VA4RM
MACHINE VISION FOR REMANUFACTURING OF AUTOMOTIVE PARTS
VA4RM IS AN AUTOMATED RECOVERY PROCESS OF AN AUTOMOTIVE COMPONENT according to its level of wear. The detection of these defects is complex due to multiple existing benchmarks, the area to be inspected, the nature of the component and the intra-class variation. Therefore, it is necessary to create complex smart algorithms that are capable to detect and measure defects quickly. The system has multiple environmental, technological and economic benefi ts which can be applied to different metal products and the automotive sector, that account from 14% to 17% of the manufacturing value in Europe, respectively.
The engineering firm EKIDE is leading VA4RM with the collaboration of VICOMTECH Technology Centre.
DRIVING FACTOR
OBJECTIVES
- Replace manual inspection with automated inspection based on AI allowing unified decision-making criteria, thereby increasing productivity in classification and diagnosis of parts and reducing errors derived from human factors in these processes.
- Improve the company’s competitiveness and production by reducing material waste, driving the transformation of plants towards eco-efficiency.
RESULTS
- Development of detection system that allows maximum highlighting the defect in any product by means of oblique illumination. The developed algorithm is based on model assembly and combines a segmentation algorithm with an object detection algorithm.
- Full validation using a set of components not used in training, obtaining 100% correct classification results.
CONCLUSIONS
- The current validation data guarantee the technical feasibility of VA4RM to solve the problem posed.
- Regarding industrialisation, further steps should involve the creation of a more complete application with a graphical interface and a communication layer with the factory database.
- Training with more samples is necessary to obtain a model that is more robust to changes and to generate a more sophisticated decision layer that can be configured by the corporate client.
ENVIRONMENTAL
TECHNICAL
ECONOMIC
COMMERCIAL
ON THE MARKET
COLLABORATORS
