Prognostics of the ultrasonic motor based on an deep learning network model considering working signals and environmental factors

Abstract

With the advantage of light weight, small volume, less electromagnetic radiation, fast dynamic response, and high energy density, ultrasonic motors have important applications in aerospace, automation and medical equipment. To realize the prognostics of an ultrasonic motor, this paper proposed a deep learning network model combining the convolutional neural network and the recurrent neural network. The convolutional neural network mines the correlation between different working parameters, while the recurrent neural network mines the temporal correlations of a single channel. Moreover, a fault injection test is carried out to obtain the baseline data and further validate the proposed prognostic model. The results show that this model can give accurate prognostic results under different environmental conditions.