Fraunhofer Institute for Mechanics of Materials IWMFor the first time in Germany, all leading research centers for constitutive materials modeling are being brought together in a consortium. The goal is to create a fully digitized, modular materials modeling platform for digital twin components - ranging from parameter identification and AI-based meta-models to real-time service life assessment. This will give industrial companies access to validated simulation tools that were previously reserved exclusively for scientific research.
Fully digitized & standardized materials modeling for real-time analysis of process and operation-induced deformation and damage behavior within digital twin components
Ongoing research project
Project description
Market needs and challenges
Industrial companies in the mechanical and plant engineering sectors - from the turbine to the pump industry - are under increasing pressure to operate components more reliably, with greater resource efficiency, and for longer periods. Digital component twins are intended to lay the foundation for this. However, high-performance constitutive material models that describe deformation and damage behavior under real operating loads are hardly usable in industrial practice: parameter identification is time-consuming, computation times often run into days, models are not standardized, and robust interfaces are lacking.
The project’s approach
DigitalModelling brings together, for the first time, all leading German research centers for viscoplastic materials modeling in a single consortium. On the MaterialDigital (PMD) platform, modular constitutive material models are made available, linked to ontologies, and converted into real-time-capable FEM subroutines using AI-based meta-models (hybrid neural networks / Gaussian processes). In addition, standardized workflows for parameter identification as well as industrially verified test and benchmarking environments are being developed.
Contribution to addressing challenges
The project enables industrial companies to use advanced material models robustly and efficiently for the first time. More precise component simulations allow for better planning of maintenance intervals, real-time monitoring of limit states, and condition-based operational optimization. The industry partners have set a specific goal of increasing materials utilization by 15% and reducing the economic costs of unforeseen damage by 25% in the medium term.
Fraunhofer IWM’s responsibilities in the project:
Ontology development and knowledge representation:
Fraunhofer IWM develops and expands domain ontologies (specifically PMD_coOntology and BWMD Ontology) for damage and deformation mechanisms. Workflows are created for the semi-automated mapping of process chains and materials characterization results as machine-readable knowledge graphs. Using causal inference, linkage rules between material model modules, microstructural descriptors, and process information are developed. Concrete result: a functional “ontology matching” workflow as a part of the PMD.
Data-driven materials modeling / extension of the DTMF model:
Fraunhofer IWM is expanding its proven DTMF fatigue life model for thermomechanically loaded components using data-driven AI methods. Production and microstructural parameters are integrated as influencing factors concerning fatigue life to significantly reduce the variance in life prediction. Concrete result: a validated, extended data-driven life model, including a workflow for industrial application.
Based on the project results, Fraunhofer IWM can offer industrial companies the following research and development services:
Modular materials model selection and configuration
Support in selecting and compiling suitable constitutive material models from the PMD’s standardized model catalog for specific materials and loading scenarios.
Ontology-based materials data integration
Development of machine-readable knowledge graphs that semantically link process parameters, microstructure, and materials properties as a basis for automated parameter identification and digital twins.
Objective parameter identification for material models
Execution of standardized, quasi-automated workflows to determine materials model parameters based on existing and newly collected experimental data, including uncertainty assessment.
Development of real-time-capable meta-model subroutines
Abstraction of complex constitutive models into hybrid AI-based meta-models (neural networks / Gaussian processes) as FEM subroutines for Abaqus, Ansys, and others to drastically reduce computation times while maintaining accuracy.
Advanced service life prediction accounting for process influences (DTMF+)
Service life evaluation of thermomechanically loaded metal components with explicit consideration of manufacturing route, heat treatment, and microstructure, with the goal of reducing variance and increasing prediction accuracy.
Validation and benchmarking of material models in component geometries
Conducting standardized unit and component test cases on realistic geometries and load sets to verify and compare material models prior to industrial use.
