Flow over a cylinder calculator
Type the cylinder diameter, the fluid and the speed. You get the Reynolds number, which flow regime that is, the Strouhal number and vortex-shedding frequency, the drag coefficient a steady solver, a 2D transient run and an experiment should each give, and a time step and run length for a transient simulation. It opens on the flow-over-a-cylinder challenge of Skill-Lync's CFD courses. The second tab checks the Strouhal number and Cd from your Fluent or SolidWorks run and names the usual mistakes. Free, no sign-up, runs in your browser.
Williamson 1988 · Fey et al. 1998 · Roshko 1954 · Fornberg 1980 · Tritton 1959 · Updated 20 Nov 2026
Set the case by Re (the velocity follows) or by velocity. The course fluid is a made-up one: ρ = 1 and μ = 0.02 make Re = 100 at U = 1 m/s on a 2 m cylinder. Domain length is inlet to outlet, for the flow-through time.
The case above is what you simulated. Type the St or frequency you got and, if you have it, your Cd. The Reference Values start at Fluent's defaults; change them to what your Reference Values panel shows.
- The twin vortices break away in turn and form a Kármán vortex street. The wake is two-dimensional and periodic: run a transient laminar solver. A steady solver either stalls or converges to the unstable symmetric solution, which has no shedding and less drag.
What is the flow over a cylinder challenge?
It is the first transient case in Skill-Lync's ANSYS Fluent course and the PG program in CFD, and 588 public student reports cover it (titled "Week 2 - Flow over a Cylinder" or "Steady vs Unsteady flow over a cylinder"). A 2 m cylinder sits in a 2D rectangular domain (60 m long and 20 m high in the 50 reports that give it), filled with a made-up fluid of density 1 kg/m³ and viscosity 0.02 kg/(m·s). Two tasks:
- Part I. Run the flow at Re = 100 with a steady and with a transient solver, and calculate the Strouhal number.
- Part II. Calculate the drag and lift coefficients at Re = 10, 100, 1,000, 10,000 and 100,000 with the steady solver, discuss how drag changes with Re, and validate against the literature within 5%.
The two numbers that describe the flow:
Re compares inertia with viscosity: how hard the fluid pushes on through, against how much it sticks. St compares the shedding frequency f with how long the flow takes to pass one diameter. With D = 2 m, ρ = 1 and μ = 0.02, Re = 100 needs U = 1 m/s, and the five Part II cases need 0.1, 1, 10, 100 and 1,000 m/s. A second version of the course uses μ = 0.05 and U = 2.5 m/s for Re = 100. The SolidWorks course has a different challenge: a small cylinder in air at 10 m/s, then 20%, 40% and 100% faster (202 more reports, below).
What Strouhal number should you get at Re = 100?
St = 0.164, so the shedding frequency in the course case is f = St·U/D = 0.0822 Hz: one vortex pair every 12.17 s. That comes from Williamson's fit to his careful laminar-shedding experiments:
Two other standard sources agree within 2%: Roshko's 1954 NACA measurements give 0.167, and Fey, König and Eckelmann's 1998 fit gives 0.165. Over 50 to 150, Williamson and Fey never differ by more than 0.0012, and Williamson and Roshko by 0.0031. Fey et al. cover the range from Re = 47 to 2 × 10⁵ in pieces, with no fit between 240 and 360, where the frequency depends on how the ends of the cylinder are set up:
| Re range | St* | m | St at the start of the range |
|---|---|---|---|
| 47 – 180 | 0.2684 | -1.0356 | 0.117 |
| 180 – 230 | 0.2437 | -0.8607 | 0.191 |
| 230 – 240 | 0.4291 | -3.6735 | 0.187 |
| 360 – 1,300 | 0.2257 | -0.4402 | 0.202 |
| 1,300 – 5,000 | 0.2040 | 0.3364 | 0.213 |
| 5,000 – 2,00,000 | 0.1776 | 2.2023 | 0.209 |
The chart in the calculator plots all three. Below Re ≈ 47 there is no shedding at all, so there is no Strouhal number to find.
Why can't a steady solver give the Strouhal number?
Because a steady solver has no time in it. Each iteration is a step towards a solution that does not change, not a step forward in seconds. At Re = 100 that steady solution exists, two symmetric vortices behind the cylinder, but it is unstable: the real flow, and a transient run, leave it and start shedding. So the steady run either converges to a wake with no shedding (zero lift, Cd = 1.058 in Fornberg's 1980 solution) or its residuals and lift keep oscillating without converging. Either way, an FFT of the iteration history gives a "frequency" per iteration, not per second.
