A KG-LLM-Based Product2inspection Mapping Approach for Module Prefabrication

Quality inspection in the production of modular construction (MC) is paramount for the success of MC projects (e.g., geometry errors are sensitive for installation). However, automatic and accurate inspection task generation is still challenging in planning and control, particularly in mass production. To this end, this study aims to develop a Knowledge Graph-Large Language Model (KG-LLM) based product2inspection mapping approach for MC production. First, a knowledge graph of inspection tasks, MC products, inspection tools, inspection frequencies, operators, and corresponding production tasks is established to enhance LLM with domain knowledge. Then, a prompt-based product2inspection mapping approach using KG-LLM is developed. Finally, the effectiveness of product2inspection is tested on an MC case project within controlled experiments. This study can improve quality inspection collaboration in module prefabrication while exploring the potential of applying Large Language Models in the construction domain.