Abstract

Tunnels offer considerable benefits over other alternative forms of transportation infrastructure. For example, in contrast to highways, tunnels ensure unobstructed land above remains available for all forms of land-use that can present notable economic, social, and environmental advantages. This feature is important in highly urbanized areas where usable land is scarce and acquiring land can put high financial pressure on available budgets. Tunnelling uncertainty is the main contributor to cost overruns which are mainly due to geological uncertainity. Therefore, accurate prediction of geological characteristics can lead to higher efficiency of tunnelling projects and prevent any cost or time overruns. In this paper, a random forest (RF) approach is utilized to model underground lithology using 65 boreholes along a tunnel project. In order to acquire best hyper parameters for the RF model a random search optimization algorithm is used. Also, different intervals along the boreholes are used to extract the lithology to find the sensitivity of the model to different interval length. The results show that the proposed model can predict the lithology with 86 percent accuracy. The model was successful in prediction of lithology type and their order along the boreholes in a case study example.

Introduction

Tunnels offer considerable benefits over other alternative forms of transportation infrastructure in terms of land-use and planning. This feature allows for greater flexibility in use of lands in highly urbanized areas where usable land is scarce and acquiring land can put high financial pressure on available budgets. Tunnelling uncertainty is the main contributor to cost over runs the which is mainly due to geological uncertainity. Therefore, accurate prediction of geological characteristics can lead to higher efficiency of tunnelling projects and prevent any cost or time overruns (Yan-lin et al. 2011; Hossin and Sulaiman 2015; Zhang and Zhu 2018).

This content is only available via PDF.
You can access this article if you purchase or spend a download.