Research On Prediction Method Of Remaining Useful Life Of Shield Machine Tools Based On Big Data

With the rapid construction and development of urban underground space,the shield method has been widely used in China’s underground engineering construction due to its inherent advantages such as fastness,safety,and environmental protection,and has gradually become one of the most important construction methods for urban underground tunnel construction.In the process of shield tunneling,the problem caused by tool degradation is one of the main difficulties faced during shield tunneling.Degradation of shield machine tools will increase the probability of accidents,affect construction efficiency,and threaten personal safety.To avoid these problems,it is necessary to open the warehouse and check the tools at a high frequency.The high-frequency opening inspection will affect the construction progress and waste human resources.Aiming at this contradiction,this paper proposes a method for predicting the Remaining Useful Life(RUL)of shield cutters.The main research contents are as follows:(1)The overall research framework of cutting tool RUL prediction for shield machine is constructed.The structural composition of the shield machine is sorted out,and the common failure modes of the shield machine tools are analyzed and summarized.Based on this,a method for predicting the RUL of the shield machine tools based on big data mining technology is proposed: firstly preprocess the data and feature engineering processing,and then based on deep learning algorithms to construct comprehensive health indicators that characterize the health status of shield machine tools,and finally use time series analysis algorithms to predict future degradation trends of health indicators.(2)Dimension reduction of the features of shield machine tools.Aiming at the features of large-dimensional and noisy data of the shield machine,on the one hand,the features of the shield machine are manually selected based on the mechanism knowledge and expert experience;on the other hand,the data is reduced by the dimensionality reduction algorithm to provide subsequent analysis a good data foundation.(3)The health indicators of shield machine tools based on Convolutional Neural Networks(CNN)is constructed.Aiming at the features of complex and changeable working conditions,poor trend of features,and large amount of data during the tunneling process of the shield tunneling machine,a method for constructing the health indicators of shield tunneling machines based on CNN was proposed.Utilize the non-linear transformation of deep learning to mine deeper information from data with poor trending.Utilizing the characteristics of sparse connections and weight sharing of CNN models improves the efficiency of model construction,and also reduces the risk of overfitting,and provides support for the trend prediction of subsequent health indicators.(4)The RUL prediction model of shield machine tools is built.Based on the health indicators,a Long Short-Term Memory(LSTM)network is used to predict the trend of the health indicators,and the time from the failure threshold is calculated based on the predicted degradation trajectory to achieve the prediction of the remaining service life of the shield cutter.Based on the above research content,an example analysis of the RUL prediction of the shield cutter is completed,and the validity of the method in this paper is verified by real historical data recorded in actual construction.