Recurrent neural networks and neuro-inverse approach for estimating the thermal diffusivity of building materials

According to both the actual energy context and the latest laws meeting EU requirements about energy certification schemes for buildings, carrying out an energy performance diagnosis is mandatory, notably when buying or selling buildings. Indeed, invisible defaults could have a detrimental effect on their insulating qualities. An in-situ estimation of thermo-physical properties allowing to locate defaults, the present work focuses on testing in simulation, as a first approach, a new and effective method based on the use of artificial neural networks to characterize building materials i.e. to estimate their thermal diffusivity using thermograms obtained thanks to a non-destructive method.