Selected work

Representative Papers

Power transformer health index and life span assessment: A comprehensive review of conventional and machine learning based approaches

Jan 01, 2025Engineering applications of artificial intelligence

Accurate health assessment and remaining useful life (RUL) prediction of power transformers remain challenging due to heterogeneous diagnostic data and model interpretability limitations. Method: This paper systematically reviews and unifies health index construction principles and RUL prediction paradigms, integrating traditional diagnostic data—including dissolved gas analysis (DGA), frequency response analysis (FRA), and dielectric loss—and machine learning models such as SVM, random forests, LSTM, and graph neural networks into a multi-source information fusion classification framework. Contribution/Results: It proposes novel applicability criteria distinguishing physics-based and data-driven methods, revealing synergistic modeling opportunities. Emphasizing interpretable modeling as critical for assessment robustness, the study identifies hybrid modeling—combining physical constraints with deep representation learning—as the key pathway toward high-accuracy, high-fidelity transformer health assessment.

5 citationsRead paper

Predicting Male Fertility Using Machine Learning: A Semen Parameters Based Analysis with the VISEM Dataset

Jul 09, 2026

Male infertility is often underdiagnosed due to the lack of objective assessment tools. This study systematically evaluates the performance of over forty machine learning models in classifying fertility status into three categories—fertile, subfertile, and infertile—using semen parameters (concentration, motility, and morphology) from the VISEM dataset comprising 85 subjects. Leveraging feature engineering, the LazyPredict automated modeling framework, five-fold cross-validation, and multiclass ROC-AUC analysis, the Nearest Centroid classifier emerged as the top-performing model, achieving an accuracy of 94.2%. Its performance significantly surpassed that of support vector machines and quadratic discriminant analysis, demonstrating strong potential as a clinical decision-support tool for male infertility diagnosis.

0 citationsRead paper

Towards Automated Solar Panel Integrity: Hybrid Deep Feature Extraction for Advanced Surface Defect Identification

Apr 13, 2026

This study addresses the high cost, low efficiency, and limited scalability of manual inspection for photovoltaic (PV) panel defects, particularly in large-scale or remote solar farms. To overcome these challenges, the authors propose a hybrid defect detection framework that integrates handcrafted features—specifically Local Binary Patterns (LBP), Histogram of Oriented Gradients (HoG), and Gabor features—with deep features extracted from DenseNet-169. The framework further enhances robustness and generalization by fusing predictions from multiple classifiers, including Support Vector Machine (SVM), XGBoost, and LightGBM (LGBM). Evaluated on an augmented dataset, the proposed method achieves a state-of-the-art accuracy of 99.17% using the DenseNet-169 + Gabor + SVM configuration, significantly outperforming existing approaches and demonstrating its practical effectiveness and innovation in real-world PV monitoring applications.

0 citationsRead paper

Recent publications

Latest Papers

Predicting Male Fertility Using Machine Learning: A Semen Parameters Based Analysis with the VISEM Dataset

Jul 09, 2026

Male infertility is often underdiagnosed due to the lack of objective assessment tools. This study systematically evaluates the performance of over forty machine learning models in classifying fertility status into three categories—fertile, subfertile, and infertile—using semen parameters (concentration, motility, and morphology) from the VISEM dataset comprising 85 subjects. Leveraging feature engineering, the LazyPredict automated modeling framework, five-fold cross-validation, and multiclass ROC-AUC analysis, the Nearest Centroid classifier emerged as the top-performing model, achieving an accuracy of 94.2%. Its performance significantly surpassed that of support vector machines and quadratic discriminant analysis, demonstrating strong potential as a clinical decision-support tool for male infertility diagnosis.

0 citationsRead paper

Towards Automated Solar Panel Integrity: Hybrid Deep Feature Extraction for Advanced Surface Defect Identification

Apr 13, 2026

This study addresses the high cost, low efficiency, and limited scalability of manual inspection for photovoltaic (PV) panel defects, particularly in large-scale or remote solar farms. To overcome these challenges, the authors propose a hybrid defect detection framework that integrates handcrafted features—specifically Local Binary Patterns (LBP), Histogram of Oriented Gradients (HoG), and Gabor features—with deep features extracted from DenseNet-169. The framework further enhances robustness and generalization by fusing predictions from multiple classifiers, including Support Vector Machine (SVM), XGBoost, and LightGBM (LGBM). Evaluated on an augmented dataset, the proposed method achieves a state-of-the-art accuracy of 99.17% using the DenseNet-169 + Gabor + SVM configuration, significantly outperforming existing approaches and demonstrating its practical effectiveness and innovation in real-world PV monitoring applications.

0 citationsRead paper

Power transformer health index and life span assessment: A comprehensive review of conventional and machine learning based approaches

Jan 01, 2025Engineering applications of artificial intelligence

Accurate health assessment and remaining useful life (RUL) prediction of power transformers remain challenging due to heterogeneous diagnostic data and model interpretability limitations. Method: This paper systematically reviews and unifies health index construction principles and RUL prediction paradigms, integrating traditional diagnostic data—including dissolved gas analysis (DGA), frequency response analysis (FRA), and dielectric loss—and machine learning models such as SVM, random forests, LSTM, and graph neural networks into a multi-source information fusion classification framework. Contribution/Results: It proposes novel applicability criteria distinguishing physics-based and data-driven methods, revealing synergistic modeling opportunities. Emphasizing interpretable modeling as critical for assessment robustness, the study identifies hybrid modeling—combining physical constraints with deep representation learning—as the key pathway toward high-accuracy, high-fidelity transformer health assessment.

5 citationsRead paper