AI-Infused BIM Systems for Generative Design and Life Cycle Optimization in Modular Construction

This work introduces a Life cycle- conscious generative architecture design system for modular architecture, combining AI-driven layout generation, environmental simulation, optimization, and BIM. With diffusion models and Particle Swarm Optimization (PSO), the system obtains a 0.83 design diversity measure, 96.1% modularity conformance, and 128 feasible designs per run on the ModArchN et benchmark. Life cycle analysis returns the minimum average carbon impact of about 387 kgCO2e/m2 and cost about $1,680/m2 across climatic zones. Convergence of optimization is achieved at 28 iterations with a fitness value of 0.89. BIM integration retains 97.2% metadata and a usability score of 4.6/5, surpassing state-of-art practices in efficiency and applicability.