With AI tools to efficient and sustainable engineering

Whether ChatGPT from OpenAI, the text-to-picture generator DALL-E or the AI assistant Bard from Google - artificial intelligence (AI) tools that generate images and texts based on short instructions, so-called "prompts", are currently receiving widespread attention.

From a company's point of view, the question arises as to how AI can be used in the best possible way in the future. In this context, there are great expectations in the use of AI tools in product development and engineering to make these corporate areas more efficient, but also more ecologically sustainable. But where are the greatest potentials of AI in engineering and what is already possible today?

In July 2023, an insight into current research activities and results on AI tools for use in engineering was provided in the webinar series "From Research to Practice" of the Digital Production & Engineering topic platform of Bayern Innovativ, which this time was conducted in cooperation with the Bavarian AI network baiosphere.

baiosphere - the Bavarian AI network

At the beginning of the webinar, Andreas Preißer introduces the Bavarian AI network baiosphere. With baiosphere, an overview of the fast-growing AI landscape in Bavaria is generated. It represents a kind of "yellow pages" of AI in Bavaria. The Bavarian AI Agency and the AI Council promote research, development and application of AI. The main goal is a responsible use of AI for the benefit of humanity. In addition, use cases are presented and stakeholders are connected via matchmaking, signature events in the AI community. A 13-member team is available for this purpose. Interested parties have the opportunity to find out more at https://baiosphere.org - and can look forward to a relaunch of the website in the fall with further functionalities.

Programming with ChatGPT: How does generative AI change software development tasks?

From e-cars that cannot be charged without functioning software to online shopping that would not be possible without functioning payment and search functions - software is indispensable and ubiquitous. It influences people's daily lives and is a decisive competitive factor for companies. In his talk, Prof. Dr. Albrecht Schmidt, Professor of Computer Science at Ludwig Maximilian University in Munich, takes a look at the importance of software development and how it has changed with the introduction of AI tools such as ChatGPT. He also highlights the challenges and opportunities this presents for companies and developers. In the past, software development was often associated with high staff costs and time-consuming processes. Today, such technologies can significantly increase developer productivity and dramatically reduce the time required for development.

The Revolution in Software Development: GPT-3 and the Future of Coding

One example is GPT-3, which represents a revolution in software development. GPT-3 is a powerful language model from OpenAI that can generate text in creative and precise ways. It enables developers to generate complex codes and solutions in a very short time. From stakeholder and persona analysis to system architecture design and code implementation, GPT-3 can serve as an interactive support. Companies, such as Google or Microsoft, use self-generated software codes to train their AI models. In this way, they increase their productivity in software development by a factor of five. Prof. Schmidt points out that many countries, and Germany in particular, must react quickly here in order not to lose touch with the market.

Prof. Schmidt also makes it clear, however, that the increasing use of GPT-3 and similar technologies also raises ethical questions. How can we ensure that these technologies are used responsibly? It is important to establish clear guidelines and rules to prevent misuse and protect the privacy of users.

Seize AI opportunities now

As he concludes, "The potential of large-scale language models such as GPT-3 for software development is enormous. It enables companies to develop software solutions faster and more efficiently and remain competitive. Nevertheless, there are also challenges, particularly with regard to security and data protection. The future of software development will undoubtedly be shaped by such innovations, and it is up to us to take advantage of these opportunities while overcoming the challenges.

Sensible use of AI tools in engineering - but where and how?

The subsequent presentation by Ms. Tihlarik, research associate at the Chair of Sociology with a focus on Technology - Work - Society at the Friedrich Alexander University Erlangen-Nuremberg, looks at the use of AI from the perspective of the sociology of work. Using examples from the BMBF research project "MoSyS" (human-oriented design of complex systems of systems, funding code 02J19B103), she shows the first concrete starting points for the implementation of AI in engineering as well as for the further procedure. Among other things, the project deals with the integration of AI tools in engineering in order to meet the growing demands on product development. The goal is to design complex systems efficiently while focusing on the needs of people.

Potentials and challenges of AI use in engineering

In qualitative interviews with employees from the engineering sector, their attitudes and understanding of AI were first recorded. This identified potential areas of application for AI support in engineering. These findings laid the foundation for the further development of use cases to optimally adapt AI tools to the needs of employees. However, implementing AI tools in engineering also presented challenges. Cost-benefit tradeoffs were complex, as not all processes benefited from AI and resources had to be allocated for training and customization. In addition, ethical and privacy issues were in focus, especially when dealing with large data sets and decisions made by AI algorithms.

