A few weeks ago, the Mek team attended the final wrap-up meeting of the EXPLAIN (EXPLanatory interactive Artificial intelligence for INdustry) project, held in Lübbenau at the offices of LEAG, Germany’s second-largest energy producer. This meeting marked the conclusion of a three-year collaboration among 15 partners, including research institutions and industry leaders, all working together to explore how Explainable AI can transform industrial applications.
Mek has successfully concluded its participation in this EU-funded project, which focused on the potential of Human-Centered Explainable AI (HCXAI) in process industries, providing valuable insights into how AI can be made more transparent, reliable, and user-friendly for operators in industrial settings.
A Step Forward in Vision-Based AI for Manufacturing
Over the past three years, Mek focused on investigating Explainable AI (XAI) for Image Analysis and Vision, a critical component in electronics manufacturing. In industries like PCB manufacturing, ensuring product quality through inspection is essential. At the End-of-Line stage, operators use both human judgment and AI tools to determine whether products meet quality standards. While AI helps flag potential defects, the challenge lies in ensuring operators understand why these decisions are made.
Explainable AI bridges this gap by providing clear, actionable insights into which inspection criteria were not met, what part of the product caused the failure, and where the issue is visible in the camera image. By improving transparency in these systems, the project aims to enhance trust and decision-making during the inspection process.
Key Insights from the Final Wrap-Up Meeting in Lubbenau
From a vision perspective, the project generated crucial insights into how image-based AI systems can become more intuitive for operators:
- Improved Transparency: Operators now have access to clear explanations of AI model outputs, with visual feedback such as heatmaps and reference-based explanations. These tools pinpoint which parts of a product are flagged as defects and why.
- Customizability and Usability: One of the key challenges for industrial applications is creating AI systems that adapt to varying user expertise levels. The project emphasized the need to provide explanations at different levels of detail, depending on the user’s role and experience.
- Continuous Improvement: The project highlighted the importance of ongoing feedback from operators to improve both models and explanations. Tracking user interactions ensures models stay accurate, and operators remain confident in the system.
- Tailored Industry Solutions: The project underscored the need for XAI solutions that align with specific industry requirements. In electronics manufacturing, AI models must integrate seamlessly with existing workflows to ensure effectiveness.
The Future of Explainable AI in Vision Systems
With the EXPLAIN project now complete, Mek is ready to continue advancing Explainable AI technologies in image analysis and vision systems. The lessons learned through this collaboration will inform the next generation of AI-powered inspection systems, making them not only more accurate but also more understandable and user-friendly.
As AI becomes a bigger part of manufacturing, trust and transparency are more important than ever. The EU AI Act, which sets requirements for explainability in high-risk AI applications, emphasizes the need for transparency. The project’s focus on Human-Centered Explainable AI ensures that operators can work alongside AI in a way that’s intuitive, safe, and efficient.
The EU AI Act highlights this by requiring explainability in high-risk AI systems. That’s why the project placed a strong focus on Human-Centered Explainable AI.
Looking ahead, Mek will continue building AI systems that meet the real-world needs of industrial users, using new technologies to improve how people interact with these tools.
Conclusion
Being part of the EXPLAIN project has given Mek valuable insight into where AI is headed in process industries, especially in the area of vision systems. As the project concludes, Mek is excited to put these learnings into action—continuing to develop transparent, effective AI solutions tailored to the needs of the electronics manufacturing industry.