Fin flutter
See how HPR Sim applies Martin’s criterion for fin flutter, where bending and twisting feed each other, to your fins and your rocket’s flight.
In short
-
What it models: whether a fin may flutter, meaning its bending and twisting feed each other until the fin shakes itself apart. HPR Sim gives two readings of one criterion. The first is Martin’s own chart check: is the fin on the flutter side of the line his flight data draws? The second is the ratio of the criterion’s flutter speed to the rocket’s airspeed at the flight’s peak dynamic pressure. The first decides; the second alone is not a safe line. When fins flutter, a guide, reads both on the guides’ rocket.
-
Sources: D. J. Martin’s criterion in NACA TN 4197 (1958), eq. 18, and his figure 3 (References). The figure separates missile and wind-tunnel wings that fluttered from those that didn’t. The fin’s stiffness enters as its shear modulus. The simulator has one, with its source, for 14 of its built-in materials.
-
How well it is validated: HPR Sim reproduces Martin’s formula and both of his worked examples (Tests). That includes his verdicts on three metals and his titanium design. His safe/unsafe line is a band that this project measured off his printed chart: 0.25 to 0.31 (The line in Martin’s data). His unfailed wings flew to at least Mach 1.3. Nothing here is checked against a hobby rocket, but for one fin seen fluttering in flight whose material is inferred (below; evidence row G22). Where the source allows two readings, the simulator takes the one with the lower flutter speed. That doesn’t make the whole result conservative: nothing shows that it is. Of the published tests the project has set against HPR Sim’s speed, at least four kinds of fin or wing fluttered below it, the flattering side. Each figure is HPR Sim’s speed (from eq. 18, with the thickness ratio taken over the root chord) over the measured onset, so above 1 means HPR Sim’s speed is too high:
- plies that couple bending to twist, as an off-axis layup’s do: two of MIT’s three laminated plates (M2 and M3), 1.13 to 1.42 over the thesis’s two sets of ply properties (measured flutter onsets, §2);
- built-up wings, a skin over spars: the RAE’s rocket-flown designs of ARC Report 2944, 0.92 to 1.41 over all 22 wings (Mach 0.44 to 0.98), median 1.13, 21 above 1, and 1.24 to 1.41 on the wings with a 25% spar flown at Mach 0.75 to 0.98, the top resting on one reading of the report’s stiffness column (measured flutter onsets, §3); and up to 1.47 on ARC Report 2949’s swept wings, their stiffness taken normal to the sweep (transonic flutter onsets, §4);
- stiffness inboard with the tip and trailing edge soft, such as a fin that tapers in thickness or a stiff core with soft edges: Langley’s rocket-model wings W2 (1.05), W4 (1.14) and D2 (1.43 or more, a lower bound) (fin flutter tests, §2);
- solid swept fins: in Langley’s transonic tunnel, tested at Mach 0.66 to 1.42, each wing’s worst point 1.21 to 1.37, from a 60° compreg wing (1.37), a 45° resin-paper wing (1.31) and a 45° solid magnesium wing (1.21), and the 60° wings read high at every Mach number tested; other readings of the 60° wing’s stiffness move its worst point anywhere from 0.91 to 1.80 (transonic flutter onsets, §2 and §3); and ARC Report 2949’s four solid spruce swept wings, rocket-flown, 1.17 to 1.31, or up to 1.42 with their stiffness taken normal to the sweep (the same note, §4).
A tunnel wing in light air reads high too: AGARD wing 445.6’s weakened model, 1.24 to 1.50 at Mach 0.90 to 1.07, where the air was thinnest; eq. 18 has no term for the air’s mass against the wing’s, so it reads high there (measured flutter onsets, §5, a test too different from a fin to count).
Flat plates and other fins of one material read below 1, the safe side, as tested: 0.14 to 0.65 on the counted plates, all clamped at the root (measured flutter onsets), and 0.35 to 0.84 on the uniform fins of fin flutter tests, a few held more softly (the ASROC plate clamped over 6.0 in of its 8.9 in root, Liège’s plates not rigid in plunge). A hobby rocket’s fin seen fluttering in flight, PSAS’s LV2c, reads 0.49 to 0.82 if its fins were G10, as inferred, and 1.23 to 1.51 if they were aluminum, at the speed it was seen fluttering, which may lie above the onset (§5). A softer root lowers the onset: the ASROC missile fin’s plate with its root restraint cut reaches 1.10 on discarded data (§3), and on two root tests the highest safe-side ratios, 0.82 to 0.84, would become 0.90 to 1.09 (§4, an inference). None of these was a fin glued to a tube, and the figures are not a bound: a fin of one of these kinds may flutter further below HPR Sim’s speed than any of them did.
