A novel swarm budorcas taxicolor optimization-based multi-support vector method for transformer fault diagnosis

Neural Networks, Journal Year: 2025, Volume and Issue: 184, P. 107120 - 107120

Published: Jan. 6, 2025

Language: Английский

---

E‐SDHGN: A Multifunction Radar Working Mode Recognition Framework in Complex Electromagnetic Environments

IET Radar Sonar & Navigation, Journal Year: 2025, Volume and Issue: 19(1)

Published: Jan. 1, 2025

ABSTRACT A multifunction radar (MFR) can operate in multiple modes and perform various tasks such as surveillance, detection, fire control, search tracking. Recognising an MFR's operating mode is critical electronic warfare intelligence reconnaissance, aiding practical threat assessment countermeasure tasks. However, current recognition methods face challenges overlapping parameters among working suboptimal accuracy under conditions with parameter errors, missing pulses false pulses. Spurred by these concerns, this paper proposes entropy‐enhanced spatial‐deformable hybrid multiscale group network (E‐SDHGN) to recognise the of MFR address challenges. E‐SDHGN employs multidimensional entropy computations construct robust features integrates deformable convolution positional encoding enhance model's ability capture complex features. Additionally, it enhances feature extraction fusion within dynamic shared residual (DSRN) module integrating KAN modules weight‐sharing strategies. adaptive margin based on attention mechanisms improves classification conditions. Experimental results demonstrate that achieves superior robustness, even challenging This underscores its value for applications electromagnetic environments.

Language: Английский

---

ACMSlE: A Novel Framework for Rolling Bearing Fault Diagnosis

Processes, Journal Year: 2025, Volume and Issue: 13(4), P. 1167 - 1167

Published: April 12, 2025

Precision rolling bearings serve as critical components in a range of diverse industrial applications, where their continuous health monitoring is essential for preventing costly downtime and catastrophic failures. Early-stage bearing defects present significant diagnostic challenges, they manifest weak, nonlinear, non-stationary transient features embedded within high-amplitude random noise. While entropy-based methods have evolved substantially since Shannon’s pioneering work—from approximate entropy to multiscale variants—existing approaches continue face limitations computational efficiency information preservation. This paper introduces the Adaptive Composite Multiscale Slope Entropy (ACMSlE) framework, which overcomes these constraints through two innovative mechanisms: time-window shifting strategy, generating overlapping coarse-grained sequences that preserve signal traditionally lost non-overlapping segmentation, an adaptive scale optimization algorithm dynamically selects discriminative scales variation coefficients. In comparative analysis against recent innovations, our integrated fault diagnosis framework—combining Fast Ensemble Empirical Mode Decomposition (FEEMD) preprocessing with Particle Swarm Optimization-Extreme Learning Machine (PSO-ELM) classification—achieves 98.7% accuracy across multiple defect types operating conditions. Comprehensive validation multidimensional stability analysis, complexity discrimination testing, data sensitivity confirms this framework’s robust separation capability.

Language: Английский

---

A Hybrid Machine Learning Framework for Early Fault Detection in Power Transformers Using PSO and DMO Algorithms

Energies, Journal Year: 2025, Volume and Issue: 18(8), P. 2024 - 2024

Published: April 15, 2025

The early detection of faults in power transformers is crucial for ensuring operational reliability and minimizing system disruptions. This study introduces a novel machine learning framework that integrates Particle Swarm Optimization (PSO) Dwarf Mongoose (DMO) algorithms feature selection hyperparameter tuning, combined with advanced classifiers such as Decision Trees (DT), Random Forests (RF), Support Vector Machines (SVM). A 5-fold cross-validation approach was employed to ensure robust performance evaluation. Feature extraction performed using both Discrete Wavelet Decomposition (DWD) Matching Pursuit (MP), providing comprehensive representation the dataset comprising 2400 samples 41 extracted features. Experimental validation demonstrated efficacy proposed framework. PSO-optimized RF model achieved highest accuracy 97.71%, precision 98.02% an F1 score 98.63%, followed by PSO-DT 95.00% accuracy. Similarly, DMO-optimized recorded 98.33%, 98.80% 99.04%, outperforming other DMO-based classifiers. demonstrates significant advancements transformer protection enabling accurate fault detection, thereby enhancing safety systems.

