SmoPa3D – Sensor-supported model-based parameterisation of 3D printing processes

Key Info

Basic Information

Duration:
01.11.2021 to 31.10.2023
Organizational Unit:
Chair of Production Metrology and Quality Management, Model-based Systems, Quality Intelligence
Funding:
German Research Foundation DFG
Status:
Running

Contact

Phone

work
+49 241 80 25466

Email

E-Mail

Contact

Name

Hanna Brings

Wissenschaftliche Angestellte

Phone

work
+49 160 90158534

Email

E-Mail
 

Additive manufacturing processes are characterized by their flexibility and possibilities for individual production. However, they are not sufficiently resilient for industrial production due to their fluctuating product quality and the necessary process knowledge.

The aim of the research project is to increase process automation and achieve reproducibly high component quality by implementing real-time control. The research project SmoPa3D approaches this with a process-integrated measuring system and a model-predictive control. In the first part of the research project, a measuring system consisting of laser light section sensors was installed in a 3D printer, which records the individual component layers with a resolution of 50 µm. Deviations in the filament geometry can be determined by comparison with a target model. The deviations of a series of tests were then used as a basis to apply machine learning methods for the prediction of quality parameters.

Based on these findings, process control is to be developed and implemented in the second funding period. On the one hand, the existing system will be improved: According to the previous proof of concept, the program codes will be adapted to such an extent that data processing can take place in real-time between the printing of two layers. On the other hand, the machine learning algorithms must be further developed so that the geometric deviations are not only detected but categorized according to quality and type. Based on this data and the printer's control parameters, the quality parameters of the following layers are predicted. As a last step, a process control is implemented which uses the knowledge of the predicted evolution of an error category to the final component quality for dynamic compensation of the machine code or the control parameters.