Institute of Engineering and Natural Sciences

Symmetry-based multi-objective AI/ML driven optimization framework for sustainable building performance

Research on the Symmetry Optimization and Detection for Architecture (SODA) framework in Dutch buildings has emerged as a critical area of inquiry due to the fundamental role symmetry plays in architectural design, heritage conservation, and urban modeling. This research develops and validates an integrated framework combining symmetry-based geometric optimization with Progressive Transfer Learning (PTL) for construction performance enhancement in Dutch buildings. The SODA framework employs computer vision algorithms to automatically identify geometric patterns from diverse data sources including point clouds and facade images. A modified Continuous Symmetry Measure adapted from molecular chemistry provides quantitative symmetry assessment, while multiobjective optimization algorithms balance competing objectives including material efficiency, structural performance, construction economy, and architectural preservation. Progressive Transfer Learning mechanisms enable systematic knowledge transfer across building projects, reducing analysis time for subsequent optimizations while maintaining solution quality. Empirical validation across 50 Dutch buildings encompassing 2.287 million m² demonstrates substantial optimization potential through systematic symmetry-based design interventions, achieving average material savings of 11.3%, construction time reduction of 10.7%, total cost savings of €87.3 million, and carbon emission reduction of 21,450 tons CO₂. The integrated SODA-PTL framework advances architectural computation toward genuine practical impact on construction industry sustainability and efficiency, providing validated design guidelines for practitioners pursuing resource-efficient building optimization.