Ontology-enabled neuro-symbolic closed-loop control of laser powder bed fusion

A geometry-conditioned, neuro-symbolic closed-loop architecture is proposed for laser powder bed fusion, in which an ontology aligned with established standards and domain vocabularies operates inside the control loop and couples symbolic reasoning with statistical learning to set the targets of a constraint-aware predictive controller. The ontology links the process objectives and constraints to the signals a controller can observe. A description-logic reasoner selects the active constraints, which are converted into references and bounds enforced on each scan. The demonstrated case is overhang dross: a quality limit on the melt pool depth, which governs quality but is not measured in the present formulation, is mapped through a geometry- and power-dependent depth-to-width ratio onto a bound on the observable width. The ratio and its empirical uncertainty are estimated by a neural ensemble. The reasoner classifies each upcoming feature and selects the active constraints, adding a lack-of-fusion floor at overhangs and an extrapolation guard when the overhang angle falls below the calibrated range or uncertainty is high. It runs only when the symbolic context changes, and the inferred classification remains cached between changes. In an Eagar–Tsai surrogate calibrated to the NIST AM-Bench benchmark for IN625, the architecture reduced the mean total defect count at 45° from 6.0 to 6.2 tracks for the geometry-conditioned baselines to 3.2 per build and avoided their shift from dross to lack of fusion. Ablation attributed the ontology contribution to explicit representation of constraint applicability, inference, and consistency checking rather than separate numerical gains. The results establish architectural feasibility. Machine-specific calibration of the depth-to-width relation and physical validation are the principal next steps.