AI Errors – 70% Fewer Artificial Intelligence errors

ai errors

Views: 15

Successful completion of the “Wege in die Forschung” project

Source Leibniz Universität Hannover: Production of individual parts, such as patient-specific implants or molds, is a major challenge as production must be planned individually. At the Leibniz Universität Hannover Institute of Manufacturing and Machine Tool Technology (LUH), Florian Uhlich has developed a 5-axis process planning method for freeform surface processing that uses the experience of previous processes to ensure that the component in the first Try the requirements.

LUH has enabled Uhlich’s project as part of its junior research program “Paths to Research.” Uhlich: “This gave me a chance to work intensively on the method for artificially-designed 5-axis process planning and thus lay the foundation for my dissertation.”

The project uses a parallel simulation to the manufacturing process that generates an independent description of the working condition component as a virtual sensor. If the information thus generated is combined with measured component inspection values, the independent component relationships between quality resources and process variables can be derived through machine learning.

These relationships form the basis for planning new processes. In the project, the expected shape defect prediction due to tool loading for components not yet manufactured was realized. It was possible to achieve very high predictive quality even if the component is designed quite differently. Prediction was also used as the basis for proactive compensation for form error. This avoids mistakes. Compensation allows a form error reduction of more than 70%.

Contact:

For more information, contact Dipl.-Ing. Florian Uhlich, Institute of Production Technology and Machine Tools at Leibniz Universität Hannover at +49 511 762 18252 or uhlich@if.uni-hannover.de.

AI Errors

When should artificial intelligence be a beneficial intelligence?

 

Be the first to comment

Leave a Reply

Your email address will not be published.


*