Virtual training of machine learning algorithm using a multibody model for bearing diagnostics on independent cart system

Independent cart conveyor system is a new technology recently proposed in the field of automatic machines. This technology uses advance linear motors for moving several carts on a close-loop path. In this paper, a multibody model of the Independent cart conveyor system is used to train a machine learning algorithm for the diagnostics of ball bearings which support the carts. The multibody model provides several simulations both of healthy and faulted bearings, which are used to create the training dataset. The input features of the machine learning algorithm are statistical parameters that proved to be effective in the analysis of real vibration data. The final tests were carried out on experimental data recorded on a test rig and the fault detection algorithm was validated both in faulted and healthy cases in an industrial environment. The aim of this activity is to virtualize the training of a machine learning system for fault detection and to test its accuracy in a real environment.