Hygrothermal performance assessment of wood frame walls under historical and future climates using partial least squares regression

The objective of this study was to develop a prediction model for predicting the mould growth risk in wood-frame walls. A machine learning algorithm, the Partial Least Squares (PLS) regression was used for developing the model. Hygmthermal simulations were performed for a wood-frame wall with brick cladding using hourly historical and projected future climate data for cities located in different climate zones across Canada. The mould index calculated at the exterior layer of OSB was used as the response variable in the PLS model and the most relevant climate parameters were used as inputs including: temperature, relative humidity, wind speed, wind-driven rain, and solar radiation normal to the facade. The model was trained using a training set comprising two climate periods and different wall orientations for three cities. It was then used to predict the mould index for other years and wall orientations without performing the simulations. The coefficient of determination (R-2), the Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) was used to evaluate the model's accuracy. Also, ranking and mould risk category analysis were carried out to assess the model's ability to rank the years according to their moisture severity. Results showed that with proper selection of training dataset, the model can be effectively used to predict the hygrothermal performance of brick-clad wood-frame wall assemblies in the selected cities and over both historical and future climates.