Facing reality : design and management of flexible engineering systems

This thesis proposes a practical approach to defining flexible design and development strategies for maximizing the expected value of engineering systems. Specifically, the approach deals with the fact that it is generally computationally impractical to explore all the possible ways a system might be developed and operated, given the large number of possible scenarios in which the system might evolve. To make the analysis tractable within the computational resources available, it proposes that designers and program managers use a catalog of representative operating plans built from combinations of design elements and management decision rules. These are associated with a range of possible scenarios of uncertain variables that might affect the system's expected value and performance. This work develops the novel methodology introduced by (de Neufville, 2006) to guide the search for catalogs of operating plans while aiming at minimizing computational effort. It assumes a model of the engineering system is available, together with several value/performance metrics such as Expected Net Present Value (ENPV) and Value At Risk and Gain (VARG). It uses an algorithm based on statistical experiment design, Adaptive One-Factor-At-a-Time (OFAT) (Frey and Wang, 2006; Wang, 2007), to search the combinatorial space in light of system's responses to a limited set of uncertain variable scenarios.