Study on Life Prediction Method of Ball Screw Base on Constructed Degradation Feature and IGWO-BiLSTM

Regarding the problem of reduced remaining useful life (RUL) due to wear ball screw in feed system CNC (computer numerical control) machine tools, a prediction method based on constructing degradation feature vector signal data and improved gray-wolf optimization with bidirectional long short-term memory (IGWO-BiLSTM) neural network regression model is proposed. Firstly, time-domain analysis complete ensemble empirical mode decomposition adaptive noise (CEEMDAN) were carried out collected cycle screw. The energy each IMF (intrinsic function) component after CEEMDAN constructed. Pearson correlation coefficient was used filter vectors construct multivariate vector. Secondly, this paper improves traditional gray wolf algorithm, adds search strategy dimension learning, combines algorithm BiLSTM model, IGWO-BiLSTM theory. A between ball-screw established. Finally, established according proposed compared other five models: LSTM, BiLSTM, BO-LSTM (Bayesian LSTM), BO-BiLSTM, IGWO-LSTM. results indicate that has high accuracy good generalization ability for predicting system.