ABSTRACT The progress of smart manufacturing systems demands the integration of multiple artificial intelligence (AI) paradigms to support the decision-making process in increasingly complex and dynamic environments. Although data-driven approaches excel in pattern discovery and process optimization, they often lack interpretability and the ability to incorporate specialized knowledge. Conversely, symbolic approaches provide transparency and explainability but may present limitations in adapting to unforeseen scenarios. This article presents toward a hybrid semantic framework for manufacturing intelligence that systematically integrates data-driven and symbolic AI approaches through a three-layer hierarchical structure: physical layer, digital layer, and cognitive layer. The framework incorporates an intelligence orchestration layer and a semantic alignment layer to enable semantic communication between heterogeneous components and harmonize knowledge from diverse sources. The experimental evaluation focuses on aerospace sheet metal manufacturing parts, demonstrating the integration between automated feature extraction and specialized knowledge formalization through ontologies. A conversational interface powered by large language models enables natural language interaction with specialized ontologies, providing automated inference capabilities and explainable decision-making. The results validate the feasibility of combining automated characteristic extraction with semantic knowledge representation, establishing a foundation for hybrid AI systems in manufacturing environments. The proposed framework addresses critical challenges in cyber-physical production systems by enabling adaptability, explainability, and real-time decision-making while maintaining semantic consistency between empirical and formal knowledge sources.
Hernandes et al. (Fri,) studied this question.