Fraunhofer Institute for Mechanics of Materials IWMThe steel industry is under enormous pressure to drastically reduce CO₂ emissions while simultaneously bringing high-performance materials to market faster. AID4GREENEST is developing six innovative AI-powered tools that replace the conventional trial-and-error approach in steel development and manufacturing with digital predictive models—from microstructure characterization in under a minute instead of over an hour to creep life prediction without thousand-hour tests. For companies across the entire steel value chain, this opens up opportunities for cost and time savings while simultaneously reducing material scrap. The AI models developed and an open data platform lay the foundation for model-based innovation that can be transferred to other energy-intensive sectors.
AI-powered characterization and modeling for green steel technology
Completed research project
Project description
Market Needs and Challenges
The steel industry is responsible for 20–25% of industrial CO₂ emissions and is under immense pressure to transform. At the same time, strategic sectors such as the automotive industry, wind energy, gas turbines, and electric transportation require increasingly high-performance steel grades, the development of which is still based on time-consuming, material-wasting “trial-and-error” methods. Scrap rates of up to 10% in production, creep tests lasting thousands of hours, and time-consuming EBSD analyses (>1 hour per measurement) significantly slow down the market introduction of new steel grades. Without fundamental digitalization and acceleration of steel development, Europe risks falling behind in global competition.
Project Approach
AID4GREENEST is developing six AI-based characterization and modeling tools that cover the entire steel innovation cycle: (1) a model-based rapid characterization method that extracts EBSD information directly from SEM images in less than one minute; (2–3) AI-based screening tools for bidirectional process↔microstructure prediction; (4) a sequential model for microstructure prediction during forging and quenching of meter-sized components; (5) an ML-based tool for creep life prediction; (6) an accelerated creep testing methodology that reduces test times by three orders of magnitude. In addition, an open, AI-supported data platform will be established in accordance with FAIR and EMMO/CHADA/MODA principles to ensure interoperability and knowledge transfer.
Contribution to Addressing the Challenges
AID4GREENEST enables a reduction in experimental test series by nearly 50%, completely eliminates the rework of defective large forgings, and replaces energy-intensive long-term creep tests. This drastically reduces CO₂ emissions, material scrap, and energy consumption in steel development. The tools accelerate the introduction of low-carbon steel grades and enable the reduction or elimination of critical raw materials in alloys.
Fraunhofer IWM’s work packages in the project:
Probabilistic ML model for reliably predicting the creep life of heat-resistant steels—with quantified uncertainty
Fraunhofer IWM is developing a probabilistic machine learning model that predicts the creep life of heat-resistant steels in a data-efficient manner and with explicitly stated prediction uncertainty. For the first time, the model integrates experimental creep test data and multiscale simulation data into a common learning framework—thereby delivering more robust life-span predictions than approaches calibrated purely experimentally or purely based on simulations. In addition, a validated, accelerated creep test dataset is being created using the stepped isostress method, which serves both as a high-quality training dataset for the ML model and as an independent validation basis for constitutive material models.
Microstructure Model and Cross-Scale Simulation for the Industrial Forging Process of Large Components
Fraunhofer IWM provides a fully calibrated mean-field model for recrystallization and grain growth that is integrated into the industrial FE process chain for forging meter-sized components and, for the first time, enables a reliable prediction of the resulting microstructure across all relevant component regions. This is based on a material-specific experimental dataset from hot-forming tests conducted at process-relevant temperatures and forming speeds, which provides a robust, directly applicable basis for the material description.
Collection and Standardization of Data and Workflows
Fraunhofer IWM coordinates the collection, harmonization, and standardization of the data and workflows generated in the project and develops common documentation standards for project-wide data exchange.
Based on the project results, Fraunhofer IWM can offer industrial companies the following research and development services:
1. ML-based creep life prediction for heat-resistant steels
Fraunhofer IWM provides a validated, probabilistic machine learning model for predicting the creep life of heat-resistant steels—with explicitly stated prediction uncertainty and without the need for standard tests lasting thousands of hours.
2. Accelerated creep testing using the Stepped Isostress Method (SSM) with model-based evaluation
Fraunhofer IWM offers a fully validated methodology for accelerated creep testing that delivers results in just a few days—results that would take up to 10,000 hours using conventional testing methods.
3. Probabilistic multi-fidelity modeling for data enrichment and efficient model development
Fraunhofer IWM offers active-learning-based workflows that intelligently combine experimental and simulation-based data, thereby enabling robust predictive models even with limited datasets.
4. Simulation of microstructure development during free-form forging of large components
Fraunhofer IWM provides a validated simulation methodology that can be integrated into the industrial FE process chain and reliably predicts grain size distribution and the degree of recrystallization across the entire cross-section of meter-sized forged components.
5. Calibration of thermomechanical material models through targeted hot forming tests
Fraunhofer IWM experimentally determines all material properties required for robust process simulation under process-relevant conditions—thereby creating the indispensable data foundation for reliable in-house simulations.
6. Consulting and implementation of AI methods in industrial materials and process development
Fraunhofer IWM supports companies in the practical development of their own ML pipelines for material predictions—from the structured database to the validated, operationally implemented model.
7. Remaining Life Assessment and Damage Prediction Under Realistic Operating Conditions
Based on validated, advanced creep damage models, Fraunhofer IWM performs remaining service life assessments for high-temperature components under realistic, variable temperature and load histories—providing a robust foundation for maintenance planning and service life extension.
