The proposed research addresses the problem of determining the purpose and functional capacities of facilities whose functions are deliberately obfuscated by their operators. As it is very difficult to obtain information that is directly related to the function of such facilities, the focus is on augmenting this traditional, direct approach to analysis with information extracted from data that are only indirectly related to facility function. More specifically, the PI proposes to use a rigorous, systematic methodology for identifying facility function in situations where observation opportunities are limited. The feasibility of this approach is based upon a newly-discovered, and rather surprising, property possessed by a broad class of complex systems: despite their complexity, it is possible to extract deep, quantitative information about these systems using only limited observations of their behavior. This property will be exploited for facility function identification by formulating the problem as one of identifying the dynamical process required to realize the facility, and by developing a multiple models strategy for efficiently extracting information from the available data sets and performing the requisite (dynamical system) identification. The feasibility of the proposed identification strategy has been demonstrated in a preliminary proof of process study involving a real world facility.
