Value Proposition:
- Efficient modeling of dynamic phenomena with negligible error rates below 1%
- Reduced computational demand, not dependent on large datasets
- Utility across different domains and scalable to multiscale challenges
- Applicable to life sciences, material science, and engineering sectors
Unmet Need:
Accurate modeling of material behavior is essential for assessing the performance, durability, and safety of both engineered systems and biological tissues under various physical and environmental conditions. Finite element analysis (FEA) remains the industry standard for simulating such behavior by discretizing complex geometries into smaller elements to capture local and global responses. However, traditional FEA techniques face significant limitations: they often rely on simplifying assumptions, require repeated simulations for varying parameters, and incur high computational costs-particularly when fine-resolution meshes are needed for multiscale phenomena. Accordingly, there exists a pressing need for a surrogate modeling framework that combines high fidelity with computational efficiency, supports multiscale resolution, and can robustly simulate dynamic, time-evolving phenomena across diverse domains, including life sciences and structural engineering.
Technology Description:
Johns Hopkins researchers have developed a cutting-edge AI-driven simulation framework that transforms how we model complex materials and structures. By combining advanced deep learning with traditional physics-based solvers, this hybrid technology delivers rapid, high-accuracy predictions of material behavior under both static and dynamic conditions-without the need for large datasets or extensive reprogramming.
Unlike conventional tools that are slow, resource-intensive, and limited to static analysis, this innovative platform supports mesh-free, multiscale modeling that dynamically adapts as physical conditions evolve. It intelligently focuses computational resources where they're most needed-such as areas with high stress or complex geometry-resulting in faster simulations and significant cost savings.
Built for seamless integration with existing engineering software, this solution enhances performance, scalability, and accessibility across industries from biomedical to aerospace. Whether predicting the response of soft tissue under impact or optimizing the performance of engineered materials, this platform empowers users with the speed of AI and the rigor of physics.
Stage of Development:
Researchers have developed computational physics and simulation software to analyze static and dynamic mechanical phenomena that can be integrated with existing technologies.
Data Availability:
Data available upon request.
Publications:
Wanga, Wei, Maryam Hakimzadeh, Haihui Ruan, and Somdatta Goswami. "Accelerating Multiscale Modeling with Hybrid Solvers: Coupling FEM and Neural Operators with Domain Decomposition." arXiv preprint arXiv:2504.11383 (2025).
See also https://github.com/Centrum-IntelliPhysics.
