Research Interests
Turbomachinery
DNS and modeling of flow in gas-turbine components.
Scientific Machine Learning
Physics-informed neural networks and operators for fluid dynamics.
High-Speed Flows
Shock–boundary-layer interactions, hypersonics and flow control.
HPC & Computation
Development of massively parallel, GPU-enabled CFD solvers.
Research Grants
≈ ₹11 crore secured across 13 sponsored projects as Principal Investigator / Co-Investigator.
Quantum Machine Learning for Reservoir Computing and PDE Solvers
Physics-Informed Machine Learning for Enhanced Turbine Blade Flow Predictions: Bridging Experiments and Computations
Flow Control for Next-Generation Aircraft Using Boundary Layer Ingestion Engines
Cyclic Thermal Testing of TBC Coated Super-alloys in a Burner Rig
Design and Development of a Model Cargo-Hyperloop Using Pipe Following Robot
Forecasting Weather and Climate using Dynamics and Artificial Intelligence
Development of a GPU-enabled LES Solver for Unsteady Hypersonic Flows Using Curvilinear Grids
Prime Minister's Research Fellowship (PMRF) Sub Project
Ganita, Darśana and Kalā – Course Content Creation
Ganita, Darśana, Nīti, Kalā and Ayurveda – Integration of IKS into Holistic School Education
Computational Investigation of Flow over Low-drag Turboprop Wings in Tractor Configuration
Numerical Characterization of Transitioning Boundary Layers in Turbomachines
Special Research Grant
Research Topics
Our group investigates various flow phenomena encountered in aerospace applications as well as in fundamental flows. We employ high-fidelity spatio-temporal simulations (DNS, LES, Hybrid LES/RANS) in conjunction with state-of-the-art analysis tools (operator-based and data-driven) for flow characterization and subsequent control prospects. .
TURBOMACHINERY FLOWS
Flows in low- and high-pressure turbines and compressors are complex due to separation, transition and relaminarization — phenomena beyond most turbulence models. We use high-fidelity DNS, LES and hybrid LES/RANS to quantify losses and heat transfer in these components.
APPLIED AERODYNAMICS
We run high-fidelity simulations of unsteady aerothermodynamics for aircraft, missiles, rockets, spacecraft and UAVs. Focus areas include high-lift flaps and slats, optimal turboprop wings with propeller interaction, upswept cargo-aircraft afterbodies, and hypersonic flows.
TURBULENCE & TRANSITION MODELING
We combine theory, high-fidelity datasets and machine learning to improve or propose transition models. Efforts target data-driven transition modelling, characterizing reverse transition (relaminarization) within RANS, and including curvature and displacement effects in boundary layers.
STABILITY & CONTROL
We apply modal (local, BiGlobal, TriGlobal) and non-modal (transient-growth) linear stability analysis to track dominant modes in laminar and turbulent flows. Resolvent and adjoint methods guide optimal control, with a Mean-Flow-Perturbation approach for complex turbulent cases.
DATA-DRIVEN ANALYSIS & ROM
High-dimensional spatio-temporal flow data, from experiment or simulation, demand optimal analysis tools. We use statistical, spectral and low-order decomposition (POD/DMD) techniques to extract insight and build reduced-order models, with parallel codes scaling to millions of grid points.
SCIENTIFIC MACHINE LEARNING
We develop physics-aware machine learning that embeds governing equations into the learning process. Built on our differentiable, JAX-accelerated framework, it spans PINNs and neural operators (FNO, DeepONet), field-inversion (FIML) for turbulence-model calibration, and ML-based inflow generation and fast aerodynamic prediction.
RESEARCH CODES
Several research codes are available in the group for high-fidelity computation and analysis purposes. These codes support flows from low subsonic to hypersonic regimes.
ANUROOP (C++ & CUDA)
- Compressible Navier-Stokes solver
- 3-D cell-centered Finite Volume Approach
- Supports complex configurations & hybrid mesh
- Kinetic energy consistent second-order central differencing
- Extensively validated and highly scalable
- GPU Compatibility
- Used by DRDO-GTRE and DRDO-ANURAG
PRAVAH (FORTRAN)
- Compressible Navier-Stokes solver with low-pass filter
- 3-D Compact High-order Finite Difference Approach with Roe Blending
- Supports complex configurations with curvilinear meshing
- Explicit & implicit temporal schemes
- Scalable, modular & easy to use
- Shock capturing
- Low subsonic to hypersonic regimes
- Derivative code Mean Flow Perturbation used for linearized analysis
List of In-house Codes
QuickStab: Local and BiGlobal Stability (Temporal & Spatial) and Resolvent Solvers.
Spectral Tools: FFT, STFT, Wavelet, Bi-coherence, HHT.
Decomposition Tools: DMD, POD/SPOD, EMD.
Linearised Analysis Tools (iterative): Mean Flow Perturbation (MFP), Shift-Invert Arnoldi.