Inversion of Soil Parameters Based on Extreme Learning Algorithm and its Research and Application in Subsidence Prediction

Abstract

The results of numerical calculation of bearing performance of bridge pile foundation are closely related to the selection of geotechnical parameters, and it is of great significance for engineering construction to obtain reasonable geotechnical parameters quickly and accurately in order to accurately analyze the bearing force of bridge pile foundation. Therefore, this paper uses a sparrow search algorithm combined with the displacement inverse analysis method of the extreme learning machine to generate a representative combination of soil parameters based on the orthogonal design using the displacement monitoring data of the top of the pile of the railroad bridge beam as the basis, and finite element simulation as the training samples for orthotropic analysis, and the sparrow search algorithm is used to find the optimization for the model of the extreme learning machine, and the mapping relationship between the inverse parameter and displacement is constructed using the ELM model, and finally the inverse results are brought into the finite element simulation to obtain the training samples. Finally, the inversion results are brought into the finite element to obtain the final displacement. In order to verify the performance of the optimized model, the model is evaluated according to the most common statistical indexes, i.e., the coefficients of determination R², and MAPE, and the evaluation indexes for the test set are: MAPE=0.0031, R2=0.9884. The results show that: the optimized ELM model of the SSA converges quickly, consumes a short period of time, and the displacement response of the parameters obtained from the inversion in the finite element is in good agreement with the detection data, and the maximum error of the inversion result is 0.9884. data are in good agreement, and the maximum error is only 7.05%, which verifies the accuracy and practicality of the method.