Synthesis of RTD models VIA stochastic procedures: simulated annealing and genetic algorithm

The general strategy adopted in the development of a computational tool performing structural and parametric identifications of models based on the residence time distribution (RTD) theory is exposed. After a brief presentation of the global methodology, the main aspect of the procedure constituted by the master problem performing the structural identification is detailed. This problem of discrimination of the various schemes is solved by two stochastic procedures based on Simulated Annealing (SA) algorithm or on Genetic Algorithm (GA). The implementation of the last algorithm is exposed. An industrial approach with experimental data is proposed for the qualification step and for the description and quantification of air flows patterns and the associated pollutant transfers observed in ventilated indoor spaces, and particularly in nuclear ventilation networks.