Prediction Of Tool Consumption In TBM Construction Based On Support Vector Machine Regression

In recent years,China’s urban subway,water conservancy tunnel,railway tunnel and other underground projects have achieved unprecedented development.TBM construction method is widely used in the construction of deep-buried long tunnels because of its significant advantages in environmental protection,safety and efficiency.The basic principle of TBM tunneling is to cut and break the rock surface by means of the dual action of the thrust of the propulsion device and the torque of the rotary cutter head.The hob is installed on the cutter head and is the main rock breaking tool for TBM tunneling.Due to the complex geological conditions and changeable tunneling parameters in the process of tunnel construction,there is a great deal of uncertainty in the prediction of tool consumption.The increase of TBM cutter consumption will increase the time of cutter replacement,maintenance and construction cost,resulting in the decrease of TBM construction efficiency.Therefore,it is of great significance to predict the consumption of TBM construction tools scientifically and reasonably for the on-site construction management of TBM projects.In this paper,the prediction of cutting tool consumption in TBM construction based on support vector machine regression is studied to improve the scientific management of cutting tool consumption in TBM project.Firstly,through literature research,construction data and expert investigation,four factors affecting tool consumption,including geological factors,tunneling parameters,construction organization and other factors,including 14 secondary indicators,are identified.Grey relational analysis(GRA),decision laboratory analysis(DEMATEL)and random forest algorithm(RF)are used to analyze the correlation of indicators,which not only fully considers the nonlinearity and lack of information among influencing factors,but also exerts the practical engineering knowledge and experience of experts to optimize the evaluation indicators.In order to solve the problem that the prediction accuracy of the model is reduced due to information redundancy;Secondly,an optimized support vector machine regression(SVR)model is established to predict the cutting tool consumption in TBM construction,and sparrow search optimization algorithm(SSA),artificial bee colony optimization algorithm(ABC)and particle swarm optimization(PSO)are respectively used to optimize the SVR core parameters and penalty factors,and the optimization algorithms that can improve the prediction performance of SVR model are compared;Then the model is applied to a diversion tunnel project in northern Xinjiang to carry out prediction analysis,144 groups of engineering data are collected for training and testing,the evaluation parameters are selected to evaluate the prediction performance of the model,and the error is compared and analyzed.The results show that the SSA-SVR model has higher prediction accuracy in the case of limited data samples and multi-factor indicators.Compared with other models,the four evaluation parameters of SSA-SVR model are optimal,and the prediction error range is smaller.The model has good practicability,simple operation and high prediction efficiency.Finally,the method of TBM tool management is put forward according to the prediction results.The intelligent prediction model of cutting tool consumption in TBM construction established in this paper has certain adaptability and applicability,and has reference significance for improving the prediction efficiency of cutting tool consumption,mastering the cost of cutting tool in the field,and speeding up the management of TBM construction schedule.