A FAULT DETECTION MODEL FOR PREDICTIVE MAINTENANCE OF A WIND TURBINE BASED ON A HYBRID DEEP LEARNING APPROACH

Paper Details

Authors: Farouk N.; Jaafar A.; Ayoub A.

DOI: 10.24874/PES08.01.019

Journal: Proceedings on Engineering Sciences

Year: 2026

Publisher: Faculty of Engineering, University of Kragujevac

Document Type: Article

Open Access: All Open Access; Gold Open Access; Green Open Access

Cited by: 0

Abstract

In the context of Industry 4.0, predictive maintenance of wind turbines is a crucial issue for optimizing their performance. This article proposes a new hybrid deep learning approach, aligned with the principles of Industry 4.0, to improve the predictive detection of failures in wind turbines. By combining long short-term recurrent neural networks (LSTM), artificial neural networks (ANN), and genetic algorithms (GA), our hybrid model (GA-LSTM-ANN) effectively captures the temporal and nonlinear complexities of sensor data. Genetic algorithms automatically optimize the model configuration. This innovative approach offers a unique solution for predictive fault detection. Applied to a real dataset, our model has significantly outperformed traditional methods, achieving 96.32% precision, 95.91% accuracy, 96.45% F1-score, and 96.41% recall for wind turbine fault detection. These promising results open up new perspectives for optimizing the management of wind farms and contributing to the energy transition. © 2026 Published by Faculty of Engineering.

Keywords

Deep Learning Networks Fault Detection; Genetic Algorithms (GA); Industry 4.0; Predictive Maintenance; Wind Turbine

Read Paper Back to Search