A technique of functional redundancy (as opposed to hardware redundancy) for detecting incipient failures in process instruments is applied to a simulation of the loss-of-fluid test pressurizer. The failure detection scheme consists of a set of five Kalman filters and a logical means for combining estimated state variables with instrument signals to produce decision functions, which identify faults, as they occur, in each of five instruments. Test data from the simulated plant show that prompt detection of both bias faults and high noise faults is possible during small transient fluctuations in the pressurizer from its nominal operating state.