BMBF funds research project on artificial intelligence in design

The Federal Ministry of Education and Research (BMBF) is funding a project with around 1.8 million euros to explore the fundamentals of digitalized product development. The project, initially planned for three years, is led by Daimler AG; participating with a funding amount of approximately 630,000 euros is the team of Prof. Dr. Klaus-Robert Müller, Professor of Machine Learning at the Technical University of Berlin.

With the increasing digitalization of industry, digital prototypes in Computer-Aided Design (CAD), Computer-Aided Engineering (CAE), or Computer-Aided Manufacturing (CAM) enable the analysis of a new product through simulations. This allows certain tests to be conducted solely on the computer before building a physical prototype. For example, it makes it possible to try out various design alternatives without having to construct a physical model.

The new research project Artificial Intelligence Aided x (AIAx) aims to help manage the digitalization of the product development process with large, complex data sets and to relieve designers from time-consuming routine tasks. The central research question is: How can existing artificial intelligence methods be further developed to support the industry’s traditional strength in developing high-quality technical products? The research project aims that, in the medium term, the terms known in industry today—CAD, CAE, and CAM—will evolve into Artificial Intelligence Aided Design/Engineering/Manufacturing – briefly AIAx.

Experience-based knowledge is valuable in the design process. However, these experiential insights are difficult to formalize and can therefore only be partially passed on to new engineers. By employing certain machine learning methods, patterns in the data of a CAD design could, for example, be recognized and utilized more efficiently.

Each simulation produces vast amounts of data, which currently still need to be analyzed by engineers to identify flaws and deficiencies in the design and to improve them gradually. Specially developed machine learning methods are intended to automatically analyze this data intelligently and also suggest possible improvements. Such automation of the analysis process would not only accelerate the design process but also give engineers more time to develop innovative designs and ideas.

The TU Berlin brings its extensive experience in fundamental research on machine learning methods (ML) and AI into the new AIAx project. The research group led by Klaus-Robert Müller has made significant contributions to Support Vector Machines as well as to the so-called “Explainable Artificial Intelligence” (interpretable AI) and has advanced the establishment of ML methods in quantum chemistry. In this project, we will mainly focus on topics such as “Efficient Deep Learning” and the “Explainability and Robustness” of the ML methods being developed, explains Klaus-Robert Müller.

Making the decision processes of ML methods transparent – i.e., explainable – aims to increase their acceptance in practice and thus enable industrial application. Why does the machine learning process make a particular decision? “Here, we want to investigate different representations of the decision process through a participant study. How should the explanation be presented? What information is meaningful? Regarding the acceptance of ML methods, explainability is an important aspect. Ultimately, the responsibility lies with the designer,” explains Klaus-Robert Müller.

AIAx unites industry partners and various research institutes in a collaborative project. The interdisciplinary nature of the project is reflected in the composition of the consortium. Besides TU Berlin, the project involves Daimler AG, DYNAmore GmbH, Endress+Hauser Maulburg SE & Co. KG, the Karlsruhe Institute of Technology, and USU Software AG.