Health State Prediction Of Lathe Components Based On Deep Learning And Digital Twinning

With the rise of artificial intelligence,the development of high-end manufacturing industry integrating information technology is the trend,and CNC machine tools,as complex and high-precision manufacturing equipment,have become an indispensable maintenance management tool for health status prediction.In this paper,we propose LSTM-CRF and Serial Informer network prediction model through the study of deep learning-based temporal prediction method to solve the current problems in CNC lathe health prediction,and establish digital twin health state prediction model to realize the response output of prediction results,so as to provide a new research idea to improve the reliability of CNC lathe.The research of this paper is mainly divided into the following contents:(1)In order to improve the prediction accuracy of the network model for time-series data,a combined LSTM-CRF neural network prediction model is proposed,the CRF framework is introduced to construct the state transfer matrix through the change relationship of adjacent data,the probability distribution of the prediction sequence is obtained by using the inverse decoding of the Viterbi algorithm,and the state sequence with the highest probability is the optimal prediction result.The prediction sequence is calculated by the offset matrix of feature data,which reflects the change trend of decreasing capacity more intuitively.The CNN network is also added to strengthen the learning of feature data at different scales,and the LSTM network is fused to capture the relational information in long time sequences.The experimental results demonstrate that the LSTM-CRF model has better prediction accuracy compared with other prediction models.(2)To increase the prediction ability of the model for long time series data,this paper proposes a Serial Informer network model based on the informer architecture,which makes the model pay more attention to the time-dependent information when predicting the problem.The serial self-focus mechanism is used to explore the change relationship between multidimensional feature data sequences and health state sequences to achieve the serial prediction capability of the target,while the existence of Encorder and Decorder in the model enables Serial Informer to complete end-to-end prediction of battery data.For the input multidimensional feature data,the corresponding health state assessment system is constructed through feature screening and correlation analysis.The feasibility and validity of the model were verified on the rolling bearing full life cycle dataset provided by the University of Cincinnati.(3)Using unity3 D to realize the geometric attribute simulation modeling of CNC lathe,build the digital twin health state prediction system from four aspects: equipment interface module,data processing module,health state prediction module and intelligent maintenance decision module,and combine LSTM-CRF and Serial Informer prediction models to realize the health state prediction of CNC lathe equipment The system combines LSTM-CRF and Serial Informer prediction models to achieve the health state prediction of CNC lathe equipment.