AI-Driven Quality Assurance Framework for Transportation Infrastructure Using BIM, Digital Twins, and ISO 19650

Abstract

This study has examined how artificial intelligence, when embedded within Building Information Modelling (BIM), digital twins, and the information-management requirements of the ISO 19650 series, affects the effectiveness of quality assurance in transportation infrastructure, and it has developed and tested a framework for AI-driven quality assurance across the delivery and operational phases of highway, bridge, and rail assets. The central problem has been that quality assurance in transportation infrastructure remains largely document-centric, sample-based, and retrospective, so that nonconformances in geometry, materials, workmanship, and asset information are frequently discovered late, after cover-up, handover, or the onset of service, when correction is most costly and least reversible, and that the digital models and sensor streams now routinely produced on such projects are seldom converted into continuous, auditable quality evidence. Guided by five research questions, concerning how AI-driven BIM and digital twin integration affects quality assurance performance, which nonconformance categories are most effectively detected, which AI and sensing techniques are most effective, what risks document-centric manual regimes carry, and how ISO 19650 compliance and common data environment governance improve reliability, the study has proposed a framework comprising five contributing constructs: BIM model maturity and information standardisation; digital twin integration and real-time asset sensing; AI-based defect detection and predictive analytics; ISO 19650 compliance and common data environment governance; and data quality, interoperability and workflow integration, with quality assurance effectiveness as the outcome. A quantitative, cross-sectional design supplemented by a prototype pipeline evaluation has been used, and data have been collected from BIM and information managers, quality assurance and quality control engineers, digital twin and data engineers, asset and maintenance managers, and project directors. Out of 292 distributed questionnaires, 249 valid responses have been retained, producing an 85.3% valid response rate. The analysis has included descriptive statistics, reliability testing, correlation analysis, multiple regression, hypothesis testing, and a prototype evaluation of detection performance, discovery latency, and information-compliance conformance across the asset lifecycle. The findings have shown that all five constructs were significantly and positively associated with quality assurance effectiveness, that the regression model was significant, F(5, 243) = 58.63, p < .001, explaining 54.7% of the variance, and that AI-based defect detection and predictive analytics and BIM model maturity and information standardisation were the strongest predictors. The prototype evaluation showed that document-centric manual regimes discovered the largest share of nonconformances after cover-up or handover, while multimodal AI detection coupled with scan-versus-model geometric checking and rule-based information checking moved discovery decisively upstream.