Automated quantitative model-based fault diagnosistic protocol via Assumption State Differences

Abstract

This treatment describes the details of a systematic protocol useful for performing optimal automated process fault analysis. This implementation generalizes the underlying Boolean logic version of the Method of Minimal Evidence (MOME) developed previously to a highly comprehensive algorithm for performing model-based fault diagnostics. This generalization allows for a more compact treatment of potential single and multiple fault situations, at all levels of possible diagnostic resolution, with both elegant and efficient uniform sensor validation and proactive fault analysis (SV&PFA) diagnostic rules for diagnosing those situations. This Assumption State Differences (ASD) Protocol version of the MOME algorithm thus automates the diagnostic reasoning necessary to continuously perform optimal process fault analysis so that only the underlying well-formulated models are required to achieve such performance. Using this algorithm consequently directly simplifies the solution of the more complicated problem of automated process fault analysis into the much more tractable, and incrementally solvable, problem of adequately modeling normal process operations.