Multiphysics simulation: Key to extreme events on structures
The REACT Focus Group advances understanding of infrastructure behavior under extreme hydrological events, enhancing design and mitigation strategies. We developed and validated 3-D multiphysics models simulating debris/mud flows and floods impacting infrastructure, optimized for high-performance computing. We also created surrogate models to reduce complexity while maintaining accuracy.
Focus Group: Digital Twins of Civil Structures and Protection Systems in a Climate Change Perspective (REACT)
Prof. Antonia Larese (University of Padova), Alumna Hans Fischer Fellow
Andi Makarim Katili (TUM), Doctoral Candidate
Hosts: Prof. Kai-Uwe Bletzinger, Prof. Roland Wüchner (TUM)
Deep learning for imaging and signal processing
In mountainous regions and flood-prone areas worldwide, communities face growing threats from debris flows and floods – rapid, destructive phenomena that can cause devastating damage. As climate change increases the frequency of extreme weather events, understanding and mitigating these hazards becomes crucial. The REACT project has made significant strides in developing innovative tools to protect our infrastructure, focusing particularly on buildings, bridges, and protective structures such as safety nets in mountainous zones.
Understanding the challenge
Structures in high-risk areas play a vital role in protecting communities. However, these structures face immense forces during extreme events like debris flows and floods. Traditionally, they have been designed considering standard static loads and dynamic pressures, but this approach has significant limitations. There is a lack of detailed understanding of the dynamics preceding the impact, which is crucial for accurately defining the dynamic load on the structure itself, and the fluid-structure interaction is simply neglected.
Our innovative approach
The REACT project tackled this challenge by developing cutting-edge computer simulations that can accurately model the behavior of debris flows and floods and their impact on structures. We approached the problem by exploiting a multidisciplinary approach. We created models that integrate various disciplines: Solid mechanics, fluid dynamics, and geomechanics integrated by cutting-edge knowledge of advanced modeling and simulations.
These models consider materials subject to large deformations and displacements, representing realistic impact scenarios.
High-fidelity simulations of the physical complexity: A key ingredient for a future predictive tool
Using state-of-the-art numerical techniques and High-Performance Computing (HPC), we have developed simulations that reproduce the real physics of the problems. This includes:
Detailed modeling of debris, mud and water flow [1][2][3] (Fig. 1.) We developed 3-dimensional free surface models to capture flow dynamics at a local scale (on the order of hundreds of square meters). Various numerical techniques (e.g., finite elements, particle-based methods) were tested, and different constitutive material models were employed to simulate the flow dynamics as a heterogeneous continuum. We solved the Navier-Stokes equations for Newtonian and non-Newtonian flows, as well as the solid mechanics equations in a large deformation regime to simulate granular flows. This latter approach is essential for representing complex constitutive laws used in geomechanics to simulate various granular materials, ranging from classic visco-plastic models to elasto-plastic laws that account for historical and field variables as well as for multiphase materials.
Simulation of the interaction between the flow and structures [4] (Fig. 2,3). We have developed various partitioned approaches, coupling several techniques to enable multiphysics simulations.
Analysis of deformations and potential structural failures (Fig. 2) with advanced finite elements models accounting for the structural response.
Given the computational complexity of high-fidelity models, we worked to create surrogate models. These simplified models maintain a high degree of accuracy while drastically improving computational efficiency, allowing for faster and larger-scale analyses. [5]
Unlike methods based solely on measured data, our approach allows for simulating scenarios never before recorded. This is crucial for creating a truly predictive tool, capable of anticipating unprecedented extreme events.
V. Singer, K.B. Sautter, A. Larese, R. Wüchner, and K.-U. Bletzinger, “A partitioned material
point method and discrete element method coupling scheme”, Advanced Modeling and Simulation in Engineering Sciences, vol. 9,
no. 16, 2022.
Key findings and implications
Our advanced simulations have provided several crucial insights:
1) Improved risk assessment: We can now more accurately predict the forces that debris flows and floods exert on structures in specific locations.
2) Optimized design: By simulating various structural configurations, we can identify designs that are more effective at dissipating impact energy and protecting infrastructure.
3) Understanding weak points: Our models aid in maintenance and reinforcement planning.
