Calculating of the tunnel face deformations reinforced by longitudinal fiberglass dowels: From analytical method to artificial intelligence

Abstract

Deformation calculation of the reinforced tunnel is always the key and difficult problem in tunnel support design. To achieve simple and rapid tunnel performance evaluation, this paper attempts to establish a novel calculation system based on the artificial intelligence, which is transitioned from an analytical method named the Convergence-Confinement Method (CCM) based on the spherical symmetry hypothesis. 200 simulations were completed by using an analytical method to describe the reinforced tunnel behavior and seven parameters were considered to calculate the tunnel face deformation. The Boruta algorithm with a random forest (RF) model was utilized to eliminate unnecessary parameters in the analytical method to build high-precision prediction model. The performance evaluation results indicated that the prediction model based the artificial intelligence has obtained excellent accuracy for estimating the deformation of the tunnel face reinforced by the longitudinal fiberglass dowels. Furthermore, the Geological Strength Index (GSI) is the most important parameter for predicting the deformations of tunnel face in the weak rock masses. Interestingly, the tentative parameter, rock type (mi), cannot be ignored when using the proposed artificial intelligence model to predict the deformations of a tunnel face reinforced by longitudinal fiberglass dowels. In general, this successful transition helps popularize the application range of analytical methods and improve the prediction accuracy for the deformations of reinforced tunnel.