Two Approaches to Optimal Sensor Locations

Abstract

Accurate modeling of thermal systems depends upon the determination of the material properties and the surface heat transfer coefficients. These parameters are frequently estimated from temperatures measured within the system or on the surface or from measured surface heat fluxes. Because of sensor errors or lack of sensitivity, the measurements may lead to erroneous estimates of the parameters. These errors can be ameliorated if the sensors are placed at points of maximum sensitivity. This paper describes two methods to optimize sensor locations: one to account for signal error, the other to consider interacting parameters. The methods are based upon variants of the normalized Fisher information matrix and are shown to be equivalent in some cases, but to predict differing sensor locations under other conditions, usually transient.