AI-Driven Regulation-Aware BIM Framework for Clash Prioritisation and Digital Twin Integration in Saudi Vision 2030 Megaprojects

Abstract

Purpose
This study develops an AI-driven, regulation-aware framework that integrates Building Information Modelling (BIM) and the Saudi Building Code (SBC) to prioritise critical clashes and enable digital twin integration within Vision 2030 megaprojects.

Design/methodology/approach
A hybrid ensemble combining Random Forest (RF), Convolutional Neural Network (CNN), Graph Convolutional Network (GCN), and Graph Attention Network (GAT) was trained using hierarchical graph processing. SBC clauses were encoded into IFC features, with calibrated probabilities and a fixed cost-derived decision threshold.

Findings
Nested leave-one-project-out (LOPO) testing across five industrial federations demonstrated consistent improvements in AUROC, AUPRC, and calibration. The framework reduced coordination time by approximately 65% compared with incumbent workflows, with statistically significant results and large effect sizes.

Originality/value
This paper presents one of the first regulation-aware AI models for BIM clash prioritisation under the Saudi Building Code. The openly released framework enables reproducibility and provides a foundation for real-time digital twins in Saudi Vision 2030 projects.