Machine learning for more quality in additive manufacturing
Additive manufacturing offers unprecedented flexibility in the production of complex geometries, but is susceptible to defects during the production process. With the ML-S-LeAF project, we have made a decisive contribution to the development of new possibilities for digital quality assurance.
ML-S-LeAF stands for an innovative approach to quality assurance in additive manufacturing. The project, which is funded by the German Federal Ministry for Economic Affairs and Climate Protection, aims to make manufacturing processes more resource-efficient.
The aim of the research consortium is to develop an automated process monitoring system based on simulated noise emissions during an industrial printing and melting process.
The project
The AI listens
The project focused on the development of artificial intelligence (AI) that acoustically monitors the production process of the powder bed fusion (PBF) method. The aim: to detect defects at an early stage before they affect component quality.
The AI analyzes systematic deviations in the sound profiles - both in the air and in the component itself.
Fig.: Measurement setup using microphones and acceleration sensor in the pressure chamber of the PBF system
The thesis that the printing of defective components can be recognized by finding acoustic deviations had to be confirmed. For this purpose, components with extremely fine lines were printed that push the test system to its limits.
Fig.: Microscopic images of defect-free (left) and defective lines (right).
The AI was then trained using a combination of real measurements from the test device and virtual sound data. A large amount of virtual data was used to offer a wide variety of deviations on which the AI can build a robust evaluation pattern.
Fig.: Comparison between measured and simulated sound data.
Due to the amount of provided training data, error detection with an accuracy of up to 98% was achieved.
Fig.: Error matrix of the AI with binary classification
Project partners
Together with a top-class consortium from industry and science, we worked on pioneering quality assurance methods:
Fraunhofer IDMT
Experts for digital media technologiesSAM and PTW at TU Darmstadt
Cutting-edge research in acoustics and production technologyOmegaLambdaTec GmbH
Specialists for data analysis and machine learningNovicos GmbH
Acoustics experts for simulation and measurement
Thanks to all sponsors
We would like to thank our consortium partners as well as the BMWK (still BMWi at the time of application) and Forschungszentrum Jülich for the funding and support of this successful project.
The project created the basis for a digital, AI-based quality assurance system that was developed specifically for use in lightweight construction. This represents a significant step towards the complete digitalization of quality monitoring in additive manufacturing.
Your contact
I will be happy to answer any questions you may have and provide further details on the implementation and results of ML-S-LeaF.
Dr.-Ing. Ömer Yildiz
Further details
We continuously share the ongoing developments and successes of research projects such as ML-S-LeAF on LinkedIn. Follow us to get the latest updates and learn how machine learning is revolutionizing the manufacturing industry.
Novicos on LinkedIn