ANSYS CFD

Mechanical Engineering MECE E3100

I. Project Overview

For our fluid mechanics course, our team modeled Nikola Tesla's 1920 valve design - a check valve with no moving parts - in ANSYS Fluent to see how well it actually resists reverse flow. We ran laminar CFD across a range of Reynolds numbers, validated our mesh against a convergence study, then modified the geometry to see if we could beat the baseline design. This project provided an introduction to ANSYS CFD.

As part of a separate paper on complex wing aerodynamics, our team worked on a computational fluid dynamics case study of the ONERA M6 wing - an industry-standard benchmark geometry used to validate CFD setups against published data.

Tesla Valve

II. Tesla Valve Simulation Setup

We quantified the valve using diodicity (D) (the ratio of reverse to forward pressure drop at the same flow rate). A diodicity of 1 means the valve does nothing. A higher value means better one-way behavior. Our baseline design hit D = 1.412 at Re = 350, and across the full range we tested (Re = 50–350), diodicity climbed from 0.975 to that value. This was consistent with published experimental ranges of 1–1.5.

III. Validation

Before trusting any of our results, we ran a grid convergence study comparing a coarse, baseline, and fine mesh. Our baseline mesh landed within 3.1% of the fine-mesh result, versus a 21.6% error for the coarse mesh — good enough to use with confidence for the rest of the study.

IV. Scaling

We tested stacking 1 to 5 repeating valve stages in series. Diodicity rose from 1.18 to 1.46, but with clearly diminishing returns per added stage - useful data for deciding when a longer valve stops being worth it.

V. Valve Redesign

We iterated on three design variables - turn radius, channel length, and entry length - to see which most improved diodicity. Shortening the channel length to 3 cm was the standout change, pushing diodicity to 2.18 (confirmed at 2.85 after remeshing at 50% finer resolution to double-check the result). Combining a smaller radius with the shorter channel actually decreased performance.

Complex Variable Wing CFD

VI. ONERA M6 Wing Simulation Setup

We modeled the flow at Mach 0.84 (a freestream velocity of ~291.6 m/s, worked out from Sutherland's law for viscosity and the ideal gas law), giving a Reynolds number of 11.72 million. The domain used a symmetry plane to simulate only half the wing, a Spalart-Allmaras turbulence model, and an unstructured tetrahedral mesh refined near the wing surface and wake - kept under one million elements to fit within the ANSYS Student solver's limits.

VII. Results & Validation

The converged solution gave C_L = 0.2539 and C_D = 0.0197. Checked against published reference data across multiple solvers, that's a slight lift underprediction (references cluster at 0.26-0.27) and a slight drag overprediction (references cluster near 0.017) - both in the expected direction for a coarser mesh and steady RANS modeling, which tend to smear shocks and add numerical dissipation. The pressure and velocity contours confirmed the physics we were looking for: a leading-edge stagnation point, upper-surface acceleration, and a sharp velocity/pressure jump marking the transonic shock.

IV. Conclusion

This project provided a valuable opportunity to use the ANSYS CFD tool to learn about fluid flow properties and computational analysis.

TeslaValve_Fluids.pdf

Fluids_WingResearch.pdf

Page updated

Google Sites

Report abuse