A personal-computer-based neutron and thermal power validation system has been developed for the Taiwan Research Reactor (TRR). An extended Kalman filter is used to reduce the neutron power signal noise, and an identified neutron power/thermal power empirical model is used instead of a theoretical physical model for the analytic redundancy model. The parity-space technique is adopted to estimate the measurements and identify sensor failure. Performance of the system is evaluated using data from a TRR simulation model and the results show that common-mode failure is successfully identified without additional sensors.