Revolutionary applications of generative AI in the production environment & quality management

Generative artificial intelligence (GenAI) represents a paradigm shift for industrial value chains. It opens up new opportunities for process optimization, product innovation and resource efficiency. This article examines the use of GenAI in production from a scientific perspective with a particular focus on Bavaria as a high-tech location. The article discusses specific use cases, examines technological and infrastructural requirements and outlines challenges and regulatory implications for responsible implementation.

Industrial production is facing profound transformations in times of digitalization, a shortage of skilled workers and sustainability goals. Generative artificial intelligence - a special class of AI systems based on deep learning architectures such as transformers - promises to address these challenges through intelligent automation and creative problem solving.
As an economically strong industrial location with a high density of automotive, mechanical engineering and electronics companies, coupled with excellent research institutions such as the Technical University of Munich (TUM), the Fraunhofer Group and the AI model location Augsburg, Bavaria offers ideal conditions for a leading role in the industrial use of GenAI.

Maximilian Holland opened the webinar with an introduction to how generative AI works. Using an example - the design of a helicopter with a payload of 800 kg - he demonstrated how modern language models, such as GPT-4, can analyze, structure and solve technical tasks. They combine logical reasoning with the ability to calculate technical parameters and generate visual representations.

KEY TECHNOLOGIES: RAG AND TOOL USE

Two central mechanisms enable the practical use of generative AI:

• Retrieval-Augmented Generation (RAG): This method combines language models with external knowledge databases. Relevant information is extracted from documents and integrated into the answer generation process.
• Tool use and agent systems: AI agents can independently call up tools, query databases or carry out simulations. This enables complex, multi-level problem solutions - for example in process planning or design.

APPLICATION EXAMPLES

The following use cases are presented:

• Automated design: From text description to CAD model generation - AI can design technical geometries and adapt them parametrically.
• Process planning: In an example of the production of fiber composite components, it was shown how AI agents plan production processes, calculate cycle times and select resources.
• Robotics: A household robot was used as an example to demonstrate how AI can react flexibly to language and context in human-robot interaction.

AI-supported quality management in practice

CHALLENGES IN QUALITY ASSURANCE

In the second part of the webinar, Ulrich Kaiser presents Qualiwise GmbH's solution: an AI-supported co-pilot for product quality. The starting point is the so-called PPF process (Production Part Approval Process), which is used for quality assurance in the automotive industry. To date, this process has been highly manual and does not yet scale optimally - particularly in the event of changes or in collaboration with suppliers.

THE SOLUTION APPROACH OF QUALIWISE GmbH

The company pursues an incremental automation approach:

1. Understanding quality documents: AI reads and interprets technical drawings, material certificates and test reports.
2. Linking documents: Information from drawings is compared with inspection plans or material requirements.
3. Automated creation and inspection: The AI creates inspection plans, carries out inspections (also voice-controlled) and automatically generates reports.

A live demonstration during the webinar will show how an inspection plan is created from a technical drawing, an inspection lot is carried out digitally and an inspection report is generated automatically. Material certificates can also be uploaded and automatically checked for conformity. The solution is web-based, intuitive to use and can be integrated into ERP systems such as SAP in the future.

Discussion and outlook

INTEGRATION CHALLENGES

In the subsequent discussion round, questions regarding the integration of AI into existing systems, variant design in CAD and the use of AI without transferring data to external clouds were discussed. Holland emphasizes that on-premise solutions with open source models, such as Mistral or LLaMA, are becoming increasingly practicable - especially for data-sensitive companies.

REGULATORY FRAMEWORK

Despite the opportunities, companies face practical and ethical challenges:

• Data security: processing large, often sensitive production data requires robust data protection concepts in accordance with the GDPR.
• Transparency and explainability: The "black box" nature of many GenAI models requires increased traceability, especially in a safety-critical environment.
• Need for specialists: Implementation requires interdisciplinary knowledge - educational institutions in Bavaria are required to develop new curricula.

Bavaria has a strong technological and political infrastructure with initiatives such as the Center for AI in Production (KI.FABRIK Bayern), the Hightech Agenda Bayern and the AI-HUB@LMU at Ludwig-Maximilians-Universität München. SMEs in particular benefit from low-threshold access to AI consulting services and real-world laboratories. There is potential in areas such as

• Mass customization through AI-controlled production logistics
• Climate-friendly production through resource-efficient process design
• Digital twins that become self-learning with the help of generative models

NEED FOR COORDINATED INITIATIVES

One key conclusion: there is currently a lack of a coordinated European ecosystem for generative AI in engineering and production. A plea was made for Europe's strengths - particularly in the field of engineering software - to be combined with generative AI in a targeted manner instead of simply replicating global language models.

Conclusion

The webinar impressively demonstrates how generative AI can already create concrete added value in industry today - from design and process planning to quality assurance. The solutions presented make it clear that the path from research to practice is not only possible, but already a reality. However, targeted investments, open interfaces and stronger networking between research, industry and politics are needed to exploit the full potential. Especially in combination with Industry 4.0 technologies, Bavaria could not only become a pilot region, but also a European pioneer in AI-supported production.