Our research aims to advance data-enabled, high-fidelity simulation frameworks that integrate physics-based modeling, machine learning, and data assimilation to enable predictive modeling and inverse analysis of complex materials and geosystems under extreme conditions, including damage and failure. Our current research focuses on the following core themes:
- Intelligent Scientific Computing: We develop novel computational methodologies and infrastructure for advancing Software 2.0 that supports high-performance modeling and simulation with integrated data-driven and learning-based capabilities
- Multiscale Mechanics: We investigate multiscale and multiphysics processes in heterogenous porous media and composite, including civil, geological, and biological soft materials, through integrating model reduction, multiscale modeling, and thermodynamics-constrained machine learning.
- Digital Twin Technologies for Natural and Engineered System Resilience: We advance scientific machine learning and reduced-order modeling frameworks for multiscale and multiphysics forward and inverse simulations, with applications to material failure, natural hazards, Earth systems, and the resilient design of materials and structures.
Research Topics:
- Differentiable & Machine Learning Enhanced Computational Mechanics
- Digital Twins for Solids and Structures (e.g., biomaterials, inelastic materials)
- Multifunctional Materials Design Across Scales
- Multiphysics Processes in Porous and Fractured Media (Subsurface Systems)
- Natural Hazards and Earth System Modeling
Data-driven engineering and mechanics
Develop novel computational methods under an optimization framework that integrates scientific computing, data-driven machine learning, and particle-based methods to the modeling of soft tissues, fracture mechanics, and nonlinear solids.
References: He, X., He, Q. & Chen, J.S., Comput Methods Appl Mech Eng (2021); He, X., He, Q. & Chen, J.S., Data-Centric Engineering (2020); He, Q. et al., Journal of Biomechanics (2020); He, Q. & Chen, J.S., Comput Methods Appl Mech Eng (2019)
Physics-informed/Scientific machine learning
Develop effective and robust deep learning methods based on physics-informed neural networks to solve multiphysics problems arising in engineering science.
References: Du et al. Computers and Geotechnics (2023); He, Q. & Tartakovsky, A., Water Resources Research (2021); Tartakovsky, A., Barajas-Solano, D., He, Q., Journal of Computational Physics (2020); He, Q. et al. Advances in Water Resources (2020)
Modeling density-driven flow for CO2 sequestration
Ref: Du et al. (2023) Modeling density-driven flow in porous media by physics-informed neural networks for CO2 sequestration. Computers and Geotechnics.
FEM-SUPG
Enriched PINN with Domain Decompostion
Machine learning enhanced meshfree modeling, multiscale modeling, reduced-order modeling
Hyper reduced-order RKPM modeling for thermal fatigue in Electronic Devices
References: Kaneko, S., Wei, H., He, Q., et al., Journal of the Mechanics and Physics of Solids (2021); He, Q., Chen, J. S., & Marodon, C., Computational Mechanics (2019)
Geohazard modeling
Differentiable Simulation
Collapse of granular column (GPU-based simulation with >2 million particles for Drucker–Prager material)
Multiscale modeling of advanced materials and manufacturing
References: Zhang, Y. et al. Int J Numer Method Biomed Eng (2020); He, Q. et al. Int J Adv Manuf Technol (2018); Wei, H., He, Q. et al. Journal of Laser Applications (2017)
Deep learning and computational methods for material design
Reference: He, Q. et al. Computational Mechanics (2014); Wang Y. et al. Computers & Structures (2013, 2014)