Artificial Intelligence And Generative Design For Performance-Driven Sustainable Retrofitting Of Existing Institutional Buildings: A Multi-Objective Framework With A Case-Study Validation On A University Academic Building
Abstract
The existing building stock is one of the largest single contributors to global energy use and carbon emissions, yet most retrofit practice remains fragmented: individual measures are sized in isolation, trade-offs between energy, daylight, water and cost are rarely resolved simultaneously, and decisions seldom exploit the parametric and data-driven capabilities now available through building information modelling (BIM). This paper develops and validates a five-layer framework that couples a parametric BIM digital twin with artificial-intelligence (AI) surrogate models and a generative, multi-objective optimization engine to drive performance-driven retrofitting of existing institutional buildings. The framework is demonstrated on a three-story academic building the Department of Architectural Engineering and Design (AED) at the University of Engineering and Technology (UET), Lahore, Pakistan using a BIM-based and standards-referenced retrofit assessment as the empirical baseline. Envelope upgrades reduced wall, roof and window thermal transmittance to within IECC 2021 limits (for example, roof U-value from 0.1019 to 0.0394 BTU/h·ft²·°F); a light-emitting-diode (LED) retrofit lowered lighting demand by about 60%; ultrasonic occupancy sensors saved a further 42%; a roof-constrained rooftop photovoltaic (PV) array of roughly 100.7 kW offset about 44% of the peak electrical load; and combined rainwater harvesting and greywater reuse recovered a rooftop collection potential near 73,200 ft³ and displaced about 53% of potable water used for flushing. These measured potentials are then re-expressed as objectives, variables and constraints for the proposed generative engine, showing how a single integrated search can reconcile competing performance goals that the conventional, measure-by-measure workflow handles only sequentially. The contribution is a transferable, standards-aware decision-support framework that bridges the gap between manual retrofit auditing and automated, generative, performance-based design for institutional buildings in rapidly urbanising, climate-stressed regions.
Keywords: Sustainable retrofitting; building performance; generative design; artificial intelligence; building information modelling (BIM); digital twin; multi-objective optimisation; institutional buildings; energy efficiency; performance-driven architecture
