Hybrid transformer architecture for detecting anomalous z-offset regimes in FFF/FDM 3D printing
Abstract
This study proposes generalized adaptive telemetry encoder for 3D printing (GATE3D), a hybrid framework for early anomaly prediction in fused filament fabrication (FFF)/fused deposition modeling (FDM) additive manufacturing using only native printer telemetry. The framework combines temporal neural representations with gradient-boosted features and was tested on the Caxton dataset using a leak-free print group partition. On an unseen-print holdout comprising 5,323 windows from 27 previously unseen prints, GATE3D achieved Macro-F1 scores of 0.887, 0.883, and 0.905 for flow rate, feed rate, and hotend, respectively, while z-offset remained near chance level (0.503). Paired bootstrap analysis of print groups showed statistically confirmed improvement over HistGradientBoosting (HGB) for the hotend only. Event-level assessment further identified the hotend as the most reliable early warning target, with an event recall rate of 0.687 and 15.714 false positives per 1,000 stable nominal windows. Overall, GATE3D provides a competitive camera-free approach to telemetry-based early anomaly forecasting, while its effectiveness remains strongly target-dependent.
Keywords
Fused deposition modeling; Fused filament fabrication; Machine learning; Multichannel telemetry; Predictive quality control
Full Text:
PDFDOI: https://doi.org/10.11591/eei.v15i6.13123
Refbacks
- There are currently no refbacks.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Bulletin of Electrical Engineering and Informatics (BEEI)
ISSN: 2089-3191
,
e-ISSN: 2302-9285
This journal is published by the
Institute of Advanced Engineering and Science (IAES)
.