The project aims to develop Digital Wolfram (DW), an embedded hardware- and software-based expert system tailored to TIG-welders for monitoring and optimizing their processes.
- Lead school School of Engineering and Computer Science
- Institute Institut für Drucktechnologie IDT
- Funding organisation Innosuisse
- Duration (planned) 01.02.2023 - 31.01.2025
- Project management Prof. Dr. Annette Kipka
- Head of project Prof. Dr. Annette Kipka
- Project staff Melike Türkes
- Partner Wolfram Industrie GmbH
- Keywords Tungsten Inert-Gas welding, TIG, live monitoring, machine learning, process optimization
Even though Tungsten Inert Gas (TIG) welding is the premium metal joining technology, aspects critical to both the weld quality and the costs such as the timely replacement of the degrading tungsten electrode to date purely rely on empirical knowledge. Despite the broad range of available sensor technologies, in-line data collection and real-time signal processing capabilities available monitoring tools generally lack built-in “intelligence” to, e.g., use live current voltage data to detect bad arc strikes to indicate a worn-out electrode. This deficiency motivated Wolfram Industrie Switzerland who optimizes TIG-welding process for customers in the aerospace, nuclear, and semiconductor industries, to investigate them in their in-house lab on a more fundamental, science-based level. In collaboration with their partners from BFH and ZHAW, they could detect various process irregularities by combining current voltage with optical and acoustic emission analyses.
Course of action
Digital Wolfram will use data- and rule-based machine learning models to analyze live sensor data, detect irregularities and suggest actions to be taken. The employed models will be trained with in-house sensor data under well-controlled conditions in combination with extensive weld quality and electrode aging characterizations performed by ZHAW and BFH. For maximum reliability, Digital Wolfram will provide unique features such as measuring the voltage during the arc ignition phase by proprietary hardware. Three key clients will act as beta testers and early adopters.