Aveiro, Portugal; and the Aristotle University of Thessaloniki, Greece and the collaboration at Kaggle website lung sounds of patients who are Normal lung functioning and with some anomaly collected. The collected lung sounds are labeled as Normal, seven classes - Abnormal. Overall, eight levels are considered to classify using a deep learning model. The paper proposes a multipath convolution neural network (CNN) model for analyzing sound after exploring all approaches, like Recurrent Neural network (RNN), Machine learning models like SVM, and other CNN models as per paper [3]. Lung sound is processed using signal processing methods and converts wav files to Mel-frequency cepstral coefficients (MFCCs), Mel spectrogram, and chromatogram from a waveform or power spectrogram. Instead of using one feature among three, the paper built a Multipath CNN model to feed all three components to model and concatenate feature maps to classify the classes. The model was evaluated with various parameters like accuracy and F1 score. The proposed model gives 96.17% for the dataset used. The result implies data, and a multipath approach helps to fuse features extracted from different images and allows models to classify well.

This study presents a comprehensive deep learning framework for diagnosing acoustic faults in electric motors. The framework uses Mel spectrograms and a lightweight convolutional neural network (CNN). The method classifies three motor states, engine_good, engine_broken, and engine_heavyload, based on audio recordings from the IDMT-ISA-ELECTRIC-ENGINE dataset. To prevent data leakage and ensure a robust evaluation, the study employed file-level splitting, session separation, 5-fold cross-validation, and repeated trials. The raw audio signals were transformed into Mel spectrograms and processed through a CNN architecture that integrates convolutional, pooling, normalization, and dropout layers. Quantitative metrics, including THD, spectral entropy, and SNR, further characterize the acoustic distinctions between motor states. The proposed model achieved a test accuracy of 99.7%, outperforming or matching state-of-the-art approaches, such as ResNet-18, CRNN, and Transformer classifiers, as well as traditional MFCC-based baselines. Noise robustness and sensitivity analyses demonstrated stable performance under varying SNR conditions and preprocessing settings. Feature-importance maps revealed that low-frequency regions (0–40 Mel bins) were key discriminative components linked to physical fault mechanisms. Computational evaluation confirmed the model’s real-time feasibility on embedded hardware with low latency and a modest parameter count. Though primarily validated on one motor type, external-domain testing revealed strong adaptability. Future work may incorporate transfer learning or multimodal fusion. Overall, the proposed framework provides a highly accurate, interpretable, and efficient solution for real-time motor fault diagnosis and predictive maintenance in industrial environments.

Aiming at solving the multiple fault diagnosis problem as well as the sequence of all the potential multiple faults simultaneously, a new multiple fault diagnosis method based on the dependency model method as well as the knowledge in test results and the prior probability of each fault type is proposed. Firstly, the dependency model of the system can be built and used to formulate the fault-test dependency matrix. Then, the dependency matrix is simplified according to the knowledge in the test results of the system. After that, the logic ‘OR’ operation is performed on the feature vectors of the fault status in the simplified dependency matrix to formulate the multiple fault dependency matrix. Finally, fault diagnosis is based on the multiple fault dependency matrix and the ranking of each fault type calculated basing on the prior probability of each fault status. An illustrative numerical example and a case study are presented to verify the effectiveness and superiority of the proposed method.

Although dissolved gas analysis (DGA) is an effective method for transformer fault diagnosis, problems with the quality and accuracy of DGA characterization datasets often arise in practical industrial applications and face difficulties in adjusting the parameters of fault diagnosis models. To address the above problems, this paper proposes a fault diagnosis model (MD-IQPSO-RF) based on Mahalanobis distance (MD) data cleaning and improved quantum particle swarm (IQPSO) optimization of random forest (RF) parameters. Specifically, the abnormal samples of the DGA dataset are first processed by MD to improve the quality and accuracy of the dataset. Then, the RF parameters were optimized using the IQPSO algorithm to adjust the model parameters in order to improve the diagnostic performance of the RF. Finally, the optimal parameters of RF are output, and the training data are used to train the RF algorithm to construct the MD-IQPSO-RF transformer fault diagnosis model. The experimental results show that the model achieves an average accuracy of 93.631% for fault diagnosis, which is 6.92% higher than the unoptimized RF model. Comparison with other similar methods also achieved good results, which further validated the effectiveness of the fault diagnosis model.

With the development of the processing capacity of the embedded chip, it is possible to implement a machine learning algorithm in the embedded system. To achieve the fault status without poor portability, tricky threshold selection, and complex rulemaking, this article proposes a multimodal deep residual filter network (DRFN) for online multiple open-switch fault diagnosis of T-type three-level inverters. It contains low-rank matrix fusion (LMF), a DRFN, and a cross-transformer mechanism. The LMF fuses the voltage signal and the current signal for obtaining the unified representation. Then, the DRFN filters noise adaptively and extracts information effectively. Finally, the cross-transformer mechanism outputs the fault state of the T-type three-level inverter. The datasets consist of the dc-link voltage and load side current of the inverter control system. The data time window is selected as 20 ms. Through the real-time calculation of online monitored data, the experimental results show the effectiveness of the proposed fault diagnosis approach. Moreover, the accuracy of fault diagnosis is 99.18%, and the average open-circuit fault diagnosis time is 21 ms.

