Adaptive filtering for calibration of redundant signals

This paper presents formulation and validation of an adaptive filter for real-time calibration of redundant signals consisting of sensor data and/or analytically derived measurements. Individual signals are calibrated on-line by an additive correction that is generated by a recursive filter. The covariance matrix of the measurement noise is adjusted as a function of the a posteriori probabilities of failure of the individual measurements. An estimate of the measured variable is also obtained in real time as a weighted average of the calibrated measurements. These weights are recursively updated in real time instead of being fixed a priori. The calibration and estimation filter is hosted in a Pentium-II platform and has been tested in real time on an operating power plant. The filter software is portable to any commercial platform. The filter can be potentially used to enhance the Instrumentation and Control System Software in tactical and transport aircraft, and nuclear and fossil power plants.