Yet at least 16 reports quote a Strouhal number for the steady case, from 0.0017 to 0.5. The right answer for Part I's steady half is: no shedding and no Strouhal number, and the steady solution is the symmetric wake you would see in a laboratory only below Re ≈ 47.
Which mistakes give the wrong Strouhal number?
Skill-Lync read all 588 Fluent-challenge reports on 20 Nov 2026. 110 state a Strouhal number in text (the rest show it only in a picture or not at all). Their last-quoted value: 29 between 0.155 and 0.175, 29 between 0.14 and 0.155 or 0.175 and 0.21, 29 below 0.14 and 23 above 0.21. Four mistakes come up again and again, and the Check tab catches each:
- Default Reference Values under the FFT. Fluent's FFT can put "Strouhal Number" on its x-axis, computed from the Reference Values panel, not from your cylinder:
so the real Strouhal number is
- Left at the default 1 m and 1 m/s, the axis just shows the frequency. In the course case (D = 2 m, U = 1 m/s) that is half the Strouhal number, 0.0822, and reports quoting 0.076 to 0.088 fit it. In the μ = 0.05 version (U = 2.5 m/s) the axis reads 0.2054, which is where the "St = 0.2, matches the literature" reports come from: right-looking, but it is f in Hz. 43 reports use the FFT's Strouhal axis.
- The wrong signal. Lift swings once per shedding cycle. Drag, and the velocity at a point on the wake centreline, swing twice: a vortex from the top and one from the bottom pass in each cycle and push the same way. Read off those, and St comes out double, about 0.33. The chart in the calculator shows the three signals.
- Counting cycles in a short window. 42 reports work out f as cycles counted divided by a time window, like "f = 4/50 = 0.08". In 50 s the course flow sheds 4.11 cycles, so one cycle more or less moves St by ±0.04: 3/50, 4/50 and 5/50 give 0.12, 0.16 and 0.20. Measure the time across 10 or more whole cycles after the start-up has died out, or take the peak of an FFT of the lift.
- A steady "Strouhal number", above.
What drag coefficient should you get?
It depends on which flow you are comparing with, and a steady solver cannot produce most of them. The drag coefficient is the drag per unit span over ½ρU²D:
where s is the span (1 m in a 2D run). Fluent prints the same force over its Reference Values instead:
| Re | U (m/s), course fluid | Steady 2D, no shedding | Measured | What a steady run gives you |
|---|---|---|---|---|
| 10 | 0.1 | – | 2.93 (Tritton) | The right answer: the flow really is steady |
| 100 | 1 | 1.058 | 1.27 (Tritton) | An unstable solution: less drag than the real 1.33 (2D transient mean) |
| 1,000 | 10 | – | close to 1 (White's fit 1.10) | Nothing physical: the real flow is unsteady and turbulent |
| 10,000 | 100 | – | close to 1 (White's fit 1.02) | Nothing physical: the real flow is unsteady and turbulent |
| 1,00,000 | 1000 | – | close to 1 (White's fit 1.00) | Nothing physical: the real flow is unsteady and turbulent |
Re = 10 is the one case where steady and real agree. At Re = 100 a transient 2D laminar run gives a time-averaged Cd of about 1.33 (Park, Kwon and Choi 1998; Posdziech and Grundmann 2007 get 1.325) with lift swinging about ±0.34 (Liu et al. 1998), against 1.058 for the steady symmetric solution. Tritton's measurements give 1.27: real flow at Re = 100 is close to 2D, and 2D runs come out a few per cent high. From 1,000 up the measured Cd stays close to 1 until the drag crisis near 2 × 10⁵ (White's curve fit gives 1.10 at 1,000 and 1.02 at 10,000; it runs up to 18% above Tritton's points below Re = 100, so treat it as a rough guide). A steady 2D laminar run up there gives numbers that are not drag: the median of the 17 reports that print one at Re = 100,000 is 0.8.
Two more things move Cd in this challenge. A 20 m high domain with a 2 m cylinder blocks 10% of the channel, and if its top and bottom are walls, that pushes Cd up, most at low Re: 14 of the 19 reports that print a Re = 10 value in text give about 3.3, against 2.93 measured by Tritton, 13% more. And the Reference Values: left at Fluent's defaults (ρ 1.225, U 1 m/s, area 1 m²), every Cd in the course case prints 1.633 times too big, so a correct 1.33 reads 2.17.
