Novel hybrid MFO-XGBoost model for predicting the racking ratio of the rectangular tunnels subjected to seismic loading

Abstract

This study proposes a novel hybrid MFO-XGBoost model that integrates the moth-flame optimization (MFO) algorithm and the extreme gradient boosting (XGBoost) to predict the racking ratio of rectangular tunnels subjected to seismic loading. For this purpose, a nonlinear finite difference model of soil-tunnel considering a realistic partial-slip condition is developed and validated against centrifuge test results. Then, 2040 dynamic simulations subjected to 85 ground motions are analyzed to cover a comprehensive suite of soil-tunnel configurations. Based on the generated database, the MFO-XGBoost model is constructed to capture the relationship between various effective parameters and the racking ratio of the rectangular tunnel. The obtained results are compared with those of four existing models to evaluate the performance of the proposed MFO-XGBoost model. The comparison reveals that the proposed MFO-XGBoost model captures well the numerical results of the racking ratio and outperforms other models. Among twelve input variables, parameters with primary and secondary influences are identified. Finally, a web application is built based on the proposed MFO-XGBoost model to calculate the racking ratio of rectangular tunnels, which is computationally more effective compared with alternative procedures.