Research on deformation prediction of tunnel surrounding rock using the model combining firefly algorithm and nonlinear auto-regressive dynamic neural network

Abstract

Tunnel surrounding rock deformation is dynamic, sensitive to time and space, nonlinear, and highly complicated. By combining the firefly algorithm (FA) and nonlinear auto-regressive (NAR) dynamic neural network method, an algorithm model was proposed for predicting dynamic nonlinear surrounding rock deformation. The FA improved the prediction accuracy of the NAR dynamic neural network by determining the optimum values of two network parameters—delay order and number of units in the hidden layer; combined with the monitoring results of Beishan exploration tunnel (BET), this is demonstrated by a comparative analysis of predictions yielded by the FA–NAR dynamic neural network and by the least squares support vector machine (LS-SVM). In general, the comparation shows that the FA–NAR dynamic neural network model yielded predictions that are fundamentally consistent with measurements and exhibits higher prediction accuracy than the LS-SVM. Results also show that the surrounding rock deformation prediction of BET for March 4, 2020 was marginally smaller than 2.43 mm.