Abstract
Current methods for compressor vane control in a turbo-jet engine rely on measurements or estimates of compressor inlet temperatures and speed to determine the most favorable compressor stator vane angles over the course of an effectively infinite-horizon control decision process. During each instance of this process, the stator vane angles must be adjusted with a regular frequency to manage compressor efficiency and stability over the course of a flight. The optimality suffers from uncertainty in compressor inlet temperature, though the extent of deviation from optimality is unknown, as is the cost-benefit trade space for addressing this issue. This is an example of a more general problem in systems engineering, which is to estimate the cost of uncertainty in information with respect to its effect on deviations from optimality in infinite- and indefinite-horizon decision processes subject to a stochastic state machine that determines decision outcome. This thesis will develop a methodology for (A) estimating the cost of uncertain information in these kinds of sequential decision processes, (B) characterizing the trade space associated with addressing this problem, and (C) application to the jet engine design refinement problem.
