The store floor as a place of production manufacturing with all its processes has long been part of the MVI Group’s portfolio. Now the MVI site in Wolfsburg is combining this expertise with the topic of artificial intelligence (AI). Customer demand in this area is increasing. Sven Gabrysch, Business Solutions, gives us an insight into the topic of shopfloor AI at MVI in an interview.

“Hardly any other area is currently influenced as much by the ongoing technological leaps as production,” says our expert, describing developments on the market, ”robotics and automated guided vehicles (AGVs) play a key role in today’s production lines.” As experts in the field of production and logistics, the MVI Group offers everything from the planning of the entire factory to the complete implementation of production and logistics processes. In this environment, the technology of our time has naturally also established itself, which is currently the focus of attention in many areas of application: artificial intelligence.

How can AI make production processes more efficient, Sven?
The core idea of using artificial intelligence (AI), especially in the field of visual recognition (computer vision), directly in the production environment (“store floor”), is to automate and improve tasks such as

• automatic quality control (e.g. detection of defects on components, spot welds)
• object recognition (e.g. for logistics control, presence control of parts)
• monitoring various production steps (primary forming, forming, joining, coating, assembly)
• Increasing occupational safety (detection of hazardous situations, monitoring of safety zones)

And what advantages does holistic solution expertise offer when it comes to AI integration?
The innovation of our approach to establishing AI in production lies in specifically addressing key challenges in the implementation of AI solutions in industry:

1. use of synthetic data: A central point of innovation is the generation and use of synthetic, photorealistic image data generated from CAx data (CAD). This overcomes the limitation that real image data is often not available in sufficient quantity, variety or for rare defect cases, or is very cost-intensive to generate. The MVI can thus specifically simulate defects and variations (e.g. exposure, surface structure, position) in order to effectively train AI models.
2. explainable AI: To increase trust in the “black box” of AI, we rely on explainable AI methods. This makes the decisions of the AI models comprehensible and transparent.
3. holistic solution expertise: MVI is not limited to software and AI, but integrates domain knowledge from consulting, IT, automation and design. They offer a “one-stop shop” service, which leads to more functional overall systems.
4. industry-capable processes: MVI follows a structured process (incl. MLOps – Machine Learning Operations) to develop AI solutions that not only work technically, but can also be operated economically, predictably, scalably and reliably in an industrial environment.

What significant added value does the approach offer our customers in concrete terms?
Companies benefit from our approach in many ways:

• Reduced need for real data: Dependence on real-world training data, which is often difficult to obtain, is drastically reduced.
• Increased system robustness and accuracy: Training with a high variance of synthetic data increases the robustness of AI systems and improves forecasting accuracy.
• Full transparency and trust: Explainable AI creates traceability and therefore trust in the automated decisions.
• Cost-effectiveness and scalability: The MVI process enables predictable development times and costs as well as scalable and reliable AI solutions for industrial use.
• Holistic and integrated solutions: Customers receive not just an AI model, but a complete, integrated solution including consulting, connection to existing systems (IT, automation) and support, which simplifies implementation and ensures effectiveness.
• Broad fields of application: The approach can be flexibly applied to many areas of production and logistics, from defect detection in injection molding production to logistics control through container recognition.

Our colleague from Tech Advisory Sven Gabrysch summarizes:
“In the world of production and plant planning, an exclusive focus on software and artificial intelligence often leads to challenges if sufficient domain knowledge is not included. Without a deep understanding of the specific production processes and technical requirements, AI solutions can be developed that are only partially functional in practical implementation. This results in inefficient processes and increased costs, as the systems are not fully aligned with the real conditions of production facilities. Our approach overcomes this challenge by integrating our extensive domain knowledge from over 50 years of experience in production into the development of our AI systems. This deep grounding in real-world production processes enables us to develop customized, intelligent solutions that are not only technologically advanced, but also seamlessly integrated into existing processes and highly effective. By merging proven production know-how with cutting-edge AI technology, we offer our customers optimized solutions that create real value and pave the way for future innovations.”

Thank you Sven!