ViKIMon completed: AI quality inspection for assembly that shows how certain it is

Quality inspection in assembly is becoming more demanding as the number of product variants grows. AI can help. However, many small and medium-sized companies are hesitant: How reliable is an AI decision, and what happens if it is wrong? The EU AI Act also requires AI systems to be accurate, robust and transparent, and to allow human oversight.

In the ViKIMon project (“Trustworthy AI-based Quality Inspection in Assembly”), the RIF Institut für Forschung und Transfer e.V. and the Department of Quality and Process Management at the University of Kassel have developed a solution. The requirements are based on interviews with industry experts and a survey of companies.

At the core of the project is a demonstrator on an assembly system where people and technology work together:

• A camera checks whether components are assembled correctly.
• An AI evaluates the course of each screwing process.
• The AI reports not only “OK” or “not OK”, but also how certain it is. If it is uncertain, the operator decides.
• A central server collects all results, monitors inspection quality and documents every decision.
• New AI models only go into operation once a person has approved them.

Operators, quality management and developers each see exactly the information they need for their tasks.

Next, the University of Kassel will test the system in experiments. The project runs until the end of February 2027. The results will reach industry through the project advisory committee, publications and events. Companies interested in the demonstrator are welcome to get in touch.

ViKIMon is funded under the Industrial Collective Research programme (IGF).

Contact: Dr.-Ing. Jan Niklas Büscher