Safety Thickness of the Tunnel And Cave Based on Morris-BP Neural Network

According to the sampling principle of Morris method, 100 groups of samples about influence parameters of minimum safety thickness between tunnel and cave were obtained, then based on the different parameters of the sample, the FLAC3D software was used to simulate the tunnel excavation in karst area. Further, combined with global sensitivity analysis method of Morris, the "elementary effect" of each parameter was calculated by the value variation of one parameter in turn. Then, the mean and variance of the "elementary effect" of each parameter were also obtained. Moreover, according to the value of a and a, the effect of each parameter on minimum safety thickness was analyzed. After that, it screened the important influence factors and optimized to obtain 80 groups of samples with important influence factors; In addition, BP neural network model for predicting the minimum safety thickness was established; At last, the optimized sample were learned and trained with neural network. Research results showed that: the sensitive degree order of the minimum safety thickness corresponding to the nine parameters is as follows: C > phi > H > D > R > gamma > E > mu = T, when affecting the model output, the order of the interaction between the parameter and other parameters is: C > phi > H > D > R > gamma > E > mu = T; Based on the global sensitivity analysis, the modeling method of neural network model of the minimum safety thickness was proposed and improved the modeling efficiency, what is more, its inversion accuracy is generally greater than 90%; Inversion performance of BP neural network model was best when the hidden layer had six neurons. Finally, the model was verified by an engineering example.