Finding Best Model to Forecast Construction Duration of Road Tunnels with New Austrian Tunneling Method Using Bayesian Inference

Abstract

Forecasting project final duration (i.e., time at completion) is crucial to project risk management and is always sought by project managers during the construction period. Because of a strong correlation between past and future performances in linear projects, past progress data are the best source of information to forecast final duration of this type of project, including tunneling projects constructed by the new Austrian tunneling method (NATM). Bayesian inference is a robust probabilistic approach that can provide accurate forecasts of final duration based on a project's past performance. However, results of research in this field have shown that selecting an appropriate model, which represents the unknown pattern of the project's actual progress well, is the most challenging and subjective part of this approach. Effective risk management necessitates looking for the best model that can forecast project final duration accurately and precisely, especially early in the project. This research was aimed at finding a best progress model for NATM tunneling projects by conducting Bayesian analysis on available data of a massive project, the Niayesh highway tunnel in Iran. The analysis showed that the dual Gompertz function (with flexible lower asymptote) was the most reliable model for this purpose. The results of this research bring advantages to the planning and risk management of NATM tunneling projects, which are discussed in this paper, and can be very useful for future NATM tunnel constructions.