Abstract:

The generalized regression neural network (GRNN) is introduced into the elastoplastic displacement back analysis in three dimension of Wushi tunnel in virtue of its merits, such as good approximation capability, fast learning speed and excellent network stability. In order to find the optimal threshold value of GRNN model during training course, the genetic algorithm (GA) is combined with it to form the GA-GRNN algorithm. After determining the optimal nonlinear mapping between the numerical model parameters and the displacements, GA is used to search the elastoplastic model parameters which can minimize the error between the calculation and measured displacements of Wushi tunnel. The parameters from back analysis are inputted into the GRNN model to forecast displacement of the next construction step; and the results are very close to the measured displacement. Therefore, it is concluded that this back analysis method is feasible in tunnel engineering and can be used in similar engineering.

Key words: tunnel, numerical calculation, generalized regression neural network, genetic algorithm, displacement back analysis