Language: Английский

---

Fault diagnosis of power transformers based on t-SNE and ECOC-TEWSO-SVM

AIP Advances, Journal Year: 2024, Volume and Issue: 14(5)

Published: May 1, 2024

Support Vector Machines (SVMs) have achieved significant success in the field of power transformer fault diagnosis. However, challenges such as determining SVM hyperparameters and their suitability for binary classification still exist. This paper proposes a novel method diagnosis, called ECOC-WSO-SVM, which utilizes White Shark Optimizer (WSO) error correcting output codes to optimize SVMs. First, t-distributed Stochastic Neighbor Embedding (t-SNE) is employed reduce dimensionality Dissolved Gas Analysis (DGA) features constructed using correlation ratio method, from 26 dimensions. In addition, effectively solve SVMs, multi-strategy fusion proposed improve WSO, incorporating tent chaos initialization, elite opposite learning, selection strategies, forming TEWSO, its superior optimization performance validated IEEE CEC2021 test functions. Furthermore, address limitations SVMs classifier, an code introduced, thus constructing multi-class model. Finally, diagnostic ECOC-TEWSO-SVM model real-world data. Results demonstrate that exhibits best compared traditional models those literature, thereby proving significance effectiveness

Language: Английский

---

Enhancing Transformer Protection: A Machine Learning Framework for Early Fault Detection

Sustainability, Journal Year: 2024, Volume and Issue: 16(23), P. 10759 - 10759

Published: Dec. 8, 2024

The reliable operation of power transformers is essential for grid stability, yet existing fault detection methods often suffer from inaccuracies and high false alarm rates. This study introduces a machine learning framework leveraging voltage signals early detection. Simulating diverse conditions—including single line-to-ground, line-to-line, turn-to-ground, turn-to-turn faults—on laboratory-scale three-phase transformer, we evaluated decision trees, support vector machines, logistic regression models on dataset 6000 samples. Decision trees emerged as the most effective, achieving 99.90% accuracy during 5-fold cross-validation 95% separate test set 400 unseen Notably, achieved low rate 0.47% 6000-sample healthy condition dataset. These results highlight proposed method’s potential to provide cost-effective, robust, scalable solution enhancing transformer advancing reliability. demonstrates efficacy voltage-based diagnostics, offering practical resource-efficient alternative traditional methods.

Language: Английский

---

Few-Shot power transformers fault diagnosis based on Gaussian prototype network

International Journal of Electrical Power & Energy Systems, Journal Year: 2024, Volume and Issue: 160, P. 110146 - 110146

Published: July 25, 2024

Language: Английский

---

HazardClassTransformer: Transformer-Based Model for Reactive Chemical Hazard Classification in Industrial Processes

Published: Sept. 13, 2024

Language: Английский

---

Research on transformer fault diagnosis models with feature extraction

Review of Scientific Instruments, Journal Year: 2024, Volume and Issue: 95(11)

Published: Nov. 1, 2024

To address the challenge of low accuracy in traditional transformer fault diagnosis algorithms, this paper introduces a novel approach that utilizes Artificial Hummingbird Algorithm (AHA) to optimize both Kernel Principal Component Analysis (KPCA) and Extreme Learning Machine (ELM). We propose use various gas concentration ratio features apply AHA algorithm fine-tune kernel function parameters KPCA, thus establishing an AHA-KPCA feature extraction model. This model takes expanded as input selects top N principal components with cumulative contribution rate above 95% form vectors for classification. Following this, is employed weights hidden layer biases ELM, leading development AHA-ELM classification Ultimately, identified by serve inputs simulation verification Experimental results indicate proposed AHA-KPCA-ELM method attains 95.73%, surpassing intelligent diagnostic methods existing advanced thereby confirming effectiveness our method.

Language: Английский

---