A direct and simultaneous estimation method of the main three dimensional thermal diffusivity tensor (" , $ , %) of orthotropic opaque materials, is presented in this paper. This method consists in coupling a non-intrusive and unique 3D flash experiment with a transient nonlinear inverse heat transfer technique. A short and non-uniform excitation is applied on the surface of an orthotropic material using a CO2 laser, while the front face temperature cartography is measured over time by an IR camera. The inverse problem developed in the present study is based on the minimization of the least-squares criterion between the outputs of a 3D thermal quadrupoles model, and the experimental measurements. In order to properly estimate the thermal diffusivities, parameters related to the thermal excitation, in terms of shape and intensity, should be also estimated. In addition to that increase in the number of unknown parameters, the discontinuity nature of the excitation justify the choice of an analytical model. Considering the large number of parameters to estimate, as well as the non-linear nature of the problem, a hybrid optimization algorithm combining a stochastic method and deterministic one, is applied. The identification method proposed in this work, named as DSEH (Direct and Simultaneous Estimation using Harmonics), is validated using an isotropic opaque material of known properties. Finally, the method is used on an orthotropic carbon fiber composite, commonly used in industries, thanks to its thermal and mechanical characteristics. The results are compared to other methods and shown to be in a good agreement with the literature values. The parameters identification is then completed by a sensitivity analysis, and evaluated in terms of robustness, accuracy, and time consumption.

This work investigates the potential of the particle swarm algorithm for the optimization of detailed kinetic mechanisms. To that end, empirical analysis has been conducted to evaluate the efficiency of this algorithm in comparison with the genetic algorithm. Both algorithms are built on evolutionary processes according to which a randomly defined population will evolve, over the iterations, towards an optimal solution. The genetic algorithm is driven by crossover and mutation operators and by a selection method. The PSO approach is based on the experience of each individual and on the group experience to control the direction of its evolution. The success of the application of an algorithm can be sensitive to the choice of operators and the relative importance attributed to them. Therefore, to make the comparison as rigorous as possible, about a dozen strategies were proposed for each algorithm and the performances were evaluated. A degraded version of the GRI-Mech 3.0 mechanism (i.e. containing some of the kinetic constants randomly modified) was generated and then optimized by the two evolutionary algorithms to recover the predictive character of the original mechanism. The results show that, for the majority of the proposed strategies, PSO is more efficient than the GA, whereas the latter is generally much more used for the optimization of detailed kinetic mechanisms.

No abstract