Defect Map optimization with the use of Artificial Intelligence for the flat-rolled Aluminium sector
Elval is the aluminium rolling division of ElvalHalcor SA (ELHA:ATH), a leading global aluminium and copper manufacturer with a strong production base across 15 plants, cutting-edge technology and a solid market presence in more than 90 countries. Elval has 50 years of rolling and recycling expertise, designing and manufacturing innovative aluminium solutions while building trusted partnerships across the globe. The extensive custom-made product portfolio serves the packaging, transportation, automotive, HVAC&R, building and construction, energy and industrial markets. Through continuous investments in R&D&I and the in-house Technology Centre, Elval invests in highly specialized human talent, responsible production and environmental protection, enabling the transition to climate neutrality and the contribution of aluminium to a circular economy.
The Challenge
D-cube was challenged by Elval, a leading flat rolled aluminium manufacturer, to further improve the defect classification results of an existing surface inspection system, operating on their premises for more than 15 years. With the use of CYRUS, the Cyber-Physical Platform for Surface Inspection, D-cube worked on an 18-month project to expand defect classification with the use of Artificial Intelligence.
The Company
Elval is the aluminium rolling division of ElvalHalcor SA (ELHA:ATH), a leading global aluminium and copper manufacturer with a strong production base across 15 plants, cutting-edge technology and a solid market presence in more than 90 countries. Elval has 50 years of rolling and recycling expertise, designing and manufacturing innovative aluminium solutions while building trusted partnerships across the globe. The extensive custom-made product portfolio serves the packaging, transportation, automotive, HVAC&R, building and construction, energy and industrial markets. Through continuous investments in R&D&I and the in-house Technology Centre, Elval invests in highly specialized human talent, responsible production and environmental protection, enabling the transition to climate neutrality and the contribution of aluminium to a circular economy.
The Problem / The Opportunity
Existing surface inspection systems on Aluminium Coil process lines operate with defect detection and classification performance which needs the addition of human supervision to meet the high-quality standards of Elval’s manufacturing process. Reducing human involvement in surface defect maps analysis is a pioneering continuous improvement step to digitalization direction, aiming to enhance product quality, improve overall production efficiency and drive customer satisfaction.
The Solution
An innovate and well-trained AI model grid that processes third-party detection and classification results and creates an accurate and reliable surface defects’ map of aluminium coils. Defect maps uploaded to AI Data Management System, enable online system performance comparison.
The Results
Integrating Artificial Intelligence (AI) into surface inspection led to a robust defect detection and classification solution, significantly reducing the statistical uncertainties. AI notably improved both the accuracy and recall metrics of classification.
A 93% accuracy on major defect categories, was achieved with the use of AI, indicating a crucial enhancement of aluminum surface quality assurance.
The Takeaways
This case study highlights key advancements in our methodologies, showcasing substantial enhancements in accuracy, defect detection, operational efficiency, and data transparency, with widespread implications across various categories.
- Significant Improvement on Detection and Classification Accuracy.
- Enhanced Detection and Classification of Critical Defects.
- Operational Efficiency and Quality.
- Data Transparency
Testimonial from Mr. Spilios Theodoropoulos, Senior Project Manager at Elval
At Elval, surface quality is our top priority and a cause for great pride. We consistently make investments in new technologies as well as in the optimization of processes, resources, and equipment in order to maintain this position. The implementation of artificial intelligence in surface inspection merits careful thought and attention, since it is already transforming whole sectors and has the potential to yield unprecedented efficiencies and competitive advantages.
Throughout our 2-year partnership with D-cube, we managed to leverage AI, showcase tangible results and identify endless possibilities for future enhancements. We are now certain that deep learning algorithms, when mastered by the appropriate partners, can achieve an amazingly higher classification efficiency than standard classifiers.
With an average weighted accuracy of 93% and numerous untapped improvements, the potential is more than obvious: AI is not only capable of improving surface defect detection and classification. It is a key-enabling technology towards greater objectives like auto coil release.
Get in touch with us for a brief on your plant or production needs and find out how you can get the most out of our cutting-edge technologies.
