Multiple Inversion Techniques with Multispectral Pyrometry for the Estimation of Temperature and Emissivity of Liquid Niobium and 100c6 Steel

The article presents the estimation of temperature and emissivities of niobium and 100c6 steel around their melting points using techniques based on the minimization of the least squares norm or on Bayesian inference. Two models were considered for the spectral emissivity, including a linear variation with respect to the wavelength and independent spectral values. The measured data used for the solution of the parameter estimation problem were radiative fluxes collected from a five-wavelength pyrometer. The experimental apparatus used in this work was designed for the characterization of metal samples with sizes of few millimeters at high temperatures, combining aerodynamic levitation and laser heating. The use of the linear spectral emissivity model provided quite consistent results with the ordinary least squares estimator and with stochastic simulations using the Metropolis-Hastings algorithm. Similarly, the model of independent spectral emissivities resulted in accurate emissivities, but it required an informative prior for temperature, such as the known melting point of the metal sample. Therefore, whatever the inverse method used, a priori in terms of emissivities or temperatures are needed due to the temperature-emissivity correlation and the unknow emissivity behavior.