Knowledge Graphs and Neurosymbolic AI for Manufacturing Intelligence: Semantic Reasoning, Ontology-Driven Analytics, and the Integration of Symbolic and Neural Intelligence

Abstract

The integration of artificial intelligence into manufacturing has generated powerful data-driven models — deep reinforcement learning policies for process control, convolutional networks for quality inspection, recurrent architectures for demand forecasting — yet these models operate as black boxes divorced from the rich domain knowledge that manufacturing engineers have accumulated over decades. Knowledge graphs — structured representations of entities (machines, products, materials, processes) and their relationships — and neurosymbolic AI — hybrid architectures that combine the pattern recognition power of neural networks with the logical reasoning capabilities of symbolic AI — have emerged as complementary paradigms that inject domain knowledge into AI systems, enabling reasoning, explanation, and causal understanding that purely data-driven approaches cannot provide. This review provides a comprehensive synthesis of knowledge graphs and neurosymbolic AI for manufacturing intelligence, examining knowledge graph construction and ontology engineering for manufacturing, graph neural networks for manufacturing analytics, neurosymbolic reasoning for fault diagnosis and process optimization, the integration of knowledge graphs with digital twin platforms, and the synthesis of knowledge-driven AI with the four preceding AI frameworks for smart manufacturing (Physics-Informed RL-MPC, Adaptive Manipulation, Quality Intelligence Architecture, and Neuromorphic Industrial Intelligence Architecture). We further connect these advances to industrial optical sensing technologies — precision 3D surface metrology and four-dimensional thermal imaging — demonstrating how structured knowledge and neural-symbolic integration enhance perceptual intelligence in manufacturing. A central contribution is the articulation of an integrated Knowledge-Augmented Manufacturing Intelligence (KAMI) framework that unifies knowledge graphs, neurosymbolic reasoning, and deep learning for trustworthy, explainable, and knowledge-driven manufacturing AI.