Research on the Optimization Algorithm Combination for Surface Deformation Prediction during Tunnel Construction

• 1、School of Management, Wuhan University of Science and Technology, Wuhan, 430081

Abstract：This paper aims to improve the prediction accuracy of surface deformation during tunnel construction, and has constructed a hybrid prediction framework (SVM-RFE-HHO-ELM) that combines support vector machine recursive feature elimination with Harris eagle optimization algorithm to optimize the extreme learning machine. This framework first selects 15 key parameters affecting surface deformation through SVM-RFE; on this basis, the HHO is used to optimize the hyperparameters of ELM to establish a deformation prediction model. Experimental results show that this model performs well in multiple indicators. Moreover, LIME, an interpretable model-independent explanation method, can identify the degree of influence of each construction parameter on deformation, reveal the key construction control factors, and the proposed hybrid framework provides an effective method for surface deformation prediction and safety control during tunnel construction process.

Keywords： Civil engineering construction Tunnel construction Surface deformation Extreme Learning Machine