Mix Proportion Design of Similar Materials for Tunnel Surrounding Rocks Based on GA-BP Neural Network
(1. Key Laboratory of Transportation Tunnel Engineering of the Ministry of Education, Southwest Jiaotong University, Chengdu 610031;2. Nation Engineering Research Center of Geological Disaster Prevention Technology in Land Transportation, Southwest Jiaotong University, Chengdu 610031)
Abstract：
To control and optimize the physical and mechanical parameters of surrounding rock similar materials in tunnel model experiments, a GA-BP neural network algorithm was developed for parameter prediction. The neural network structure comprises a three-node input layer, a seven-node hidden layer, and a three-node output layer.The Genetic Algorithm (GA) was employed to optimize the weights and thresholds of the BP neural network. The input parameters included the content of river sand, fly ash, and motor oil, while the output parameters included density, cohesion, and internal friction angle. Using measured data as samples, the model's performance was thoroughly evaluated by comparing the mean square error, absolute error, and relative error of the BP neural network before and after GA optimization. Based on the analysis, a mix proportion design method for surrounding rock similar materials under a given similarity ratio was established.The results indicate that the GA-BP neural network algorithm can effectively fit and predict the physical and mechanical parameters of surrounding rock similar materials. Compared to the traditional BP neural network, the GA-BP neural network achieves lower prediction errors and higher accuracy. The prediction model based on the GA-BP neural network can quickly and accurately determine the range of raw material mix ratios under a given similarity ratio, significantly reducing the number of repeated experiments.
ZHANG Xinyang1,
2 SHEN Yusheng1,
2 CHANG Mingyu1 etc
.Mix Proportion Design of Similar Materials for Tunnel Surrounding Rocks Based on GA-BP Neural Network[J] MODERN TUNNELLING TECHNOLOGY, 2024,V61(6): 82-91
2024, Vol. 61 