Adrian Sandu
Virginia Polytechnic Institute and State University, VA·Computer Science
Complex engineered and natural systems are understood largely through computer simulation, and two difficulties limit how good those simulations can be. Different physical processes inside one system run at wildly different speeds, and models must be reconciled with imperfect sensor measurements.
- Fields
Computer Science
Mathematics & Statistics
Engineering
- Topics
digital twin
jet engine
data assimilation
- Grants
- 2 known grants, through Aug 2027
What they work on
- Develop high-fidelity multiphysics models of jet engine combustion and flow together with scalable reduced-order models, both with quantified uncertainties, as the computational core of a predictive digital twin.
- Construct particle-surface interaction models that quantify how erosion and deposition from ingested particles degrade jet engine performance.
- Design hierarchical data assimilation algorithms based on variational approaches, ensemble Kalman filters and transport map particle filters, validated against a JetCat P100-RX engine as the physical twin.
- Build a hybrid time integration framework that mixes discrete and continuous internal stages within each step, so multiscale systems can be co-simulated with component-specific discretizations and time steps.
- Establish local truncation error estimates and higher orders of accuracy than existing co-simulation methods for partial differential equation systems, and release open implementations for community use.
More about the research
Two projects attack these problems from the mathematical side. One constructs predictive digital twins of jet engines that fuse simulation with measurement. The other designs new algorithms for advancing simulations in time when their parts evolve at different rates.
Why it matters
Better digital twins of jet engines could help manufacturers and operators catch developing faults, plan preventive maintenance and improve aircraft safety, while exploring conditions too hazardous to test physically. More accurate and efficient time integration for systems with fast and slow components could improve simulations across atmospheric chemistry, oceanography, plasma physics, aeronautics and environmental modeling, and the resulting software will be openly available.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Anil Deane
University of Maryland, MD·Earth & Environmental Science
Build a digital twin linking the University of Maryland's 3-meter liquid sodium geodynamo experiment to high-order spectral element simulations (Nek5000/RS) of the flow inside it.
Alexander Robel
Georgia Institute of Technology, GA·Earth & Environmental Science·Early-career
Build an open-source software package that applies transient ensemble Kalman filtering, a method that repeatedly corrects many model runs against observations, to a wide range of ice sheet models.
Yiwei Wang
University of California-Riverside, CA·Mathematics & Statistics
Develop high-order variational Lagrangian schemes, which track moving parcels of material, for generalized diffusion models, built directly from the energy and dissipation description of the fluid.
The school
Virginia Polytechnic Institute and State University: 35 labs on PI Pages, $657M a year on research, 496 research doctorates a year.
More about Virginia Polytechnic Institute and State University
Reach out
A start for your email. Fill in the brackets.