Lubrication State Diagnosis of Ball Screw Pairs Based on Gramian Angular Field and Parallel CNN

Aiming at the difficulty in diagnosing the lubrication state of ball screw pairs, a lubrication state diagnosis method for ball screw pairs based on Gramian Angular Field （GAF） and parallel Convolutional Neural Network （CNN） is proposed. The lubrication states of ball screw pairs under different working conditions are defined as well-lubricated, moderately lubricated and poorly lubricated. Vibration signals corresponding to the three lubrication conditions are collected. One-dimensional time-series signals are converted into two-dimensional time-frequency images by Gramian Angular Field transformation, to deeply extract the inherent features of Gramian Angular Summation Field and Gramian Angular Difference Field that contain lubrication state information. The fused image features are fed into the parallel convolutional neural network to implement adaptive feature learning and state classification. Experimental results demonstrate that the proposed method possesses high accuracy and favorable stability for the lubrication state diagnosis of ball screw pairs.