One of the obstacles in the field of building performance prediction and optimization is how to realize the performance prediction of building form under specific generation rules by means of machine learning model training and how to transfer it to other performance prediction tasks corresponding to other building forms. A complex building form generation method based on multi-dimensional parameter combination is proposed, and the method is applied to the preparation of solar radiation performance prediction data set of building exterior surface. The obtained data set is used for machine learning model training, and the prediction function of solar radiation performance index of building exterior surface is realized. The results show that the model training results are good and the prediction data fit well. Through the prediction of untrained data set and the error comparison experiment, the results show that the data set formed by the method can improve the generalization ability of machine learning models, help to apply the model to different architectural design scenarios, and then provide theoretical support for the realization of general artificial intelligence technology to assist building performance prediction and optimization.
关键词
建筑形态生成/生成式设计/机器学习/建筑性能预测/泛化能力
Key words
building form generation/generative design/machine learning/building performance prediction/generalization ability
万方数据