Abstract: Traditional construction quality management is often plagued by heavy reliance on manual intervention, operational inefficiencies, and delayed risk identification. To bridge these gaps, this study develops and validates an intelligent decision support system (IDSS) integrated into a "smart construction site" framework. The system facilitates real-time monitoring, precise diagnosis, and collaborative mitigation of quality risks throughout the construction lifecycle. By leveraging building information modeling (BIM), IoT sensing, big data, and AI, we propose a three-tier architecture encompassing data acquisition, processing, and decision-making. Specifically, convolutional neural networks (CNNs) are employed to automate the detection of quality defects. Experimental results demonstrate that the system achieves a 94.2% accuracy rate in identifying non-compliant behaviors, with a false alarm rate under 3.1%. When deployed in tunnel engineering projects, the system reduced safety incidents by 20%~35% and boosted management efficiency by 22%~35%. These findings suggest that intelligent decision-making significantly accelerates hazard response and minimizes human-induced quality variances, offering a robust technical roadmap and empirical foundation for the digital transformation of construction management.