Machine learning-based wind pressure prediction of low-rise non-isolated buildings

This paper proposes a novel machine learning-based wind pressure prediction model (ML-WPP) for low-rise non-isolated buildings. ML-WPP combines a gradient boosting decision tree (GBDT) and the Grid search algorithm (GSA) to automatically predict the wind pressure parameters. In comparison with existing ML models, this model is more straightforward and interpretable. The Tokyo Polytechnic University (TPU) non-isolated low-rise wind tunnel dataset is used to develop and test the ML-WPP model. ML-WPP considers the mean pressure coefficient, fluctuating pressure coefficient, and peak pressure coefficient to reflect the wind pressure among the roof area. The ML-WPP model obtained a low mean-squared error and a high coefficient of determination for all wind tunnel test configurations of the non-isolated low-rise buildings. A time history interpolation is proposed in this paper as well. This technique is the first of its kind as it is the first time an ML model has been used in the wind engineering field to deal with wind pressure prediction while considering the effect of the neighboring buildings. With the advantages of ML-WPP, it is possible to reduce the reliance on physical wind tunnel tests. ML-WPP yields a robust, efficient, accurate alternative to predicting the pressure of structures under wind loads while considering the effects of neighboring structures.