Design and implementation of an industrial software cloud platform integrating knowledge graphs and large language models

Industrial software is evolving from standalone tools toward collaborative services, increasing the need for heterogeneous resource coordination and industrial knowledge reuse. Integrating knowledge-based planning with shop-floor execution remains a challenge in complex manufacturing workflows. To address this challenge, this study presents the design and engineering implementation of an industrial software cloud platform integrating knowledge graphs and large language models (LLMs). Deployed in a private cloud, the platform comprises resource, enabling, driving, and application layers within a cloud-fog-edge-endpoint framework. A knowledge-model-data-resource operational loop connects data governance, knowledge retrieval, task planning, verification, execution, and feedback. Knowledge graphs provide domain evidence and constraints, LLMs generate candidate plans, and mechanistic models and deterministic software services support verification and execution. Agents coordinate service invocation and feedback across these stages. Two engineering applications demonstrate the platform’s implementation in five-axis blisk computer-aided manufacturing (CAM) and CNC machine-tool prognostics and health management (PHM). Existing CAM records show reductions of 40.27% in aggregate programming and preparation time and 12.29% in single-piece machining time. The PHM application demonstrates the functional integration of data acquisition, fault reasoning, maintenance planning, and knowledge updating. These cases demonstrate the platform’s engineering implementation in manufacturing and maintenance workflows. The current evaluation is based on case-specific CAM records and PHM functional demonstrations, with broader performance and cross-enterprise validation remaining outstanding.