ABSTRACT. Dynamic load testing (DLT) is a widely used field testing method for evaluating pile load capacity. The Genetic Algorithm (GA) is potential technique that can be used for DLT analysis while eliminating the influence of initial guesses and manual intervention based on engineering judgement and experience. However, GA requires extensive function evaluations, resulting in high computational costs. Surrogate-Assisted Genetic Algorithm (SAGA), by constructing surrogate models, enables the analysis program to more efficiently explore potential solutions and thus improves the efficiency of GA. In this study, SAGA was integrated into a DLT analysis program and validated. The results demonstrate that, under identical optimization settings, the use of surrogate model improves the accuracy of the analysis. The results of the parametric study are also presented to illustrate how different surrogate model parameters affect the accuracy of the optimization outcomes. Furthermore, a case study based on field data confirms that the SAGA DLT analysis program holds considerable promise for practical engineering applications.

ABSTRACT. With the increasing severity of population aging and labor shortages, the construction industry is facing significant challenges, including insufficient on-site workforce, limited management capacity, and difficulties in transferring technical expertise. To address these issues, this study proposes an automated and visualized on-site construction progress control framework that integrates AI-based image recognition and mixed reality (MR) technologies. The proposed framework encompasses key technical processes such as physical–virtual model alignment, real-time Image Recognition, automated data updating, and visualized progress presentation, providing a practical and intelligent solution for construction management.