Utilizing Artificial Intelligence and Finite Element Method to Simulate the Effects of New Tunnels on Existing Tunnel Deformation

Abstract

Buildings and infrastructure should be constructed while maintaining the safety of the existing infrastructure. Since tunnels are an essential component of urban infrastructure, building new tunnels next to existing ones may have unintended consequences. This study aims to examine the effects of building a new tunnel next to an existing tunnel under different conditions. Two-dimensional models have been developed using the finite element method (FEM) for the existing tunnel, with a diameter of 5 meters, and for the new tunnel, with a diameter that differs from the existing tunnel. Then, the effects of two factors, namely the distance between the two tunnels and the diameter of the new tunnel, were examined on the horizontal, vertical, and total deformation of the soil and the existing tunnel. For the first time, this study attempts to predict the effect of new tunnels on existing tunnels using artificial intelligence (AI) methods. Due to the importance of tunnels and the increasing number of tunnels being constructed in urban areas, this issue will be of great interest. For analyzing the feasibility of using mathematical methods to predict tunnel deformation, the multiple linear regression (MLR) method and an AI technique, namely classification and regression random forests (CRRFs), were used to utilize the generated database. Analyzing FEM results represented that by increasing the diameter of the new tunnel from 3 to 5, horizontal and vertical displacement of the existing tunnel increased by approximately 5 and 10 times, respectively. Further, by reducing the distance between the new tunnel and the existing tunnel from 11 meters to 6 meters, the intensity of horizontal and vertical deformation of the existing tunnel increased by about 2 and 3 times, respectively. Moreover, the results of mathematical models demonstrated that the CRRF method was more accurate than the MLR method, with R of 0.94 and mean absolute error of 2.89 for the testing database, which indicated its proper performance. Received: 1 July 2022 | Revised: 5 August 2022 | Accepted: 12 August 2022 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data available on request from the corresponding author upon reasonable request.