A novel framework for multi-sensor data fusion in bearing fault diagnosis using continuous wavelet transform and transfer learning

Intelligent fault diagnosis (IFD) is crucial in industrial settings, leveraging big data from various sensors and machine learning advancements to monitor critical components such as rolling bearings. While IFD-based deep learning and multi-sensor fusion offer promising solutions, challenges remain in integrating heterogeneous data and managing computational complexity. Transfer learning from pre-trained models can mitigate these issues, particularly with limited labeled datasets common in industrial applications. However, integrating transfer learning with multi-sensor fusion for diagnosing complex fault scenarios, especially combined bearing defects under varying operational conditions, remains underexplored in current research. This paper proposes a novel multi-sensor fusion approach for bearing fault diagnosis that combines vibration and acoustic signals within a transfer learning framework. Continuous Wavelet Transform (CWT) is applied to multi-sensor inputs, and the resulting wavelet coefficients are fused using the Maximum Energy to Shannon Entropy Ratio (ME-to-SER) criterion to fine-tune pre-trained Convolutional Neural Networks (CNNs). The effectiveness of the proposed method is validated on the Spectra Quest Machinery Fault Simulator (MFS) across various bearing fault scenarios, including combined faults, under variable speeds. The proposed approach achieves high accuracy (up to 100%) using multi-modal fused data, outperforming single-modality methods. It excels in complex fault classification and maintains robustness under various operational conditions. The fusion approach efficiently handles heterogeneous data to enhance diagnostic reliability, whereas transfer learning effectively addresses limited labeled datasets and reduces the computational burden of training deep CNNs from scratch.