NOSCNN: A robust method for fault diagnosis of RV reducer

Operating conditions of RV reducer, such as speeds and loads, are frequent to change. In order to identify the fault of RV reducer under different operating conditions, a noise deep convolution neural model (NOSCNN) is proposed in this paper. The NOSCNN model follows the idea of modular design to simplify the structure. The whole NOSCNN model consists of five blocks with the same structures and a full connection layer. Moreover, a random noise layer is developed and added to the blocks of NOSCNN model to improve its capacity of resisting disturbance. Effectiveness and feasibility of the NOSCNN model are validated by datasets under various conditions. By comparing to experimental results, the present NOSCNN model is confirmed to be more robust than other algorithms.