Laser welding inspection systems are frequently used in industrial manufacturing. Most techniques are based on thermal radiation that is emitted during the process. The need for parameterization of these systems to each new welding task causes additional work, often it even leads to an increase in false detections. If it would be possible to make such systems self-learning defects could be detected with only little additional effort. The aim is to create a system with assisted self-learning capabilities to detect close to 100 % of defects and to reduce the false detection rate below 10 %.
First, a tailor-made sensor package by the assessment partner Cicrosa is integrated with hardware by the assessment partner Empiric. The next step sees the testing of a first ICT-prototype for its ability to differentiate between faulty and non-faulty seams during the laser welding process. Cartif’s traceability system will be adapted to document all steps in the laser welding process and to record the data that is necessary to support the self-learning approach. Advanced information processing of this data will enable the discovery of defects that have not been considered during the supervised learning phase. At the premises of Cicrosa, the system will undergo a validation in a production like environment to demonstrate its capability.
The CICERONE system will help to reduce scrap rate and speed up production ramp-up time. It is expected that the system can be transferred to several other welding applications with minimum effort.
Equipment for detecting defects in laser welding of steel has been developed accomplishing most goals set. The system features a small camera and easily configurable analysis software implementing a self-learning module. The system can be transferred to several other welding applications with minimum effort.