Sensor supported model-based parametrization of 3D-printing processes
Engineering Design, Machine Elements, Product Development
Final Report Abstract
In the research project SmoPa3D, methods for process-integrated quality control in Fused Deposition Modeling (FDM) were developed. The aim was to dynamically regulate printing parameters to meet defined quality objectives using laser line sensor technology (LLS) and data-driven modeling. The project was divided into two phases. In the first phase, possibilities for data modelling and data reduction were investigated to enable efficient data processing. Based on these models, quality metrics such as gaps, distances, or surface roughness can be derived by comparing target and actual data. Where simplified two-dimensional models facilitate categorization of errors and allow inferences about the printing state. The modeling of quality target variables can be realized using both white-box approaches, such as geometrical filament extrusion models, or black-box approaches, such as neural networks. These allow for a prediction of the printing state based on selected process parameters and form the basis for process control when combined with extended data foundations. In the second phase, a new FDM printer was equipped with modifiable G-code along with two laser line sensors and a Jetson to ensure real-time capabilities. An ontology for a cloudbased database was established to accurately assign measurement data to process parameters and clearly categorize the generated data. In addition to the real data generated, a pipeline for generating synthetic data was implemented to increase the data basis for training machine learning methods. For defect prediction, the PointNet++ framework was employed, achieving a test accuracy of 95.4 %. Furthermore, error propagation across consecutive layers could be successfully modeled. The real-time system and the developed machine learning models offer promising approaches for process control. However, a complete predictive control model could not be achieved within the project period due to the high complexity in database development and data synthetization. Future work will be built upon the developed methods and open data structure.
Publications
-
Qualitätsorientierte modellbasierte Prozessparameter für das Fused Deposition Modeling. Apprimus Verlag, Aachen. 2018
Fuhrmann, M.
-
Data-driven Prediction of Surface Quality in Fused Deposition Modeling using Machine Learning. Production at the leading edge of technology, 473-481. Springer Berlin Heidelberg.
Sohnius, Felix; Schlegel, Peter; Ellerich, Max & Schmitt, Robert H.
-
Influence of single filament dimensions on geometrical density as a quality criterion for fused filament fabrication APAM 2019
Schmitt, R. H., Geiger, K., Wolfschläger, D. & Peterek, M.
-
In-line optical measurement system for 3D process monitoring of a fused filament fabrication (FFF) printer. Holistic Innovation in Additive Manufacturing Conference June 25-26 2020
Wolfschläger, D., Briele, K., Ellerich, M. & Schmitt, R. H.
-
Voxel-based description model of quality-related data for a holistic quality assurance in additive manufacturing Proceeding of the Joint Special Interest Group meeting between euspen and ASPE Advancing Precision in Additive Manufacturing September 2023
Großeheide, J., Wolfschläger, D. & Schmitt, R. H.
