Research On Fault Prediction Model For Shield Construction Based On Deep Learning

With the development of urban construction in China,the construction projects of subways and underground pipe corridors in major cities are increasing.Shields are used as the main excavation equipment for underground engineering tunnel construction.Due to the complexity of the shield system and the harsh underground working environment,the shield often has various faults and has one or more faults occur at the same time during the construction process.These failures will not only affect the progress of the project,but also cause economic losses.In serious cases,it will also endanger the safety of the staff and the construction site.Therefore,relevant research on shield construction failure can help the operators to find problems in time,make effective decisions,adjust the construction status of the shield,and help the managers to carry out the project schedule and quality safety control of the shield construction,which helps the shield project management work of construction engineering.A large number of sensors distributed throughout the shield can generate and collect a large amount of construction parameter data during the excavation process.By combing the relevant research,it is found that there is a close relationship between construction parameters and faults of shields,but the related research is not deep enough.Therefore,this paper uses the data science method to explore the relationship between shield construction parameter data and shield faults.This research is conducive to controlling the progress,quality and safety of shield construction projects.For the shield construction,there are many kinds of shield faults at the same time in the actual construction process.Based on the method of data science and based on the related models and theories of machine learning and deep learning,this paper studies the method of predicting shield failure by using shield construction parameter data.Especially for the problem of how to identify the shield machine fault information hidden in the shield construction time series data,a prediction model based on Long-Short Term Memory(LSTM)is designed.Because the construction process of the shield is very complicated,the configuration of the excavation parameters,the external environmental state,and various geological forms are various dimensions that affect the construction effect of the shield.Therefore,this paper firstly builds a shield construction database based on various types of data which are collected during the shield construction process by different sensors.The parameters of the actual machine equipment collected during the shield construction process,the fault alarm data and the external environment data of the shield are integrated,and the shield construction database is established to provide training data for model supervision learning for shield fault prediction.Considering that the collected data contains the changes of shield construction parameters before and after the shield failure,and taking the advantages of LSTM to handle the time series data,the shield faults prediction process design based on LSTM model is carried out.The data preparation process of the shield failure prediction model is described,and the multi-label based classification evaluation index is introduced as the evaluation index of the model prediction result.Finally,the shield fault multi-label prediction model based on PCA-LSTM is established.Obtaining and integrating the construction data of the parameters of each subsystem of the shield from the shield construction database.The PCA method is used to remove the correlation between construction parameters,reduce the dimensionality of the construction parameters.And then the data integrate with the external environmental data and fault data as input to the LSTM model.The learning of faults correlation is performed by introducing a loss function adapted to multi-label.Experimental results show that the model has a good prediction effect on single fault and multiple faults.Shield construction failure is an important factor that affecting the risk management,schedule management and quality management of shield construction.This study is conducive to the project management of shield construction.