Digital Pre-Print Verification of Functional FDM Geometry Based on G-Code: Prediction, Profile Selection, and Selective Cad Correction

Introduction. The dimensional accuracy and functional suitability of fused deposition modeling (FDM) components depend on part geometry, technological parameters, and slicing algorithms. Nominal CAD geometry does not fully represent the material distribution defined by extrusion toolpaths, particularly for small functional features and mating interfaces. This study presents Guardian, a G-code-based digital workflow for inspecting local geometry, estimating digital acceptance status, and evaluating profile or CAD changes before printing. Materials and Methods. Extrusion geometry was reconstructed from G-code using a capsule-based representation of deposited paths. Feature-specific criteria were applied to local geometric elements and mating interfaces to determine their digital functional status. Machine learning methods supported suitability prediction and process-profile ranking, while technological adaptation and local CAD correction were evaluated through repeated slicing and geometric verification. The experimental framework included 108 transfer cases, 440 local-feature measurements, 552 mating-interface measurements, independent confirmation series, cross-slicer comparisons, and external STL models. Results. On 36 cases from three previously unseen geometric families, incorporating process parameters increased predictive AUC from 0.807 to 0.960, with F1 = 0.971 and MCC = 0.946. Exhaustive evaluation of eight technological configurations identified a positive digital alternative for all 22 initially unsuitable local features. After all eight configurations had been measured, retrospective per-model selection yielded positive digital status for the monitored features in 20/20 affected models. Digital-status agreement between PrusaSlicer and OrcaSlicer was 19/72 (26.4%) under nominally matched settings, indicating sensitivity to the evaluated slicer domains. The external STL catalog provided 190 valid measurements, while process-profile adaptation recovered 6 of 43 initially unsuitable features, highlighting the influence of surrounding geometry on corrective effectiveness. Conclusions. Guardian integrates G-code-based geometric reconstruction, feature-specific functional assessment, predictive profile selection, and corrective actions with repeated digital verification. The computational results support further evaluation of Guardian as a pre-print screening tool for prototypes, local functional features, and mating interfaces; physical fit was not tested. Further development requires physical metrological calibration, prospective profile selection, and validation across different materials, printers, and slicers.