Accuracy assessment of the Non-Ideal Computational Fluid Dynamics model for siloxane MDM from the open-source SU2 suite

Giulio Gori, Marta Zocca, G. Cammi, Andrea Spinelli, Pietro Marco Congedo, Alberto GuardoneEuropean Journal of Mechanics - B/Fluids 2019 · DOI

TL;DR

This paper presents the first experimental validation of a Non-Ideal Compressible-Fluid Dynamics (NICFD) model for siloxane MDM fluid, using the open-source SU2 CFD suite. The study demonstrates that the NICFD model accurately predicts supersonic flows in a converging-diverging nozzle, even under conditions close to the liquid-vapor saturation curve, which are crucial for Organic Rankine Cycle (ORC) applications.

What Problem Does This Paper Address?

Computational Fluid Dynamics (CFD) is a powerful tool for simulating fluid flows, but most existing CFD solvers are designed for ideal gases, which follow simple equations of state. However, many real-world applications, particularly in Organic Rankine Cycles (ORCs) that use complex fluids like siloxanes, involve flows that deviate significantly from ideal gas behavior, especially near the critical point. These "non-ideal" flows require specialized models and solvers. While NICFD models exist, there's a significant lack of experimental validation for these models in CFD codes. This gap makes it difficult to trust the predictions of NICFD simulations for designing and optimizing systems operating in these non-ideal regimes. This paper aims to bridge this gap by rigorously assessing the accuracy of a NICFD model within the SU2 CFD suite against experimental data for siloxane MDM.

Methods

The researchers employed a multi-pronged approach:

  • NICFD Model Implementation: They utilized the open-source SU2 CFD suite, which was extended to handle Non-Ideal Compressible-Fluid Dynamics. This involved incorporating advanced thermodynamic models (specifically, a Span-Wagner equation of state optimized for MDM) and modifying the solver's core to account for the complex fluid behavior.
  • Experimental Setup: The experiments were conducted using the TROVA test-rig, a blow-down facility designed to study the expansion of molecular complex fluids in NICFD conditions. This rig simulates supersonic flows through a planar converging-diverging nozzle, mimicking conditions found in ORC turbines.
  • Uncertainty Quantification (UQ): A crucial aspect of this study was the rigorous application of UQ. The researchers acknowledged that experimental measurements and operating conditions are never perfectly known. They used a non-intrusive Polynomial Chaos Expansion (PCE) method to propagate these uncertainties (primarily from inlet conditions like pressure and temperature) through the CFD simulations. This allowed them to not only predict the mean flow behavior but also to quantify the uncertainty associated with these predictions.
  • Grid Sensitivity Analysis: Before the main assessment, the authors performed a grid sensitivity analysis to ensure that their numerical results were independent of the mesh resolution, a standard practice in CFD to confirm the reliability of the simulation.

Datasets

The study utilized experimental data collected from the TROVA test-rig facility at Politecnico di Milano. This dataset includes:

  • Nominal operating conditions: Total pressure and temperature measurements at the inlet for 10 different test cases (5 for discharge 1, 5 for discharge 2).
  • Experimental pressure measurements: Static pressure readings at specific locations (p1-p4) along the nozzle centerline for each test case, along with their associated expanded uncertainties.
  • Schlieren images: These optical images provided qualitative information about the flow and were used to estimate Mach number distributions.

This dataset represents a novel benchmark for validating NICFD models in supersonic nozzle flows, specifically for siloxane MDM, and differs from previous studies that often focused on simpler fluids like CO2 or lacked comprehensive UQ.

Key Insights and Ideas

  • First Experimental Validation: This paper marks the first time a NICFD CFD model has been rigorously validated against experimental data for a complex fluid like siloxane MDM. This is a significant step towards building confidence in NICFD simulations.
  • UQ is Crucial: The study strongly emphasizes the importance of UQ. By propagating experimental uncertainties, they could provide not just a single prediction but a range of likely outcomes, making the comparison with experimental error bars much more meaningful.
  • Robustness of the Model: The NICFD model in SU2 proved to be remarkably robust. Even with uncertainties in inlet conditions, the predicted flow properties (Mach number, pressure) closely matched experimental values, with the UQ analysis showing limited variability in the numerical results.
  • Influence of Inlet Conditions: The UQ analysis revealed that uncertainties in the inlet total pressure had a greater impact on the static pressure predictions than uncertainties in the inlet temperature, especially in the supersonic flow region.
  • Numerical Dissipation is Low: Despite the complex flow and the use of an inviscid assumption (with some justification), the numerical dissipation was found to be minimal, as indicated by the small variations in entropy and enthalpy.

Key Results

  • Accurate Predictions: The numerical simulations using the NICFD model in SU2 accurately predicted the experimental pressure and Mach number trends for all 10 test cases across a range of non-ideal conditions (compressibility factor Z from 0.633 to ~1).
  • Limited Uncertainty: The UQ analysis showed that the numerical predictions had narrow uncertainty bands (±2σ), indicating that the model's output was not overly sensitive to the uncertainties in the inlet conditions.
  • Good Agreement with Experiments: The mean numerical solutions consistently fell within the experimental error bars for both Mach number and static pressure.
  • Grid Independence: The grid sensitivity analysis confirmed that the results for Mach number and pressure were largely independent of the mesh resolution, especially for finer meshes.

Limitations and Future Work

  • Focus on Inviscid Flow: The simulations were performed assuming inviscid and adiabatic flow. While the authors justify this by noting that viscous effects are negligible on the centerline for this specific geometry and flow regime, a future work could explore the impact of viscosity, especially near the walls.
  • Limited Thermodynamic State Space: The assessment primarily focused on conditions close to the saturation line. While representative of many ORC applications, further validation across a broader range of thermodynamic states would be beneficial.
  • Specific Fluid: The study focused on siloxane MDM. Extending this validation to other fluids used in NICFD applications would broaden the applicability of the findings.
  • Future Work: The authors suggest that this work is a crucial preliminary step towards a comprehensive validation of the NICFD solver. Future work could involve a more detailed treatment of numerical uncertainties, including truncation errors, and potentially extending the validation to more complex geometries or flow phenomena.

Summary updated

BibTeX
@article{gori2019accuracy,
  author = {Giulio Gori and Marta Zocca and G. Cammi and Andrea Spinelli and Pietro Marco Congedo and Alberto Guardone},
  title = {{Accuracy assessment of the Non-Ideal Computational Fluid Dynamics model for siloxane MDM from the open-source SU2 suite}},
  journal = {European Journal of Mechanics - B/Fluids},
  year = {2019},
  doi = {10.1016/j.euromechflu.2019.08.014},
}
Figure