Data-driven meets knowledge-driven: an LLM‑agent framework for quality control in metal additive manufacturing

Quality prediction in metal additive manufacturing (AM) has conventionally relied on data-driven models that map process parameters to defect classes or quality metrics. However, these models often fail to generalize across different machines, materials, and process regimes. The emerging knowledge-driven approach based on large language models (LLMs) can interpret literature and expert guidance, yet struggles to deliver quantitative decisions tied to part-specific parameter sets. To bridge this gap, we propose AM-Agent, a hybrid neuro-symbolic LLM agent framework that unifies data-driven prediction and domain knowledge through Model Context Protocol tools organized as Knowledge Services (KS) and Data-driven Prediction Services (DPS). KS integrates literature-based static knowledge with a dynamic digital twin (DT) context powered by Asset Administration Shell, enabling real-time queries of printer status and historical build records through LLM-DT interaction. DPS exposes melt-pool regressors and defect classifiers using a model-per-condition design, where the AM-Agent adaptively selects the appropriate pretrained predictor at runtime based on current working conditions. To harmonize the outputs of DPS and KS, a reliability-weighted fusion strategy is proposed to resolve conflicts by dynamically weighting numerical uncertainty against semantic confidence. The proposed AM-Agent is intended as a part-specific, pre-built decision support system for human operators during process planning. It is able to select suitable printers, forecast quality risks, and recommend parameter adjustments. Statistical experiments show that this neuro-symbolic integration significantly outperforms both purely data-driven and LLM-only baselines under domain shifts, validating that hybridizing neural perception with symbolic knowledge provides a promising path toward generalizable and interpretable AM quality control. The full project repository is publicly available on GitHub.