Hierarchical symbolic analysis and particle swarm optimization based fault diagnosis model for rotating machineries with deep neural networks

Abstract Fault diagnosis (FD) is considered as a hot research topic for prognostics and health management of machinery by the detection and identification of faults. The diagnosis of faults in rotating machinery is important for improving safety, enhancing reliability and reducing maintenance cost. Therefore, different research works have started to concentrate on the design of FD models with high automation and accurateness. This paper presents an FD model by integrating hierarchical symbolic analysis (HSA) and particle swarm optimization with a convolutional neural network (PSO-CNN) named HPC model. The presented HPC model initially undergoes feature extraction process using HSA. Then, the PSO-CNN model is utilized for learning the complicated non-linear relationship between the features and health conditions in an automated way. The PSO-CNN exhibits a fast convergence rate and involves a direct encoding mechanism as well as velocity operator to allow the optimal usage of PSO with CNN. For validation, a centrifugal pump dataset is employed. The simulation outcome showcased the superiority of the presence HPC model over the compared methods under different measures. From the detailed experimental outcome, it is shown that the presented FD model offers excellent results by attaining a maximum classification accuracy of 98.97 and 99.09 under two dataset namely dataset 1 and dataset 2 respectively.