Dynamic-model-based anomaly diagnosis as an operator aid

An approach is described which is a combination of simple lumped parameter modelling, parsimonious search through a regression model lattice and forward chaining through a production system. Four separate functional requirements are identified, each function is outlined and its role elucidated by referring to the diagnosis of one particular abrupt anomaly in a liquor storage area.