Traditional supply chain diagnostic methods that rely on manual analysis and expert experience are no longer able to meet the needs of modern supply chain management. The relationship between human decision-making and system operation in the supply chain has also shifted from one-way dependence to intelligent collaboration. Therefore, this paper proposes a supply chain intelligent diagnosis framework driven by Large Language Model (LLM), which integrates LLM, optimization solver, multi-agent collaboration framework, and supply chain fault mode library to achieve an automated analysis model of supply chain data. LLM technology is used to intelligently transform natural language into optimized code, thereby achieving automated generation from problem description to optimization solutions, simplifying the complex modeling process in supply chain management, and enabling the framework to have the ability of self-learning and iterative optimization.