Application of optimised neural networks models in gears and bearings faults diagnosis

Gears and bearings are some of the most important machine components in the industrial world and detection of their faults has become a major trend. In the present article, in order to bring a reliable methodology for monitoring and diagnosis of rotating machinery failures, a test rig is implemented. However, to diagnose gears and bearings combined faults, a monitoring system based on neural network model (NNM), is proposed. To train and test the NNM, the principal high frequency indicators, determined with the collected time domain vibration data, and codes of defects are used respectively as input and output data. A comparison of two learning algorithms, optimised by the Taguchi method, was done to determine the best NNM. Therefore, the proposed method is effective to study other various industrial cases.