Abstract
The applicability in practice of a diagnostic tool is strongly related to the physical transparency of the un- derlying models, for the interpretation of the relationships between the involved variables and for direct model inspection and validation. In this work, a methodology is developed for transforming an opaque, fuzzy clustering-based classification model into a fuzzy logic model based on transparent linguistic rules. These are obtained by cluster projection with appropriate coverage and distinguishability constraints onto the fuzzy input partitioning interface. The methodological approach is applied to a diagnostic task con- cerning the classification of simulated faults in the feedwater system of a nuclear Boiling Water Reactor.