A dual attention mechanism network with self-attention and frequency channel attention for intelligent diagnosis of multiple rolling bearing fault types

Different fault types of rolling bearings correspond to different features, the classical deep learning models using a single attention mechanism (AM) has limitations in feature diversity capturing. Therefore, a novel dual attention mechanism network (DAMN) with self-attention (SA) and frequency channel attention (FCA) is proposed for rolling bearings fault diagnosis, in which the SA mechanism is used to capture global relationships between the input features and the fault types, and the FCA mechanism applies mutli-spectral attention to learn the local useful information among different input channels. Results of the ablation study of the effects of FCA blocks show that including a proper combination of multiple frequency components is helpful to achieve higher accuracy. Experiments on the diagnosis of rolling bearings with multiple fault types were carried out. Results show that compared with the current fault diagnosis models, the proposed DAMN has better comprehensive performance on diagnosis accuracy and model convergence speed. It is also demonstrated that the backbone of DAMN based on dual AM can achieve better performance than the backbone based on single AM.