RECAST - Monitoring welding quality using artificial intelligence
DLR
The use of artificial intelligence (AI) to increase plant productivity is recognised as having great potential. These positive assessments are currently only offset by a few available options for rapid implementation. Ready-made solutions in the field of machine and process technology are only marketed by manufacturers in combination with corresponding systems. Complex AI projects therefore have to be realised as part of expensive individual projects. Overcoming these obstacles represents a major financial hurdle for SMEs, especially if the prospects of success are unclear.
In this project, the project partner WOGRA AG (https://wogra.com/) aims to create an easy-to-use yet adaptable framework for process support. The framework allows existing systems to be upgraded cost-effectively by digitally mapping the production processes. Experienced and technologically adept employees of small and medium-sized enterprises (SMEs) are to be enabled to integrate Kl into existing process chains independently.
To achieve this, production data from components is to be recorded throughout the entire production process with the help of low-cost mobile and embedded devices. The collected data will be used to evaluate individual process steps in terms of their automation potential using AI solutions. If the framework determines on the basis of the collected process data that a process step can be automated (e.g. quality control), this is suggested to the person responsible for the process.
During the project, AI will be used and evaluated in particular for sensor data from the continuous ultrasonic and resistance welding process of thermoplastic composite materials supplied by the DLR. The aim is to provide automated and non-destructive quality assurance of the strength of the weld and to support process optimisation. DLR is analysing the results achieved with regard to statistical significance and classification of the reliability of the AI predictions.
The developed framework is to be made available as open source in order to achieve a wide distribution and thus accelerate the use of AI in production.
Supported by:
