Abstract

Masked stereolithography (MSLA) is widely used for fabricating high-resolution polymer parts, but its reliability is limited by layer delamination, insufficient curing and support failure, which are difficult to observe because the process occurs beneath an opaque resin surface. This paper presents a non-invasive, force-based monitoring framework for real-time quality assessment in commercial MSLA printers, conceived as an IoT-enabled sensing layer for Industry 4.0 additive manufacturing cells. A strain-gauge module was integrated into the Z-axis cantilever of a commercial MSLA printer to measure peeling forces during each lift-retract cycle, and the resulting signals were synchronised with firmware layer information via SDCP (Smart Device Control Protocol V3.0.0). Employing a structured experimental campaign with intentionally induced failure modes produced an extensive labelled dataset of layer-wise force waveforms. Two Random Forest classifiers were trained to perform binary classification of healthy versus faulty layers. A geometry-assisted model, using a hybrid feature set that combines geometric descriptors from the sliced file with statistics and peeling-phase descriptors of the force waveform, achieved 96.65% accuracy on unseen print jobs with satisfactory precision for faulty layers and an extremely low false positive rate. A second, sensoronly model that operated solely on normalised raw force samples, without access to print metadata, still attained 94.04% overall accuracy, although recall on faulty layers was lower (66%) due to class imbalance and subtler defect signatures. Feature-importance analysis revealed physically meaningful spatial effects and confirmed the dominant role of peeling-peak topology. The results demonstrate that layer-wise peeling-force signatures provide a robust basis for in-situ MSLA monitoring and that a standalone, plug-and-play sensing system is feasible without deep integration into the printer software stack, providing a building block for digital-twin and simulation-driven process optimisation in emerging MSLA production technologies.