Towards Multimodal Industrial System Evaluation: Integrating Local Knowledge Bases with Vision-Language Models for Digital Transformation in Shipbuilding

The digital transformation is accelerating the integration of new-generation information technologies with manufacturing, which is key to China’s manufacturing strategy and high-quality development during the 14th Five-Year Plan. Industrial system digital transformation evaluation plays a critical role in optimizing efficiency and advancing transformation. With the rise of large language models (LLMs) and vision-language models (VLMs), AI applications in industry are expanding toward multimodal understanding. However, general-purpose LLMs and VLMs lack industry-specific knowledge, limiting their effectiveness in evaluating digital transformation tasks. To address this, this study proposes digital transformation evaluation indicators for shipbuilding enterprises, integrated into LLMs via a local knowledge base (LKB) and extended with VLMs for scene image interpretation. By combining structured textual knowledge with visual scene descriptions, the framework enhances both accuracy and contextual richness in digital transformation evaluations. Practical results from W Shipbuilding Co. demonstrate that this multimodal approach improves the efficiency, interpretability, and quality of system transformation assessments. This study provides a novel intelligent path for industrial digital transformation evaluation, highlighting the adaptability of LLMs and VLMs for optimized decision-making in real-world industrial scenarios.