Novel deep learning framework for aeroengine-bearing fault diagnosis: Optimizing VMD parameters with TTAO

Abstract

Early and accurate bearing fault diagnosis is crucial for aeroengine safety and operational efficiency. This paper proposes a novel deep learning framework for aeroengine-bearing fault diagnosis, leveraging the strengths of variational mode decomposition (VMD), convolutional neural networks (CNN), and residual networks (ResNet). The framework’s key innovation lies in optimizing critical VMD parameters using a triangle extension and aggregation optimization (TTAO) algorithm and integrating a ResNet architecture into the CNN framework. This approach enhances the model’s accuracy and robustness in identifying bearing faults, even under significant noise and vibration. The optimized VMD decomposes vibration signals into intrinsic mode functions (IMFs), fed into the CNN-ResNet architecture for feature extraction and classification. Extensive experimental evaluations using laboratory-acquired aeroengine-bearing datasets demonstrate that the proposed VMD-CNN-ResNet approach outperforms traditional methods and other deep learning architectures. The model accurately identifies various bearing faults, including inner ring, outer ring, and rolling element faults. Cross-dataset migration tests demonstrate the model’s ability to adapt to different data sources, suggesting potential for practical aerospace applications. Further validation through extensive testing and real-world data is needed to assess its suitability for engineering scenarios. This research presents a reliable framework for aeroengine-bearing fault diagnosis, enhancing predictive maintenance strategies.