A Robust Failure Diagnostics Scheme for Nonlinear Thermofluid Processes

The catastrophic failure of many physical systems is often precipitated by degradation of minor system components. A real time microprocessor based diagnostic scheme may detect system abnormalities and apply various strategies to determine whether compensatory action is required. In this paper, a robust innovations-based failure detection methodology for improved heat pump performance is presented. A sixth-order nonlinear model and an extended Kalman filter were used to reconstruct the system´s state based upon limited sensor feedback. The estimated mass flow rates and pressures were statistically monitored and failures predicted based upon the repeated violation of an alarm threshold. A series of heat pump failures are experimentally induced and representative results presented and analyzed to demonstrate the potential advantages of this diagnostic approach.