Portrait of Saeed Salehi

Saeed Salehi

Associate Professor of Fluid Mechanics, Docent

Division of Applied Thermodynamics and Fluid Mechanics · Department of Management and Engineering · Linköping University, Sweden

I study complex fluid flows with high-fidelity computational fluid dynamics (CFD), and combine it with data-driven methods and machine learning to build efficient and reliable tools for simulation, reduced-order modelling, flow control and uncertainty quantification. The applications range from canonical flows to turbomachinery and energy systems.

I am first and foremost a CFD specialist and an active OpenFOAM developer. My current focus is deep reinforcement learning for active flow control, coupled directly to the flow solver.

Research themes

All research →
Schematic of a reinforcement learning agent interacting with a turbulent flow simulation through states, rewards and actions

Reinforcement-learning flow control

Deep reinforcement learning coupled with CFD for active control of unsteady and turbulent flows.
Close-up of the vortex rope in the draft tube of a Francis turbine

Hydraulic turbine transients

Resolved simulation of start-up, shutdown and load changes, with new mesh-motion methods in OpenFOAM.
Annotated vortex rope structures in a draft tube, coloured by axial velocity

Modal analysis and reduced-order models

POD, SPOD and DMD to extract coherent structures and dominant frequencies from simulation data.
Architecture of a physics-informed neural network with periodic input layer and loss function

Physics-informed neural networks

Multi-fidelity PINNs for solving partial differential equations, including free-surface waves.
Temperature contours on a baseline cooled turbine vane and on a robust-optimum design

Uncertainty quantification

Sparse and multi-fidelity polynomial chaos methods and robust design under uncertainty.
Vortex rope below the runner of a Kaplan turbine model at part load

Flexible hydropower

Flow-induced loads, turbine lifetime and contra-rotating pump-turbines for energy storage.

News

All news →

Sep 2026

The application period for the PhD position in AI-based flow control for hydropower has closed, and assessment and recruitment are under way. We received a large number of applications, so this will take some time. Thank you to all applicants for your patience.

Aug 2026

New review paper: State-of-the-art implementations of PINNs for the lid-driven cavity problem, in Results in Engineering.

Jul 2026

Talk on closed-loop control of the vortex rope in a swirl generator with deep reinforcement learning, and session chair, at the 21st OpenFOAM Workshop in Guimarães, Portugal.

Jun 2026

Appointed Section Editor of the OpenFOAM Journal.

Jun 2026

Announced: a PhD position in AI-based active flow control for hydraulic turbines, funded by the ÅForsk Foundation. The call has since closed (31 August 2026); see the project page.

Current project

ÅForsk Early-Career Research Grant

Enabling flexible hydropower operation by AI-based active flow control

Deep reinforcement learning coupled with high-fidelity CFD to suppress harmful flow instabilities in hydraulic turbines at off-design operation. Funded by the ÅForsk Foundation, 2027–2031, with a doctoral student being recruited at Linköping University.

About the project →

Students

Thesis projects and PhD opportunities

I supervise MSc and BSc theses in CFD, OpenFOAM development and machine learning for fluid flows, and I support applications for external PhD and postdoctoral fellowships.

Opportunities →

Recent publications

All publications →
  1. State-of-the-art implementations of PINNs for the lid-driven cavity problem – a critical review with future perspectivesM. Sheikholeslami, S. Salehi, W. Mao, A. Eslamdoost and H. NilssonResults in Engineering, 32, 112399 (2026) DOI
  2. Lifetime analysis of hydro turbines with focus on fatigue damage in a renewable energy system – A reviewM. Nobilo, S. Salehi and H. NilssonRenewable and Sustainable Energy Reviews, 228, 116578 (2026) DOI
  3. Physics-informed neural networks with hard and soft boundary conditions for linear free surface wavesM. Sheikholeslami, S. Salehi, W. Mao, A. Eslamdoost and H. NilssonPhysics of Fluids, 37(8), 087158 (2025) DOI
  4. An efficient intrusive deep reinforcement learning framework for OpenFOAMS. SalehiMeccanica, 60(6), 1673–1693 (2025) DOI
  5. Transfer learning strategies for accelerating reinforcement-learning-based flow controlS. SalehiarXiv preprint arXiv:2510.16016 (2025) arXiv