Predicting wind-driven rain catch ratios in building simulation using machine learning techniques

Wind-driven rain catch-ratios are an important boundary condition for the study of the hygrothermal behaviour and durability of building envelopes. Measurements are time-consuming, expensive and of limited applicability to other facades of other buildings and sites. CFD simulations are accurate, but time consuming and simplified calculation have large uncertainty. This work focuses on improving the use of WDR catch-ratios in building simulation using artificial neural networks (ANNs). Results obtained indicate that an ANN can predict WDR catch-ratio with an uncertainty of 0:07 for a confidence interval of 95%. ANNs have the ability to combine results from multiple experiments/simulations to provide catch ratios at any position at the facade and extrapolate them to a range of facade's aspect ratios.