As the complexity of complex product system design continues to increase, Model-Based Systems Engineering (MBSE) has demonstrated significant advantages in the process of system modeling and simulation. In recent years, Generative AI (GenAI) has introduced an innovative modeling paradigm to MBSE. However, the scarcity of high-quality, structured simulation model datasets remains a significant bottleneck, hindering further advancements of large models in this field. This paper constructs a system simulation model dataset aimed at industrial complex system modeling and simulation. The dataset, based on X language, encompasses multi-level modeling information such as system requirements, use cases, architecture, and physical behaviors, boasting excellent structural consistency and executability. We have designed a comprehensive data generation process, including domain task setting, LLM-driven model generation, syntax verification, structural normalization, and manual review. Experimental results indicate that the dataset demonstrates high performance in terms of syntactic accuracy and structural completeness, and can be successfully applied to fine-tuning tasks for large language models. This dataset provides high-quality training material for intelligent system modeling tasks and offers a reliable baseline support for subsequent research.