A Multi-Task-Based Deep Multi-Scale Information Fusion Method for Intelligent Diagnosis of Bearing Faults

The use of deep learning for fault diagnosis is already a common approach. However, integrating discriminative information types and scales into models rich multitask feature still deserves attention. In this study, multitask-based multiscale fusion network model (MEAT) proposed to address the limitations poor adaptability traditional convolutional neural complex jobs. performed multidimensional extraction through convolution at different obtain levels information, used hierarchical attention mechanism weight features achieve an accuracy 99.95% total task six classification, considered two subtasks in classification discriminate size type multi-task mapping decomposition. Of these, highest reached 100%. addition, Precision, ReCall, Sacore F1 all index 1, which achieved accurate bearing faults.