Interference wind pressure prediction of high-rise buildings with square section based on machine learning
To predict the wind load of the high-rise building with square section under interference,wind tun-nel interference tests are conducted under 576 working conditions.Three kinds of machine learning methods are used to train,test and verify the prediction model of wind load in the principal building.The prediction re-sults show that decision tree regression(DTR),random forest(RF),and gradient boosting regression tree(GBRT)models can predict the wind load of the principal building effectively,and the prediction perform-ance for the average wind load is better than that for the extreme wind load.The GBRT model has the best performance in predicting wind loads,and the R2 obtained by the model for predicting minimum and average wind loads are 0.994 0 and 0.999 7,respectively.The GBRT model with hyperparameter optimization,whether interpolated or extrapolated,can show good prediction performance.The comparison shows that the prediction performance for the wind pressure distribution is better on the windward side and the two sides,while the prediction effect is relatively weak on the lee side.GBRT model can provide an economical and ef-fective machine learning method for predicting wind loads of high-rise buildings under interference,which can partially replace traditional wind tunnel test and numerical simulation.
high-rise buildingsinterference effectwind pressure coefficientmachine learninggradient boosting regression tree
万方数据
维普
