A risk-based dynamic decision support system for tunnel construction

Efficient tunnel construction planning requires that contractors determine the optimal sequence of tunnel excavation methods and primary support systems based on available information. Important tunneling factors, including geologic uncertainty, uncertainty in the productivity of tunneling processes, the dynamics of tunneling operations, and a contractor's risk sensitivity influence a contractor's decisions significantly and must be addressed directly.; This dissertation develops a computerized risk-based dynamic decision support system that addresses the aforementioned issues by structuring a contractor's tunneling decisions as a risk-sensitive dynamic probabilistic decision process based on three interrelated models. The probabilistic geologic prediction model characterizes both uncertainty and variability of geologic conditions along the tunnel profile in the probabilistic form of ground class transitions. The probabilistic tunnel cost estimating model evaluates tunneling time and cost performance for different excavation and support methods as applied to different prevailing ground classes by using discrete-event simulation. The resulting ground class transition probabilities and tunneling unit costs determined by both models provide inputs for the risk-sensitive dynamic decision model. To capture a contractor's risk sensitivity, the tunneling unit costs are converted to a utility unit using the contractor's exponential utility function. Based on the assumption of constant risk aversion, the risk-sensitive stochastic dynamic programming model can be solved by using backward recursive fixing algorithm. The final outputs are optimal tunneling policies and estimated costs for the project, both of which reflect the contractor's risk preference. The results from applying the proposed system in a case study show that it can determine optimal tunneling policies and cost estimates as a function of a contractor's degree of risk sensitivity.; The significance of this dissertation stems from the fact that it has accomplished several firsts. It incorporates uncertainty in both geologic prediction and performance of tunneling processes. It establishes a probabilistic estimating procedure based on a contractor's work breakdown structure. It is also the first system that can be used to systematically price the risk associated with tunneling and to investigate the effects of a contractor's risk preference on his optimal tunneling policies and cost estimates.