Generative Optimization of Building Blocks for Density, Solar and Structural Performance

Abstract

This study addresses the challenge of performance-informed building blocks generation by developing a generative design framework that simultaneously optimizes building massing, density distribution, and solar and structural performance. As energy consumption, carbon emissions, and material efficiency become increasingly critical in building engineering, there is a growing need for integrated methodologies that combine architectural form exploration with quantifiable performance objectives. The aim of this research is to formulate and validate a modular, cell-based algorithm that generates building configurations optimized for solar gain, thermal comfort, and structural efficiency. The methodology employs parametric design tools, including Grasshopper and Python, alongside simulation engines such as Ladybug for solar radiation analysis and Karamba for finite element structural evaluation. Multi-objective optimization is conducted using the Octopus application to identify Pareto-optimal solutions across competing criteria. The proposed approach is validated using a mid-rise residential block case in Tehran, demonstrating its effectiveness under real-world regulatory and climatic constraints. Findings show significant improvements in seasonal solar performance and reductions in structural deflection, with up to 248% more winter solar gain and 4.6% lower displacement compared to conventional designs. The key contribution of this research lies in its integration of environmental and structural simulation within an automated generative workflow that ensures both design adaptability and engineering feasibility. The novelty of the study is in bridging early-stage form generation with detailed performance feedback, providing a scalable method for sustainable and structurally sound building design. The proposed framework is adaptable to various site contexts and can inform future advances in computational building engineering.