Reduced-order turbulent flow solver to simulate streamwise periodic fins with iso-thermal walls

This paper presents and validates a reduced-order streamwise periodic turbulent flow solver implemented in the open-source SU2 CFD suite, demonstrating its ability to accurately and efficiently simulate the thermo-hydraulic performance of heat exchangers with repeating structures, such as offset circular fins, by significantly reducing computational costs compared to full array simulations.

The Gist

Imagine you are trying to design the perfect cooling system for a supercomputer or a car engine. These systems use heat exchangers, which are essentially giant, intricate puzzles made of thousands of tiny metal fins. The goal is to blow air through these fins to cool down hot fluids.

To design these efficiently, engineers usually use powerful computer simulations (CFD) to see how air and heat move. But here's the problem: Simulating the whole thing is like trying to count every single grain of sand on a beach. It takes a massive amount of computing power and time—sometimes days or weeks just to run one simulation. This makes it very hard to test new, better designs quickly.

The "Treadmill" Solution

This paper introduces a clever shortcut, a "reduced-order" solver. Think of it like this:

Imagine you are running on a treadmill. Even though you are running for miles, you are actually staying in the exact same spot. The scenery (the air, the heat, the flow) repeats itself over and over again.

In a heat exchanger with repeating fins, the air flow eventually settles into a pattern. After the first few fins, the air doesn't "remember" where it started; it just sees the same pattern of fins repeating endlessly.

The researchers developed a mathematical trick to simulate just one single fin (the "unit cell") and pretend that the air is running on a treadmill. Instead of simulating 100 fins, they simulate one, but they add a special "source term" (a mathematical instruction) that tells the computer: "Hey, pretend this air is part of an infinite line of fins, and the pressure is dropping steadily as it goes."

The New Discovery: Hot Walls and Turbulence

Previous versions of this "treadmill" trick worked well for smooth, calm air (laminar flow) or for walls that were being heated by a constant flame. But this paper solves a specific, difficult puzzle: What happens when the walls are a constant temperature (like a hot metal plate) AND the air is swirling chaotically (turbulent flow)?

Until now, no one had figured out the exact math to make the "treadmill" work for this specific combination. The authors derived the new equations and built them into a free, open-source software called SU2.

The Proof: Does the Shortcut Work?

To prove their shortcut was accurate, they did a "taste test":

1. The Full Meal: They simulated the entire heat exchanger with 11 fins using the old, slow, heavy method.
2. The Sample: They simulated just one fin using their new "treadmill" method.

The Result: The two simulations matched perfectly. The speed of the air, the pressure, and the temperature distribution were identical.

Why This Matters (The "Aha!" Moment)

The difference in time was staggering:

• The Old Way (Full Array): Took about 24 hours (1,440 minutes) on a powerful computer.
• The New Way (Treadmill): Took only 30 minutes.

The Analogy:
If designing a heat exchanger is like trying to find the best recipe for a cake, the old way was like baking 1,000 full cakes to see which one tasted best. The new way is like baking a tiny spoonful of batter, but using a special chemical formula that tells you exactly how the whole cake would taste.

The Takeaway

This research gives engineers a superpower. They can now test thousands of different fin shapes and arrangements in the time it used to take to test just one. This means we can design heat exchangers that are smaller, lighter, and more efficient, which is crucial for everything from electric cars to data centers.

In short: They found a way to simulate an infinite ocean of fins by studying just one drop of water, saving massive amounts of time and computing power without losing any accuracy.

Technical Summary

1. Problem Statement

The thermal and hydraulic design of heat exchangers (HEX) with complex, repeating geometries (such as fin-and-plate or pin-fin arrays) is computationally expensive when using standard Computational Fluid Dynamics (CFD).

• Challenge: Accurately predicting flow and heat transfer requires fine mesh discretization of the entire geometry. Simulating a full array of fins leads to high computational costs, hindering the use of CFD in design optimization loops.
• Current Limitations: While "Streamwise Periodic" (SP) modeling is a well-established technique for reducing computational domains by simulating only a single unit cell, it has historically been limited to laminar flows or constant heat-flux boundary conditions.
• Gap: There is a lack of open literature providing the mathematical formulation and implementation of SP source terms for turbulent flows with isothermal (constant temperature) walls, which is a common boundary condition in evaporators and condensers.

2. Methodology

The authors developed a Streamwise Periodic Reynolds-Averaged Navier-Stokes (SP-RANS) solver implemented in the open-source CFD suite SU2.

