An integrated data-driven framework for multi-objective optimization of building forms in early-stage

This study proposes an innovative framework for multi-objective optimization of building form in early-stage architectural design, integrating Artificial Neural Networks (ANN) and the NSGA-II algorithm within the Grasshopper environment. Unlike previous frameworks that often apply sequential optimization to building components, this study introduces a concurrent, multi-scale workflow embedded in a visual design environment, offering near real-time integration of performance and aesthetics. Focusing on an office building in Tehran, the results demonstrate that ANN models can accurately predict key performance indicators, including Energy Use Intensity (EUI), Useful Daylight Illuminance (UDI), photovoltaic energy generation (PV), and solar radiation potential during cold seasons. The models achieved high validation accuracy, with R² values exceeding 0.90 for EUI and cold-season solar potential, 0.78–0.84 for UDI, and above 0.85 for PV. The optimization process revealed that medium densities (∼50 %) offer the best trade-off between EUI and UDI, while higher densities enhance solar energy potential. This framework supports real-time feedback and interactive Pareto exploration, enabling architects to align aesthetic decisions with performance goals. The findings highlight the method’s strong potential for advancing sustainable and efficient architectural design.