Thermophysical parameter estimation of multi-layer walls with stochastic optimization methods

The problem of the estimation of thermal conductivities of multi-layer walls is studied using genetic algorithms and simulated annealing. Parameter estimation is shown for the cases of three- and five- layer walls, comparing the two stochastic approaches and also contrasting them with deterministic gradient-based methods. It is shown how stochastic methods permit better performances than classical ones when initial estimations of parameter values are not available or when the problem becomes complex.