Sequential estimation of high temperatures of liquid metals by using particle filter methods

Abstract

This work follows up our previous studies dedicated to the estimation of the transient temperature of niobium and 100c6 steel samples, in ranges that include solid and liquid phases, as well as phase changes. In the experimental setup, a metallic sample of few millimeters is aerodynamically levitated and then heated by a powerful laser, while a collimator focused on the sample collects the radiative flux that is split towards five pyrometers of different wavelengths. The measured spectral fluxes are used in an inverse analysis to estimate the sample temperature. Difficulties in the inverse problem solution related to the dependence of the sample emittance with temperature and wavelength could be overcome by reducing the number of model parameters using a simple linear variation of the emittance with respect to the wavelength. In this work the transient temperature of the sample is estimated sequentially by solving a state estimation problem with two algorithms of the particle filter method, namely: the Sampling Importance Resampling (SIR) and Auxiliary Sampling Importance Resampling (ASIR) algorithms. Results obtained with these algorithms and the radiative spectral fluxes exhibited small discrepancies with respect to the temperature measurements of a reference bichromatic pyrometer.