For situations in which the speed of thermal propagation cannot be considered infinite, a hyperbolic heat conduction equation is typically used to analyze the heat transfer. The conventional hyperbolic heat conduction equation is not consistent with the second law of thermodynamics, in the context of nonequilibrium rational thermodynamics. A modified hyperbolic type heat conduction equation, which is consistent with the second law of thermodynamics, is investigated in this paper. To solve this equation, we introduce a numerical scheme from the field of computational compressible flow. This scheme uses the characteristic properties of a hyperbolic equation and has no oscillation. By solving a model problem, we show that the conventional hyperbolic heat conduction equation can give physically wrong solutions (temperature less than absolute zero) under some conditions. The modified equation does not display these erroneous results. However, the difference between results of these two models is negligible except under extreme conditions.

The Levenberg-Marguardt algorithm is used to study effects on convergence for inverse heat conduction in the unsteady state. In this model, the finite volume method is usedto obtain anestimated temperature, which is necessary for minimizing inverse error. To validate the model, constant thermal conductivity (k) and heat capacity (ρCpC) are identified from a semi-infinite slab subjected to constant heat flux. These properties are inserted into the theoretical equation for a semi-infinite slab, and an analytical solution is obtained by solving the theoretical equation including the two identified properties. The analytical solution and the identified resultare in very good agreement. Three simulations were performed to investigate the sensitivity of computation time and conversion to initial thermo-physical values by changing three different damping ratios of the Levenberg-Marquardt algorithm. Our results show that agood initial guessallowsgood convergence, but convergence time decreases as the value of damping ratio decreases.A poor initial guess results in more convergence time, and causes divergence when a small damping ratio is used. Once the simulation converges, our model shows that results areobtained within an error of 0.01%.