INVERSE ESTIMATION OF GRAY-BAND EMISSIVITY IN A THREE-DIMENSIONAL ENCLOSURE USING COMBINED SIMULATED ANNEALING AND REPLATING ALGORITHM Afif Tajouri 1 , Khalil El Khoury 1,2 , and Maroun Nemer 1 1 MINES ParisTech, PSL—Research University, Center for Energy Efficiency of Systems, Palaiseau, France 2 Mechanical Department, Lebanese University, Roumieh, Lebanon A methodology for parametric identification of gray-band emissivity is developed and tested on a radiant oven heating a metal sheet. Direct modeling uses the component interaction network approach that allows fast and accurate simulations. Thermal radiation modeling is done by the replating algorithm, which achieves important savings in the calculation time. The optimization is carried out with the simulated annealing algorithm. Monte Carlo simulations are used for the design of the experiment. Sensitivity analyses are done, concerning the number of temperature sensors and the measurement errors. The results show that the proposed methodology is promising. INTRODUCTION Infrared ovens are widely used in steel thermal processing. Heat transfer mod- eling and simulation in this kind of system plays a major role because it allows pre- diction of thermal behavior and provides valuable help in making decisions for production or pre-design purposes. At the same time, technological advances in this field are now making oven systems more sophisticated, and therefore models become more complicated. Unfortunately, the increase in the number of components and heat exchanges often compromise the precision of simulation results. Consequently, calibration and validation processes become difficult and affect the model reliability. This failure may be due to the accumulation of uncertainties surrounding input para- meters that are associated with thermal properties of materials such as radiative properties, thermal conductivity, and heat capacity. When these properties are not available in the literature, especially for uncommon materials, experimental measure- ments should be done to retrieve them, directly or by inverse methods. The latter involve using measurements for the estimation of unknown parameters that are not directly measured. This is done by the minimization of the residual between simulated and measured quantities [1]. Thus, inverse methods require a reliable Received 20 July 2014; accepted 13 October 2014. Address correspondence to Afif Tajouri, MINES ParisTech, PSL—Research University, Center for Energy Efficiency of Systems, Z.I. Les Glaizes - 5 rue Le ´on Blum 91120, Palaiseau, France. E-mail: afif.tajouri@mines-paristech.fr Numerical Heat Transfer, Part A, 68: 268–287, 2015 Copyright # Taylor & Francis Group, LLC ISSN: 1040-7782 print=1521-0634 online DOI: 10.1080/10407782.2014.977165 268