The complexity of mechanical equipment and the harsh working environment make bearing faults susceptible to strong background noise interference. As a result, it is challenging to learn the crucial information from vibration data, leading to a significant decline in the performance of most fault diagnostic models. To overcome the above deficiencies, a novel fault diagnostic model based on signal denoising and dual branch feature fusion under strong noise interference (SDFF) is proposed. Firstly, the raw vibration signals are decomposed using Variational Mode Decomposition (VMD), followed by computing the Permutation Entropy (PE) to identify noise components. Subsequently, effective signal components are selected for signal reconstruction to achieve effective denoising. Secondly, a dual-branch convolutional neural network (CNN) model is constructed, and a multi-modal fusion mechanism is applied to integrate the fault features extracted by the dual-branch CNN model. This fusion process enhances the comprehensiveness of the features by combining information from multiple modalities. Finally, a softmax classifier is employed to identify different types of bearing faults. Two bearing datasets are adopted to validate the validity of SDFF. Compared to various baseline methods, SDFF owns superior noise resistance, feature extraction capabilities, and better diagnostic performance.
Li et al. (Thu,) studied this question.