GENERIC MODELLING OF HVAC PLANT FOR FAULT DIAGNOSIS

摘要

A methodology is presented for creating models whichrnare suitable for use in fault detection and diagnosisrnschemes in applications where it is impossible to obtainrndata from the actual plant. Generic qualitative modelsrnbased on fuzzy rules are used to describe the basicrnfeatures of the behaviour of a class of plants of similarrndesign. The generic models are identified off-line fromrntraining data produced by computer simulation ofrntypical plant designs. The method uses a datarnclustering algorithm to assist in identifying thernstructure and a fuzzy identification algorithm tornestimate the parameters of the model from the trainingrndata.- A measure of the degree of similarity betweenrnfuzzy models is introduced to determine the extent tornwhich the behaviour of individual plants is similar tornthat of the generic model.rnThe method is used to generate generic models of arncooling coil subsystem when it is operating correctlyrnand when the coil has become fouled. Results arernpresented which show that the generic models can bernused to describe a class of plant designs withoutrngreatly increasing the ambiguity associated with thernfault diagnosis.