Importance of human-AI collaboration in engineering

The results from the "MoSyS" project show, according to Ms. Tihlarik, that standardized and digital routine activities in engineering are ideal application areas for AI tools. Here, AI systems can increase efficiency and take over repetitive tasks, while workers can focus on complex and other challenges, such as creative ones. During implementation, however, it is essential to utilize the expertise of employees and to include their needs in the design process depending on the respective work process.

This is the only way to successfully respond to the growing complexity in product development.

Resource efficiency in development and production with method and engineering AI

In the third presentation, Peter Stirnweiß, Lead Engineer and Six Sigma Black Belt at mts Consulting & Engineering GmbH, gives an insight into how resource efficiency in development and production can be achieved with the help of an AI system. For this purpose, the company uses the Analyser® tool, which it developed in-house. This was awarded the Bavarian Resource Efficiency Prize in 2021.

At the beginning of a project, the first step is to collect data and analyze the processes in order to define the important quality characteristics. To do this, measurable influencing factors are identified and the processes are digitized to collect and evaluate the data in a database. Labeling helps to clearly assign the process steps and parts. AI is then used to analyze the data and create predictive models. This makes it possible to understand the relationships between influencing factors and quality targets and to predict quality. The optimal process parameters are derived as a so-called "best setting" to ensure high quality and resource efficiency.

The practical benefits of AI

Mr. Stirnweiß now shows some practical examples of how the AI systems used achieve resource efficiency. A customer in the automotive industry wanted to change the substrate material for dashboards and trim parts to an environmentally friendly and less expensive plastic. Predictive modeling and optimization resulted in resource savings of about 60% by reducing scrap and rework while maintaining quality appearance. In another example from mechanics, the goal was to make the drivetrain and transmission in electric vehicles quieter. With 3200 influencing variables and only 10 parts available, it was possible to achieve consistently low noise levels across all speed ranges. This resulted in resource savings of up to 30% through reduced scrap and rework. The third example comes from the aerospace industry: For an expensive component with 830 process parameters, quality prediction was to be made possible at an early stage. Transparent cause-effect relationships and the ability to regulate the process enabled a significant increase in resource efficiency to be achieved here as well.

All in all, these examples show how the targeted use of AI in development and production can lead to a significant increase in resource efficiency. The data-driven approaches enable sustainable optimization of processes and products, which is not only economically beneficial but also makes a positive contribution to environmental protection.

Final discussion

The intensity of the participants' engagement with the topic of AI was also evident in the final discussion round. Here, the main issue was the intellectual property of content generated by an AI tool. And how companies should behave today, even in the face of a largely unclear legal situation.

Prof. Schmidt points out, among other things, that when generating codes, it is important to use classic search engines to check whether a code already exists or whether something may have been copied unintentionally, similar to the way texts are generated. It must be ensured that no intellectual property is infringed.

His basic recommendation and that of the other presenters: The creators of the future are those who create AI models, and there is some risk in waiting for legal certainty. Those with the necessary resources should consider training their own models. Otherwise, there is the option of using free open-source models or even commercial models. It is important, he said, to take a first step and use AI models that ensure more efficient processes - an important point for competitiveness.

The webinar series "From Research to Practice"

In the event format of the Digital Production & Engineering theme platform of Bayern Innovativ, research institutions and companies provide insights into current research activities and discuss them with participants. The aim is to support small and medium-sized enterprises in particular in making meaningful use of digital technologies in their production processes and engineering.

Contact details of speakers

Andreas Preißer, Head of Business Relations Bavarian AI Agency
Contact by Mail

Prof. Dr. Albrecht Schmidt, Professor of Computer Science, Ludwig-Maximilians-Universität München
Contact by Mail

Amelie Tihlarik, Research Associate, Chair of Sociology with a focus on Technology - Work - Society, Nuremberg Campus of Technology, Friedrich-Alexander-Universität Erlangen-Nürnberg
Contact by Mail

Peter Stirnweiß, Lead Engineer and Six Sigma Black Belt, mts Consulting & Engineering GmbH
Contact by Mail