Mathematical Derivation

• Governing Equations: The solver solves the incompressible Navier-Stokes equations.
• Momentum Source Term: A constant pressure gradient source term ($-\Delta p / L$) is added to the momentum equation to drive the flow through the periodic unit cell.
• Energy Equation Source Term (Novel Contribution):
  • The authors derived a specific source term ($S_{\theta}$) for the energy equation under isothermal wall conditions.
  • They utilized the concept of dimensionless reduced temperature ($\theta$), defined as $\theta = (T - T_w) / (T_b - T_w)$, where $T_w$ is the wall temperature and $T_b$ is the bulk temperature.
  • Laminar Source Term: Derived based on existing literature (Patankar et al.).
  • Turbulent Source Term: The paper presents a novel derivation for the turbulent source term, accounting for turbulent viscosity ($\mu_{turb}$) and turbulent Prandtl number ($Pr_{turb}$). The term is expressed as:
    $$S_{\theta, turb} = S_{\theta, lam} - \theta \lambda_L \frac{\nabla \mu_{turb} \cdot \hat{i}_L}{\rho Pr_{turb}}$$
  • The decay rate of the bulk temperature difference ($\lambda_L$) is calculated iteratively by integrating the energy equation over the unit cell, solving a quadratic equation derived from the periodicity constraints.

Numerical Setup & Verification

• Test Case: An offset circular pin-fin array with staggered arrangement.
• Comparison Strategy:
  1. Full Array Simulation: A standard RANS solver simulated a domain containing 11 fin elements to establish a "ground truth" and identify the region where flow becomes fully periodic.
  2. SP Simulation: The new SP solver simulated a single unit cell using the derived source terms.
• Flow Regimes: Two cases were tested:
  • Laminar: Reynolds number ($Re$) = 100.
  • Turbulent: Reynolds number ($Re$) = 10,000.
• Turbulence Model: $k-\omega$ Shear Stress Transport (SST) model.
• Grid: Structured quadrilateral meshes with $y^+ \approx 1$ to resolve near-wall effects.

3. Key Contributions

1. Novel Mathematical Formulation: First derivation of the energy equation source term for SP turbulent flows with isothermal walls. Previous works focused on laminar flows or constant heat flux.
2. Implementation in SU2: Integration of these source terms into the open-source SU2 CFD suite, making the tool accessible to the research community.
3. Comprehensive Verification: Rigorous validation of the solver against full-array simulations for both laminar and turbulent regimes, confirming that the reduced-order model accurately reproduces full-domain physics.

4. Results

• Accuracy:
  • Laminar ($Re=100$): The SP solver results for velocity, pressure, and reconstructed temperature fields matched the full-array simulation perfectly. The flow was found to be fully periodic starting from unit cell #5.
  • Turbulent ($Re=10,000$): The SP solver accurately reproduced the full-array results. The flow became fully periodic later than in the laminar case (starting around unit cell #7-8), highlighting that entrance effects are more pronounced in turbulent flows.
  • Quantitative Metrics: The pressure drop ($\Delta p$) and temperature decay rate ($\lambda_L$) calculated by the SP solver matched the full-array data within negligible error margins.
• Physical Insights:
  • Heat Transfer: The laminar case showed a higher temperature decay rate ($\lambda_L \approx 0.23 m^{-1}$) compared to the turbulent case ($\lambda_L \approx 0.03 m^{-1}$), indicating more effective heat transfer per unit length in the laminar regime for this specific geometry.
  • Pressure Drop: Interestingly, the laminar case exhibited a higher pressure drop ($6.41$ Pa) than the turbulent case ($3.24$ Pa) for the specific operating conditions tested.
• Computational Efficiency:
  • Full Array: ~1440 minutes (24 hours).
  • SP Solver: ~30 minutes.
  • Speedup: Approximately 48x reduction in computational time.

5. Significance

• Optimization Enabler: The drastic reduction in computational cost (from days to minutes) makes the SP-RANS solver highly suitable for design optimization of complex heat exchangers. It allows engineers to iterate through thousands of geometric variations (e.g., fin pitch, height, shape) that would be prohibitively expensive with full-array simulations.
• Generalizability: By addressing the isothermal wall condition for turbulent flows, this work bridges a critical gap in CFD modeling, enabling accurate simulation of phase-change heat exchangers (evaporators/condensers) where wall temperature is constant.
• Open Science: The implementation in SU2 ensures that the methodology is reproducible and available for further development by the broader scientific community.

In conclusion, this paper successfully validates a reduced-order modeling technique that significantly lowers the barrier for high-fidelity thermal-hydraulic optimization of heat exchangers, specifically addressing the previously unmodeled scenario of turbulent flow with isothermal boundaries.