Augmented-Reality-Assisted Bearing Fault Diagnosis in Intelligent Manufacturing Workshop Using Deep Transfer Learning

Digital twin is a significant way to realize intelligent manufacturing, and provides important supports for the development of fault diagnosis. In the actual workshop, it is quite difficult to achieve to sufficient data to train a diagnosis model with high reliability, and the test data do not follow the same distribution under different working condition which traditional data-based fault diagnosis rely on. In this paper, we present a digital-twin-assisted bearing fault diagnosis method to solve the difficulty in achieving fault data and training fault diagnosis model under dynamic working condition. Firstly, by relying on the digital twin virtual workshop, massive fault data is provided to train the fault diagnosis model, and transfer this model to the actual workshop to avoid the waste in time and knowledge. An experiment is conducted in an actual machining workshop, which verifies the proposed method.