4) Early warning potential: Our large-scale simulations could contribute to improved early warning systems for communities downstream of potential hazard areas.
The tools and knowledge developed in the REACT project have immediate practical applications:
Engineers can design more resilient structures, potentially saving lives and reducing infrastructure damage.
Local authorities can better prioritize upgrades or replacements where needed.
The multidisciplinary approach we developed can be adapted to study other types of natural hazards.
Conclusions and future perspectives
While the REACT project has made significant advances, there is still more work to be done. Future research directions include:
1) Incorporating a more diverse range of materials and flow types into our models to represent an even wider range of real-world scenarios.
2) Exploring how climate change might alter debris flow and flood patterns and intensities, allowing for long-term infrastructure planning.
3) Integrating our simulations with real-time monitoring systems to create more responsive early warning capabilities.
The REACT project represents a significant leap forward in our ability to understand and mitigate the dangers of extreme events in vulnerable environments. By combining advanced physics-based simulations with practical engineering knowledge, we have created tools that can help protect communities and infrastructure.
As climate change continues to alter weather patterns and increase the frequency of extreme events, the importance of this work only grows.
Through continued refinement and application of these models and methods, exploiting the power of AI technology, we can work toward a future where vulnerable communities are more resilient in the face of natural hazards, preserving both lives and livelihoods.
In close collaboration with Dr. Nicolò Crescenzio (Università di Padova) and Dr. Veronika Singer (TUM).
[1]
V. Singer, T. Teschemacher, A. Larese, R. Wüchner, and K.-U. Bletzinger (2023).
[2]
L. Moreno, R. Wuechner, and A. Larese (2024).
[3]
V. Singer, “Partitioned Coupling Strategies to Simulate Granular Mass Flows Impacting Flexible Protective Structures”, Ph.D. dissertation, Chair of Structural Analysis, School of Engineering and Design, TUM, Munich, Germany, 2024. [Online]. Available: https://mediatum.ub.tum.de/doc/1743069/1743069.pdf
[4]
V. Singer, K. B. Sautter, A. Larese, R. Wüchner, and K.-U. Bletzinger (2022).
[5]
D. Pasetto, D. Kumar, E. Spricigo, M. Putti, and A. Larese, “An RBF Approach for Enhanced Surrogate Modeling of a Debris Flow,” In Proc. EGU General Assembly Conference 27 April/May 2025, Vienna, Austria.
Selected publications
- V. Singer, T. Teschemacher, A. Larese, R. Wüchner, and K.-U. Bletzinger, “Lagrange multiplier imposition of non-conforming essential boundary conditions in implicit material point method,” Computational Mechanics, vol. 73, no. 6, pp. 1311–1333, Nov. 2023, doi: 10.1007/s00466-023-02412-w.
- V. Singer, K. B. Sautter, A. Larese, R. Wüchner, and K.-U. Bletzinger, “A partitioned material point method and discrete element method coupling scheme,” Advanced Modeling and Simulation in Engineering Sciences, vol. 9, no. 1, Aug. 2022, doi: 10.1186/s40323-022-00229-5.
- L. Moreno, R. Wuechner, and A. Larese, “A mixed stabilized MPM formulation for incompressible hyperelastic materials using Variational Subgrid-Scales,” Computer Methods in Applied Mechanics and Engineering, vol. 435, p. 117621, Dec. 2024, doi: 10.1016/j.cma.2024.117621.
- D. Dehghan-Souraki, D. López-Gómez, E. Bladé-Castellet, A. Larese, and M. Sanz-Ramos, “Optimizing sediment transport models by using the Monte Carlo simulation and deep neural network (DNN): A case study of the Riba-Roja reservoir,” Environmental Modelling & Software, vol. 175, p. 105979, Feb. 2024, doi: 10.1016/j.envsoft.2024.105979.
- D. Dehghan-Souraki, U. C. Goñi, R. Z. Martínez, E. B. I. Castellet, and A. Larese, “Three-Dimensional finite element modeling of thermal stratification in the Riba-Roja Reservoir confluence: a Fluid–Thermal Multiphysics approach,” Water, vol. 17, no. 5, p. 674, Feb. 2025, doi: 10.3390/w17050674.