Knowledge-driven preliminary cost estimation of temporary shoring systems for linear construction

Cost estimation for temporary shoring systems in urban deep excavation projects is challenging during the planning and early design stages because of geometric uncertainty, changing construction conditions, and repeated design revisions. Research on temporary shoring systems remains limited because of their high variability and project-specific characteristics. This study proposes a knowledge-driven cost estimation framework for temporary shoring systems in linear construction projects such as roads. The framework formalizes expert design knowledge in a reusable knowledge-driven database through parametric modeling, structural analysis, and itemized quantity takeoff procedures. Based on South Korean design codes and standards, 739 cases were generated by varying excavation geometry, geotechnical conditions, and decking requirements for traffic maintenance. MLR and GBR models were developed and subjected to internal validation using 10-fold cross-validation. Both models achieved high internal validation performance (R 2 > 0.95), and GBR more closely approximated the relationships encoded in the combined dataset. This internal result does not establish superior real-world prediction or a specific advantage for non-decking cases. Preliminary external validation was then conducted using 13 sections from two construction projects that applied the same temporary shoring method. The external validation results were mixed: MLR produced lower prediction errors for the available decking sections, whereas the non-decking sample was too small for a meaningful model comparison. Accordingly, the external validation did not establish the consistent superiority of either model. The principal contribution is demonstrating that expert design logic can be formalized in a reusable knowledge-driven database for early-stage cost estimation. The workflow can be adapted to other regional design standards and construction practices.