The objective of our proposal is to develop and demonstrate high-performance embedded algorithms for smart valves that exhibit improved ability to detect damage (i.e., leaks and ruptures) to shipboard fluid networks. Phase I of the proposed effort will focus on synthesizing non-deterministic smart valve sensor models and parametric and non-parametric detection algorithms. Specifically, we shall derive and validate non-deterministic models that describe how smart valves estimate the states of the fluid system (e.g., flows and pressures) during normal conditions, and when fluid network damage exists. Development of these models will assist us in applying decision theoretic and detection theoretic algorithms to detect and classify damage events. They will permit us to compare these approaches to current techniques for detecting fluid network damage by comparing their detection capabilities (e.g., detectible size and location of ruptures; effect of sensor placement and sensor noise on rupture detectability). In Phase II we shall also experiment with multi-hypothesis and “intelligent” detection techniques such as neural networks and genetic algorithms in conjunction with smart valve technology to further improve the performance of damage detection algorithms (especially in the areas of response speed and computational complexity).
Keywords: Smart Valve, damage detection, Automated damage control, Flow estimation, Fluid system automation
