An improved CS-SVM model for slope deformation forecasting under a BIM-based monitoring system

Abstract

This study aims to develop an efficient slope monitoring and deformation prediction method to address the geological disaster problems caused by slope deformation in infrastructure construction. The slope monitoring system and deformation prediction model were studied and constructed based on the building information model, support vector machine, and cuckoo search algorithm. Experiments showed that the improved cuckoo search algorithm had the best convergence and the optimal solution quality. The hypervolume reached 0.977, the inverse substitution distance reached 0.108. Moreover, this method had the fastest solution time, and the highest coverage of inflection points in the solution set. The optimized deformation prediction model had a strong generalization ability and achieved the best prediction effect in practical applications. The deviation between the predicted value and the true value was the smallest. The prediction accuracy of the optimized model was the highest, and the sum of squared errors reached 0.121. The prediction effect and efficiency of this method were the best, and the prediction time reached 6.46 seconds. This research can promptly identify potential safety hazards, provide theoretical support for formulating scientific slope treatment plans, and reduce casualties and property losses.