Enhanced Diagnosis of Wind Turbine Main Bearing Faults Through Fusion of Multi-source Signals with a Hybrid MTF-CNN-NSGAII Approach

This paper proposes an innovative and efficient lightweight multi-source signal fusion method specifically for the precise fault diagnosis of wind turbine main bearings. Ingeniously, this method integrates the Markov Transfer Field (MTF) and Convolutional Neural Network (CNN) to adeptly achieve the transformation from complex multi-source signals to images, significantly enhancing the efficiency of fault feature extraction and recognition. This strategic approach not only considerably enhances the model’s capability to accurately learn fault features in time series signals but also dramatically reduces the model’s parameter count and computational complexity through its thoughtfully designed lightweight structure, making it exceptionally suitable for rapid and reliable fault monitoring and diagnosis. Through meticulous experimental verification on the detailed bearing dataset of Paderborn University, its effectiveness and superior performance in the fault diagnosis of wind turbine main bearings have been thoroughly demonstrated. Importantly, this article not only introduces a groundbreaking method for the fault diagnosis of wind turbines but also pioneers a new pathway for the application of deep learning in the complex field of multi-source signal processing and fault pattern recognition.