In this paper, a new approach to alarm processing in power systems is proposed based on the set covering theory and a refined genetic algorithm (RGA). Although alarm processing and fault diagnosis have different explanations in power system community, from the viewpoint of artificial intelligence community, the alarm processing problem is a typical Multiple Fault Diagnosis (MFD) problem. Thus, we first attribute the alarm processing problem to a MFD problem, and discuss the MFD problem in general with special emphasis on the evaluation criteria. And then, a probabilistic criterion is introduced to describe the alarm processing problem, and this criterion is more reasonable man the currently used criteria. Thirdly, a new method is presented to solve the alarm processing problem using a RAG. Finally, two sample examples are used to demonstrate the feasibility and efficiency of the developed method. Test results suggest that the developed RAG-based probabilistic method to the alarm processing problem is promising.

Fault diagnosis of rotating machines is an important task to prevent machinery downtime, and provide verifiable support for condition-based maintenance (CBM) decision-making. Deep learning-enabled fault diagnosis operations have become increasingly popular because features are extracted and selected automatically. However, it is challenging for these models to give superior results with rotating machine components of different scales, single and multiple faults across different rotating components, diverse operating speeds, and diverse load conditions. To address these challenges, this paper proposes a comprehensive learning approach with optimized signal processing transforms for single as well as multiple faults diagnosis across dissimilar rotating machine components: gearbox, bearing, and shaft. The optimized bicoherence, spectral kurtosis and cyclic spectral coherence feature spaces, and deep blending ensemble learning are explored for multiple faults diagnosis of these components. The performance analysis of the proposed approach has been demonstrated through a single joint training of the entire framework on a compound dataset containing multiple faults derived from three public repositories. A comparison with the state-of-the-art approaches that used these datasets, shows that our method gives improved results with different components and faults with nominal retraining.

Bayesian network offers a powerful map framework that can process probabilities inference. It can be used in inference and express of uncertainty knowledge. This paper introduce a new electrical equipment fault diagnosis method based on Bayesian network (BN). For example, power transformer is very important in power system as a electrical equipment. But, it's very difficult to diagnose the fault exactly because power transformer's complexity configuration. Now, dissolved gas analysis (DGA) is the most effective and convenient method in transformer fault diagnosis. However, the codes of DGA is too absolute, so this paper advances a new transformer fault diagnosis method based on Bayesian network (BN). This method introduces BN method into transformer fault diagnosis and presents a new idea of finding out transformer faults rapidly and exactly. Then, the transformer fault diagnosis model based on Bayesian network and DGA is constructed. Finally, the application examples in the fault diagnosis of transformer are given which shows that this method is effective.

As the mark of sonic source category, acoustic signal also carries some architecture and running status information of transformer to be a natural character. In fault diagnosis techniques, acoustic recognition as one of the important methods was used in mechanical equipments fault diagnosis and observation system. In transformer fault diagnosis, its acoustic signal time-frequency changing has close connection with the active state. A new method of transformer fault diagnosis based on Hilbert transform and wavelet contour is developed. The fault signals and normal signals that have near and far sound field noise are extracted, and then de-noise it. Using the Hilbert transform and wavelet contour processing the measured signal, Hilbert spectrum and wavelet contour, which attribute the time-frequency changing of transformer acoustic signal, is obtained. The test result shows the character of transformer running state. It is demonstrated by actual analysis that the new detection method is one effective way for the condition evaluation and fault diagnosis of transformer.

Gearbox is a key component of mechanical equipment, which has a complex structure, harsh working conditions, and a higher probability of failure. Therefore, gearbox fault diagnosis is vital to ensure the efficient operation of the whole mechanical equipment. However, traditional gearbox fault diagnosis methods mainly rely on manual feature extraction. To address this issue, a novel end-to-end fault diagnosis model that combines convolutional neural network (CNN), long short-term memory network (LSTM) and attention mechanism (AM) is proposed. Firstly, Spatial features are extracted from the original input data using CNNs, and then temporal features are extracted even further from the spatial features using LSTMs. The attention mechanism improves the network's attention to the global key features by assigning weights, and finally, the fault diagnosis results are derived from the fully connected layer and Softmax classifier. The experimental results show that the method is able to adaptively extract features and ensure higher diagnostic accuracy, validating the effectiveness of the proposed method.

The failure of water chiller will cause a series of problems such as increasing energy consumption, decreasing comfort of users and decreasing life of equipment. The fault diagnosis of the chiller has an important role in the stable operation of the air conditioner. In order to improve the fault diagnosis level of water chillers and solve the problems of difficulty in determining the parameters of neural network and great randomness, this paper presents a fault diagnosis method of long short-term memory (LSTM) network based on whale optimization algorithm (WOA) optimization. The fault diagnosis performance was verified by RP-1043 dataset, and compared with the results of recurrent neural network (RNN) and BP neural network. The results show that the WOA-LSTM method can effectively identify five typical chillers, and the diagnostic accuracy is as high as 99.27%. Compared with RNN and BP neural network, the proposed method is helpful to detect faults as early as possible and reduce the loss. It is especially obvious to improve the detection efficiency of small faults and complex faults.

In a chiller plant, primary or critical sensors are used to control the system operation while secondary sensors are installed to monitor the performance/health of individual equipment. Current sensor fault detection and diagnosis (SFDD) approaches are not applicable to secondary sensors which usually are not involved in the system control. Consequently, a hybrid multiple sensor fault detection, diagnosis and reconstruction (HMSFDDR) algorithm for chiller plants was developed. Machine learning and pattern recognition were used to predict the primary sensor faults through the comparison of the weekly performance curves. With the primary sensor signals reconstructed, the secondary sensor faults were estimated based on mass and energy balance. By applying the algorithm with various logged plant data and comparison with site checking results, a maximum of 75% effectiveness could be achieved. The merits of the present approach were further justified through off-site sensor testing which reinforced the usefulness of proposed HMSFDDR algorithm.