Artificial Intelligence for Material Models.

Optimization of Material Modeling by means of Machine Learning

Logo of the project "Artificial Intelligence for Material Models" (AIMM)
Ihr Ansprechpartner (Kopie)

Dr.-Ing. Said Jamei

Mercedes-Benz Group AG


Artificial Intelligence for Material Models

AIMM proposes new methods for innovative product development by creating machine-learning-based material models, enabling the resource-efficient use of new materials.

Material Optimization in Product Development

Nowadays, digital configuration methods are the central tools in product development. For this purpose, realistic simulation models already exist for process simulation (e.g. injection molding simulation) and structure simulation (e.g. NVH and crash simulation), which precisely map physical phenomena and thus enable virtual tests. Using topology optimization, it is now possible to find material-efficient lightweight solutions in a highly automated process.

The latest algorithms in commercial simulation programs are also capable of optimally selecting and using different materials. The simulations are based on material models that capture specific material properties such as plasticity, anisotropy and damage. However, the creation as well as conditioning of classical material models is currently a tedious and time-consuming process.

Two factors in particular show that material models need to be made more efficient: On the one hand, the experimental determination of the parameters to be considered is time- and cost-intensive, and on the other hand, the simulation processes require high computing power. However, if the models were simplified or their complexity reduced, the prediction quality of the simulation could suffer.

So how can the processes of material modeling and ultimately the entire product development be made more innovative and efficient?

Development of a Machine Learning-based Method for Material Modeling

The objective of the AIMM project is to supplement or replace classical material modeling with an alternative data-driven and thus digital material modeling. By establishing machine learning (ML) methods in simulation-based vehicle development, complex modeling processes are to be simplified, leading to a reduction in development times and enabling digital functional validation (by a larger group of users). In addition, computation times can be reduced, as iterative solutions of material equations can potentially be avoided by using ML models.

Thus, computations can be faster, they require fewer hardware resources, and multiple variants can be considered in the same amount of time. Against the background of the use of new and complex materials, the limitations of conventional material modeling should be overcome and a fast material description for new materials should be possible.

Accelerating the Introduction of Innovative Materials through Interdisciplinary Collaboration

The interdisciplinary collaboration in ARENA2036 is revolutionizing the use of machine learning (ML) in the field of material modeling. Here, among others, the disciplines of artificial intelligence, structural mechanics, numerical mathematics, test execution as well as those of metrology come together to insert data-driven models into the process chain of product development.

The data-driven modeling process developed in the project enables significantly faster and more cost-effective introduction of new materials. The complete automation of data generation and evaluation is an enabler for the full utilization of the method's potential. Complex modeling processes, which currently require a lot of time, technical and human resources, are simplified and the digital predictability of material behavior is improved. This will advance both the development and the use of innovative materials and material combinations.

AIMM Project Objectives at a Glance

  • Development of more accurate and efficient material models by complementing or substituting classical material models with ML-based models
  • Development of new specific experimental concepts for the efficient generation of necessary training data
  • Process automation to shorten the characterization and modeling phase
  • Accelerated description and rapid deployment of novel materials

Partners and Funding of AIMM


  • Large-scale enterprises: Mercedes-Benz AG, ElringKlinger AG
  • SME: DYNAmore GmbH
  • Research institutes: TU Berlin-IDA, FhG EMI, Uni Stuttgart-IFB und IFU
  • Associated partners: GOM GmbH, Renumics GmbH

AIMM receives funding from the German Federal Ministry of Economics and Climate Protection (formerly BMWi) under the funding code 19|20024A.