Data-Based Modeling Methods for Fault Detection and Isolation in Heat Exchangers

A multivariate analysis method is developed for processing measurements, and for detecting and isolating faults and monitoring performance degradation in heat exchanger control loops. A heat exchanger in a typical temperature-to-flow cascade loop s considered. A mechanistic thermal-fluid model for the components in the system is developed and compared to an installed laboratory heat exchanger control loop. A supplemental model for condenser hear transfer is included. The mechanistic model generate data to develop a data driven model using the Group Method of Data Handling (GMDH) approach. The GMDH model matches the mechanistic model well.

A Fault Detection and Isolation (FDI) rule-base is formulated from results of simulations performed using these models. The rule base allows the identification of faults in a heat exchanger control loop given suitable process measurements. The mechanistic model matches the physical system performance well and is used to create a Fault Detection and Isolation (FDI) algorithm for the system.