Real-Time Estimation of Some Thermodynamics Properties During a Microwave Heating Process

This work considers the prediction in real time of physicochemical parameters of a sample heated in a uniform electromagnetic field. The thermal conductivity (K) <i style='mso-bidi-font-style:normal'> ( K ) and the combination of density and heat capacity terms (pc)  were estimated as a demonstrative example. The sample (with known geometry) was subjected to electromagnetic radiation, generating a uniform and time constant volumetric heat flow within it. Real temperature profile was simulated adding white Gaussian noise to the original data, obtained from the theoretical model. For solving the objective function, simulated annealing and genetic algorithms, along with the traditional Levenberg-Marquardt method were used for comparative purposes. Results show similar findings of all algorithms for three simulation scenarios, as long as the signal to noise ratio sits at least at 30 dB. It means for practical purposes, that the estimation procedure presented here requires both, a good experimental design and an electronic instrumentation correctly specified. If both requirements are satisfied simultaneously, it is possible to estimate these type of parameters on-line, without need for an additional experimental setup. This work considers the prediction in real time of physicochemical parameters of a sample heated in a uniform electromagnetic field. The thermal conductivity <i style='mso-bidi-font-style:normal'> ( K ) and the combination of density and heat capacity terms ( <i style='mso-bidi-font-style:normal'><span lang=EN style='font-family:"Cambria Math","serif"; mso-ansi-language:EN'> ρc ) were estimated as a demonstrative example. The sample (with known geometry) was subjected to electromagnetic radiation, generating a uniform and time constant volumetric heat flow within it. Real temperature profile was simulated adding white Gaussian noise to the original data, obtained from the theoretical model. For solving the objective function, simulated annealing and genetic algorithms, along with the traditional Levenberg-Marquardt method were used for comparative purposes. Results show similar findings of all algorithms for three simulation scenarios, as long as the signal to noise ratio sits at least at 30 dB. It means for practical purposes, that the estimation procedure presented here requires both, a good experimental design and an electronic instrumentation correctly specified. If both requirements are satisfied simultaneously, it is possible to estimate these type of parameters on-line, without need for an additional experimental setup.