BIM and ANN-based rapid prediction approach for natural daylighting inside library spaces.

This article focuses on developing an integrated approach combining Building Information Modeling (BIM) and Artificial Neural Network (ANN) technology for rapid and accurate prediction of daylight factors in large public buildings at the early design stage. Using the Chengdu University of Technology Library as a case study, the authors created a dataset through Autodesk® Revit and Ecotect Analysis simulations, which was then used to train a backpropagation-based ANN model to predict optimal window design parameters—windowsill height, window height, and window width—for maximizing natural daylighting. The trained ANN models demonstrated low root mean squared errors (below 0.1), indicating high prediction accuracy, and enabled the evaluation of over 225,000 parameter combinations much faster than traditional simulation tools. Additionally, a smartphone app was developed to allow designers without modeling experience to visualize and analyze daylight factor predictions, facilitating energy-efficient architectural design decisions in public buildings.