Noncontact temperature measurement. II. Least squares based techniques

A technique for the noncontact measurement of temperatures is described. The technique is based on the measurement of the emitted intensity at multiple wavelengths and the simultaneous calculation of emissivity and temperature through the use of least squares curve fitting techniques. The technique is shown to make no assumptions regarding the emissivity of the target except that it be possible to model it with an analytic function. A theory is developed to predict the errors in the estimation of temperatures based on both linear and nonlinear least‐squares techniques. It is shown that the maximum error in the predicted temperature is related to the noise in the measured intensities in a quantifiable manner. It is shown using computer simulations that the theory and algorithms developed here can predict both the temperatures and the uncertainty associated with each temperature prediction with a very high accuracy. An instrument was developed to test this theory. This instrument, referred to as the MITTMA, was used to measure absolute temperatures of various sources from 900 °C to 2300 °C with an average accuracy of approximately 0.5%.