Research Seminar on AI: Unsupervised Tool Condition Monitoring in Fine Blanking

  Logo AI-Center and IoP

Contact

Phone

work
+49 241 80 22958

Email

E-Mail

Tuesday, May 31, 2022, 4:00pm

Tool wear during fine blanking impairs the quality of the sheared part. Fine blanking companies report low planning reliability, since tool wear is e.g. highly dependent on material conditions, which are largely unknown to the worker. Acoustic emission (AE) data has proven to contain valuable information about the tool wear state in past research efforts. Tool wear effects and the resulting changes in the tribological system cause changes in the AE signal. However, the analysis of industrial acoustic emission data is challenging due to the sheer volume of the data and the sparsity of meaningful labels in an industrial context. For his PhD thesis, Martin Unterberg extracts, preprocesses and analyzes several hundred thousand fine blanking process cycles from an industrial fine blanking line and utilize unsupervised representation learning to identify, track and interpret changes in the AE signal. Through sparse labels from a log file written by the worker at the industrial fine blanking line and validation experiments at WZL in a controlled setting, he wants to identify and interpret links between changes in the AE signal and changes of the underlying tribological system. 

Martin Unterberg originally studied mathematics in cologne and initially worked as a math teacher. After picking up a teaching assignment at the faculty for mathematics, computer science and natural sciences of RWTH Aachen university in 2017 he decided to pursue a new career and started his PhD in 2018 at the laboratory of machine tools and production engineering (WZL) of RWTH Aachen University. After working within the IoP cluster of excellence he is currently the project lead for the project SPAICER (funded by the Federal Ministry for Economic Affairs and Climate Action) at WZL. Specifically, he works with companies from the sheet-metal forming sector and use machine learning to extract knowledge about the tool wear state from sensory data in a sheet-metal forming process called fineblanking. His research interest lies mainly in the application of unsupervised representation learning in an industrial context. 

 
 

Further information

To stay up to date with information regarding the AI seminar, you can subscribe to its mailing list by sending an e-mail with the subject "subscribe" to ai-seminar-join@lists.rwth-aachen.de