Intelligent Algorithm Empowers Civil Engineering Construction: Dynamic Control Driven by Multi-source Data

Traditional construction control methods face challenges such as data fragmentation, decision-making lag, and rigid response, making it difficult to cope with complex and changing on-site environments. Therefore, this study proposes a dynamic intelligent control method driven by multi-source data. Firstly, it integrates BIM, IoT, and drone data to build a real-time digital twin; Next, the Attention-enhanced Spatial-Temporal Graph Convolutional Network (ASTGCN) is used to dynamically predict the construction process; Then, the NSGA-III (Non-dominated Sorting Genetic Algorithm III) algorithm is used for multi-objective collaborative optimization of project duration, cost, and resource balance; Finally, PPO (Proximal Policy Optimization) reinforcement learning agents are introduced to achieve adaptive regulation. The experiment showed that the prediction accuracy of the ASTGCN model improved by 46.0%, the quality of the solution set obtained by NSGA-III increased by 51.1%, and the PPO agent reduced the critical path impact by about 60% under sudden interference, verifying the significant advantages of the proposed method in improving the resilience, efficiency, and scientificity of construction control.