What time step and how long a run?
Resolve each shedding period with at least 20 time steps, better 40 or more, and run until the start-up has washed out before you measure. For the course case, T = 12.17 s, so Δt = 0.30 s gives 40 steps a period. The flow takes 60 s to cross a 60 m domain once at 1 m/s, and shedding usually needs a few flow-through times to settle from a symmetric start. Then average Cd and measure f over 10 or more periods: another 122 s. The calculator works these out for your case.
The SolidWorks version: is the frequency even resolved?
The SolidWorks Flow Simulation challenge (202 reports, "Flow over a cylinder") runs a small cylinder in air, typically D = 0.04 m at 10 m/s, then 20%, 40% and 100% faster, transient. In 32 of the reports that state it, the total analysis time is 0.1 s, and 24 save results every 0.01 s. Air at 20 °C:
| U (m/s) | Re | St (Fey et al.) | Shedding f (Hz) | Cycles in 0.1 s | Saved frames per cycle at 0.01 s |
|---|---|---|---|---|---|
| 10 | 26,494 | 0.191 | 47.8 | 4.8 | 2.09 |
| 12 | 31,792 | 0.190 | 57.0 | 5.7 | 1.75 |
| 14 | 37,091 | 0.189 | 66.2 | 6.6 | 1.51 |
| 20 | 52,987 | 0.187 | 93.6 | 9.4 | 1.07 |
Saving every 0.01 s is 100 samples a second, so nothing faster than 50 Hz can be seen in the saved results (the Nyquist limit). At 10 m/s the wake sheds at about 48 Hz, just under it; at 12, 14 and 20 m/s it sheds faster than 50 Hz, and any frequency read from those frames, or any animation of them, is aliased: it shows a slower, false oscillation. And 0.1 s holds only about 5 to 9 cycles, part of them start-up. Save at least 20 frames a period (every 0.001 s or less here) and run long enough for 10 settled cycles. The calculator gives f for any speed; set the fluid to air.
What this calculator does not cover
- Blockage and domain size corrections. The reference values are for a cylinder in an unbounded fluid. A narrow channel raises both Cd and St; use a domain of 20 D or more each side if you want to compare with them directly.
- Turbulence-model results. What URANS or LES should give at Re = 10⁴ to 10⁵ depends on the model, the mesh and the span; the page gives the measured values to compare with.
- The critical and supercritical range (above about 2 × 10⁵): the drag crisis depends on surface roughness and free-stream turbulence, and shedding has no single frequency.
- Other shapes and moving cylinders. Square cylinders, cylinders near a wall, rotating or vibrating cylinders (lock-in) all change St and Cd.
Flow over a cylinder is the first case where a simulation has to be time-accurate, and getting its Strouhal number right is how you learn reference values, time steps and validation. It is week 2 of Skill-Lync's Fluent course.
Frequently asked questions
What is the Strouhal number for flow over a cylinder at Re = 100?
0.164 (Williamson 1988, from experiments; Fey et al. 1998 give 0.165 and Roshko 1954 0.167). Good 2D laminar transient simulations land at 0.164 to 0.166. For the Fluent challenge (D = 2 m, ρ = 1 kg/m³, μ = 0.02 kg/(m·s), U = 1 m/s) that is a shedding frequency of 0.0822 Hz, one cycle every 12.17 s.
At what Reynolds number does vortex shedding start?
At about Re = 47 (Henderson 1997; Williamson gives 49). Below it the wake is two steady, symmetric vortices and there is nothing to shed. From about 190 the wake turns three-dimensional, and above about 2 × 10⁵ the boundary layer turns turbulent (the drag crisis).
Why is my Strouhal number 0.08 instead of 0.16?
Almost always the Reference Values. Fluent's FFT "Strouhal Number" axis is f × length / velocity from the Reference Values panel. Left at the default 1 m and 1 m/s, it shows the frequency in Hz, which for D = 2 m and U = 1 m/s is half the Strouhal number: 0.0822. Set the reference length to D and the velocity to U, or read the frequency and compute fD/U yourself.
Why is my Strouhal number about 0.32?
You read the frequency from the drag coefficient or from the velocity at a point on the wake centreline. Both oscillate twice per shedding cycle, because a vortex from the top and one from the bottom pass in each cycle. Use the lift coefficient: it swings once per cycle. Halve the 0.32 and you are back at 0.16.
