Back Analysis of Geomechanical Parameters Using Hybrid Algorithm Based on Difference Evolution and Extreme Learning Machine

Abstract

Since the geological bodies where tunnels are located have uncertain and complex characteristics, the inverse problem plays an important role in geotechnical engineering. In order to improve the accuracy and speed of surrounding rock identification, the back analysis objective function with usage of the displacement and stress monitoring data has been constructed, with a hybrid algorithm proposed. An extreme learning machine (ELM) is employed with optimal architecture trained by the difference evolution (DE) arithmetic. First, the three-dimensional numerical simulation is used in the creation of training and testing samples for ELM model construction. Second, the nonlinear relationship between rock parameters and displacement is constructed by numerical simulation. Finally, the geophysics parameters are obtained by DE optimization arithmetic taking into consideration the monitoring data including both displacement and pressure. This method had been applied in the Fusong highway tunnel in Fusong City of China’s Jilin Province, with a good effect obtained. It takes full advantage of DE and ELM and has both calculation speed and precision in the back analysis.