Bayesian-Based Hybrid Simulation Approach to Project Completion Forecasting for Underground Construction

Abstract

Real-time simulation is powerful in forecasting the completion probability of long-term projects with repetitive tasks but fails to consider the time-varying uncertainty of inputs caused by construction process variabilities. In this paper, an improved method is introduced for predicting the time-varying probability of project completion of ongoing underground cavern group projects using Bayesian updating techniques. Within a tailor-made hierarchical simulation model, the Bayesian approach is adopted to constantly update duration distributions of unfinished project activities according to onsite data. The probability of project completion can therefore be increasingly refined during the process. The methodology is further explained in a case study where its feasibility and advantage over traditional approaches are verified. The success may also be replicated in addressing other similar time-varying uncertainty issues inherently present in almost all construction projects.