Assessment of Tapered Roller Bearing Assembly Status with Friction Torque Based on Stacked Deep Autoencoder and BP Neural Network

The magnitude of rolling bearing frictional torque is an important performance parameter that affects bearing power consumption and life. In this paper, the tapered roller bearing in transmission device is investigated as the research object, and the assembly test is designed to study the influence law of bearing assembly tolerance, temperature, speed, radial load and preload force on the friction torque of tapered roller bearing. Firstly, the monitoring signals are processed by a stacked deep autoencoder to automatically extract the critical features. Secondly, a back propagation (BP) neural network model is employed to learn the mapping connection between the parameter information and the friction torque. And then the prediction of the tapered roller bearings friction torque is realized. It is verified that the relative errors of the predicted friction torque under three different working conditions are ≤ 0.1, which satisfies the requirements of engineering applications.