Why is my drag coefficient 2.17 at Re = 100?
Fluent divided the drag force by its default Reference Values (density 1.225, velocity 1 m/s, area 1 m²) instead of your fluid (ρ = 1) and the cylinder's area (D × 1 m = 2 m² in 2D). That makes every coefficient 1.633 times too big: the time-averaged 1.33 prints as 2.17. Set Reference Values to compute from the inlet and type the area and length yourself.
Can a steady solver give the drag of a cylinder at Re = 1,000 to 100,000?
Not reliably. Above Re ≈ 47 the real flow is unsteady, and above a few hundred the wake is turbulent. A steady 2D run either never converges or settles on a solution that does not exist in nature, so a 5% match with the measured drag (close to 1 in this range) is luck. Run transient, with a turbulence model at the higher Re, and average the drag over many shedding cycles.
Related on Skill-Lync
- Free lessons: GUI based CFD using ANSYS Fluent (includes the 2D flow over a cylinder lesson)
- Student project: Week 2, flow over a cylinder
- ANSYS project ideas and SolidWorks project ideas
- Fluid mechanics book: boundary layers and drag
- Converging-diverging nozzle calculator (the MacCormack assignment)
- 2D heat conduction solver
- CFD courses compared
- All free engineering tools
Sources
- C. H. K. Williamson, "Defining a universal and continuous Strouhal-Reynolds number relationship for the laminar vortex shedding of a circular cylinder", Physics of Fluids 31 (1988) 2742: St = −3.3265/Re + 0.1816 + 1.6 × 10⁻⁴ Re.
- C. H. K. Williamson, "Vortex dynamics in the cylinder wake", Annual Review of Fluid Mechanics 28 (1996) 477-539: the regimes.
- U. Fey, M. König, H. Eckelmann, "A new Strouhal-Reynolds-number relationship for the circular cylinder in the range 47 < Re < 2 × 10⁵", Physics of Fluids 10 (1998) 1547 (table as reproduced in Falcato, University of Coimbra 2020, Table 2.1).
- A. Roshko, "On the development of turbulent wakes from vortex streets", NACA Report 1191, 1954.
- R. D. Henderson, "Nonlinear dynamics and pattern formation in turbulent wake transition", Journal of Fluid Mechanics 352 (1997) 65-112: shedding from Re ≈ 47, 3D transition at 190 and 260.
- B. Fornberg, "A numerical study of steady viscous flow past a circular cylinder", Journal of Fluid Mechanics 98 (1980) 819-855: steady symmetric Cd 2.000, 1.498, 1.058, 0.829, 0.722 at Re 20, 40, 100, 200, 300 (Dennis and Chang's 1970 values at 20, 40, 100 agree within 2.3%).
- D. J. Tritton, "Experiments on the flow past a circular cylinder at low Reynolds numbers", Journal of Fluid Mechanics 6 (1959) 547-567: measured Cd at Re 10 to 100, as tabulated in arXiv 2007.07347.
- Re = 100 two-dimensional benchmarks as compiled by Qu, Norberg and Davidson, Journal of Fluids and Structures 39 (2013): Park, Kwon and Choi 1998 (Cd 1.33, St 0.165), Posdziech and Grundmann 2007 (1.325, 0.1644), Liu, Zheng and Sung 1998 (1.350, lift ±0.339).
- F. M. White, Viscous Fluid Flow: the cylinder drag curve fit Cd ≈ 1 + 10 Re^(-2/3), quoted for 1 < Re < 2 × 10⁵.
- Ansys Fluent User's Guide (Release 12.0, sections 29.11.4, 30.11, 33.10): force coefficients use the reference area, density and velocity; the FFT's Strouhal number is f·L_ref/U_ref from the Reference Values; computing Reference Values from a boundary does not set the length and area; default air 1.225 kg/m³ and 1.7894 × 10⁻⁵ kg/(m·s).
- Skill-Lync, all 961 public flow-over-a-cylinder reports (588 Fluent, 202 SolidWorks, the rest other software), read 20 Nov 2026: Strouhal numbers, signals, Reference Values, drag tables, time steps.
- Skill-Lync, scripts/verify/cylinder.py: re-implements every fit and table from the papers, checks the fits against each other, works the course, SolidWorks and mistake cases, and writes the numbers this page checks itself against at build time.