Abstract

The safety and durability of civil infrastructure depend heavily on the timely detection and assessment of structural defects in concrete elements. Traditional inspection methods rely on manual visual assessment, which is often time-consuming, subjective, labor-intensive, and prone to human error. Recent advancements in Artificial Intelligence (AI), Computer Vision, and Large Language Models (LLMs) have enabled the development of intelligent systems capable of automating structural health monitoring tasks. In particular, Multimodal Large Language Models (MLLMs) have emerged as powerful tools capable of processing and reasoning across multiple data modalities, including images, text, and sensor information. This paper presents a Prompt-Driven Multimodal Large Language Model framework for concrete defect identification. The proposed system combines image-based defect detection with prompt engineering techniques to enable contextual understanding and automated interpretation of structural defects such as cracks, spalling, delamination, honeycombing, corrosion staining, and surface deterioration. The framework integrates image acquisition systems, multimodal feature extraction, prompt generation mechanisms, and large language model reasoning to provide accurate defect classification and maintenance recommendations. The study discusses the architecture, implementation methodology, hardware requirements, experimental evaluation, advantages, limitations, and future research directions. Results indicate that prompt-driven multimodal learning significantly improves defect recognition accuracy, interpretability, and decision support capabilities compared to conventional computer vision approaches.