A Multi-concurrent Fault Diagnosis Scheme for the Parallel Shaft Gearbox Based on Resnet Neural Network and Image Recognition Approach

The main objective of this paper is to propose a fault diagnosis scheme for the multi-concurrent faults of the parallel shaft gearbox. To be specific, the normal and failure data sets are preprocessed and translated into black and white pixel images first. Then, the data sets are divided into training sets and verification sets with different batch sizes. By applying the Resnet Neural Network, a fault diagnosis scheme is presented for monitoring the gearbox. Finally, a simulating experiment with multiple failure modes is conducted to illustrate the effectiveness of the developed results.