Integrating experimental, Machine Learning and simulation approach to predict and validate the thermal performance of Ventilated and Permeable tiled Roofs

Abstract

Ventilated Permeable Roofs (VPR) are an effective passive cooling strategy for reducing summer heat gains and cooling energy demand improving indoor thermal comfort. However, their performance is difficult to quantify due to the irregular geometry of tiles and the complexity of the underlying physical processes, which hinder the direct measurement of key aerodynamic parameters such as effective opening area (A) and discharge coefficient (Cd). To address this limitation, the present research activity involves a multi-step methodological approach combining tile experimental characterisation, machine learning, energy simulations and real roofs data. A standardised test method developed within the LIFE SUPERHERO project was used to characterise tile air permeability, while a Gradient Boosting Machine model was trained on reference openings to estimate Cd and A for each configuration. The estimated parameters were implemented in the EnergyPlus Exterior Naturally Ventilated Cavity model and calibrated against measured roof surface temperatures and heat fluxes by tuning the convective heat transfer coefficient and buoyancy height scale. The optimal configuration was then used to simulate the summer performance of the tested tiles. Based on the calibrated simulations, the cooling potential of VPR systems was expressed through an additional conductive thermal resistance, defined for each tile configuration to quantify the reduction of incoming heat flux and external roof surface temperatures induced by air permeability effects. The results demonstrate that higher tile permeability leads to an increase of thermal resistance values, highlighting the effectiveness and applicability of VPR solutions for both building simulation and regulations.