The invention discloses an intelligent
ball screw life prediction method based on a self-attention mechanism. The intelligent
ball screw life prediction method comprises the following steps: step 1, constructing vibration feature capture models of different time scales; 2, constructing a
health index prediction model; 3, constructing a model training sample
library; 4, performing model training; step 5, acquiring actual operation data of the
ball screw; step 6, predicting health indexes; and 7, life prediction: according to the
health index and the degradation point of the ball screw to be subjected to life prediction, calculating to obtain the final
life time of the ball screw to be subjected to life prediction. According to the invention, by integrating various sensors, detecting key parameters such as temperature, vibration and
noise of the ball screw in real time, and combining vibration feature capture models and
health index prediction models of different time scales, service life prediction of the ball screw is realized, so that the operation reliability and maintenance efficiency of the ball screw can be effectively improved, and the service life of the ball screw is prolonged. And the
downtime is reduced.