Sparse-CNN: An Effective Combination to Improve Performance of Classification of Fault of Rotating Machinery

Fault of rotating machinery is very common in mechanical systems. It is caused by long-time operation of mechanical systems. So it is important to use different methods to discover the reason and the form of fault. This paper tries to develop a more effective model to predict the type of vibration signal of rotating machinery. Based on sparse deconvolution model and CNN model, we create a new model called Sparse-CNN, which combines the advantages of sparse deconvolution and CNN. Then, we compare 4 different kinds of models of Sparse-CNN in simulation and experiment. The accuracy and loss function are compared in each experiment. Finally, we verify the effectiveness of our proposed model for the case of strong noise and the performance in coping with different datasets.