Research On Fusion Prediction Model For Shield Equipment Fault Based On Ensemble Learning

With the rapid development of urban underground space construction in China,the shield machine has become the most important construction machinery and equipment for developing urban underground space,but its structure is complex and difficult to operate.At the same time,construction in a complex underground environment,the shield machine cannot avoid the influence of many harsh external factors,which can make the equipment wear and tear consumption serious,or even multiple failures at the same time.The state of shield tunneling machine directly affects the progress,quality and safety of construction projects,and then affects the economic benefits of enterprises.Therefore,this paper studies the improvement of equipment maintenance and management methods for large mechanical equipment with complex structure such as shield machine,and establishes a fault predictive maintenance management model for shield machines based on an integrated learning approach.Through modeling and analysis of data from the construction process,it predicts possible failures in advance,helps equipment operators respond early,and assists equipment maintenance managers in making targeted equipment maintenance management decisions.In this paper,the equipment predictive maintenance management,shield equipment failure prediction and other related domestic and foreign literature are reviewed,the characteristics of equipment predictive maintenance management are summarized,and the failure predictive maintenance management mode for shield equipment failure is proposed.At the same time,the core module of the mode--the failure prediction model is designed.Considering the complex structure of the shield machine,the many construction parameters and the harshness of the construction environment,which can trigger the problem of multiple equipment failures,if a single machine learning method is used to construct a model for failure prediction,the accuracy of its prediction will be limited.Therefore,in this paper,a multi-model integrated shield fault prediction method based on Stacking is designed in combination with integrated learning related prediction techniques.The designed prediction model is trained using fault data from different projects of different shield machines to accurately predict impending shield machine faults.In order to make the prediction of the constructed Stacking shield equipment fault prediction integrated model more accurate,this paper uses Bayesian optimization algorithm to optimize the model.In this paper,the principle of Bayesian optimization algorithm is introduced in detail and the flow of shield fault prediction model based on Bayesian optimization Stacking is presented.The experimental results of the four single classification models and the Stacking integrated model are compared and analyzed before and after optimization.The practice shows that the Bayesian optimized Stacking shield fault prediction integrated model has higher accuracy than the four single classification models and can predict the faults to occur during the operation of the shield more accurately.Finally,in order to make the shield equipment fault prediction model constructed in this paper have auxiliary management function,this paper designs and implements the prototype of shield equipment fault prediction visualization system.Combined with the prediction results,the importance and function of different construction parameters in shield equipment fault prediction are analyzed by using SHAP method,so as to help the equipment operators to make quick and effective decisions and assist the equipment maintenance managers to make or improve the equipment maintenance management plan.To sum up,this study proves the feasibility and effectiveness of applying integrated learning method,Bayesian optimization method and other technologies to the maintenance and management field of large equipment such as shield machine.The research results of this paper can make construction enterprises become more intelligent and automated in equipment maintenance and management,and provide an effective tool for making reasonable equipment predictive maintenance decisions.