-
What it leaves out: sweep, and how the fin is mounted. Martin’s wings were clamped at the root, and a fin glued to a tube is less stiff there. Also left out are stall flutter at high angles of attack and Mach-number effects, such as a dip near Mach 1. So are the rocket body’s own bending and fins that aren’t trapezoids. G10/FR-4, the commonest fiberglass fin sheet, has no built-in modulus (Shear moduli).
The model
What flutter is, Martin’s criterion for its speed, and what that speed means against the line his data draw.
What flutter is
Air pushing on a fin twists it a little. The twist changes the fin’s angle to the air, so the push changes, and the fin bends. Below a certain speed the fin’s stiffness damps this out. Above it, each cycle feeds the next, and a fin can fail within a second. The speed depends on the fin’s shape and thickness, how stiff its material is in shear, and the air it flies through.
The criterion
Martin starts from Theodorsen and Garrick’s flutter speed for a wing that bends and twists (his eq. 1, from their NACA Report 685 of 1940). He reduces it to a few numbers from one fin’s outline. For a trapezoidal fin with root chord , tip chord , span and thickness :
| Symbol | Meaning | For a trapezoid |
|---|---|---|
| Panel aspect ratio: span over the chord halfway out | ||
| Taper ratio: tip chord over root chord, 0 to 1 | ||
| Thickness ratio | ||
| The fin’s effective shear modulus, Pa | the material’s (Shear moduli) | |
| , | The air’s static pressure and speed of sound | from the atmosphere |
| Where the section’s mass sits: this fraction of the chord behind the quarter chord | 0.25, at mid-chord | |
| Air’s ratio of specific heats | 1.4 |
Eq. 18 gives a flutter speed through a denominator , in pascals:
Martin prints the constant in pounds per square inch (psi) at sea-level pressure ,
which he rounds to 39.3. His (eq. 19) is psi, and . The constant is derived, not fitted. Stiffer or thicker fins flutter faster: grows as and as . At a fixed speed of sound, thinner air raises it as .
A flutter dynamic pressure. The air enters only through , and the dynamic pressure is . So the criterion fixes a dynamic pressure at , the same at every height:
A fin flying at dynamic pressure is below by the ratio . That makes the flight’s least ratio the one at its peak dynamic pressure, “max q”, which Flight metrics already finds between the integrator’s steps.
The line in Martin’s data
Eq. 18’s is not the speed at which a fin is known to flutter. Martin plots against for missiles and wind-tunnel models (his figure 3). Wings that fluttered or failed lie above a shaded band, and wings that flew to at least Mach 1.3 without known failure lie below it.
This project measured the band on a 250 dots-per-inch scan of the figure. Both log axes were calibrated on their tick marks. The band’s edges were traced in 69 pixel columns from = 0.05 to 10 × 10⁶ psi. The band runs at = 0.25 to 0.31 all along that range, from wood to steel. In eq. 18’s terms that is = 3.2 to 4.0, so Martin’s line sits where is 1.8 to 2.0 times the speed of sound. HPR Sim calls this ratio the figure 3 ratio:
| Figure 3 ratio | Martin’s data |
|---|---|
| above 0.31 | mostly wings that fluttered or failed (a few that didn’t lie there too) |
| 0.25 to 0.31 | the band: marginal |
| below 0.25 | wings that flew to at least Mach 1.3 without known failure |
Martin takes where the wing flies. At the launch site’s pressure, the highest a flight sees, the ratio is at its largest. Figure 3’s axis runs from 0.05 to 20 × 10⁶ psi (0.34 to 138 GPa) in . Balsa’s modulus is left of it, so its ratio is an extrapolation.
Which reading decides. The figure 3 ratio does: it is Martin’s own check, and a fin must be below the band. alone is not a safe line, at 1 or at any fixed number. At the flight’s max q the two are tied by , with the Mach number there. So the band is at from about to . A fin is below the band only if is above about : 1.5 at Mach 1.3, but 4.0 at Mach 0.5. Martin’s data show nothing about a fin above the band on a rocket slower than Mach 1.3: treat it as not shown to be safe.
The halved constant. Loft, this project’s predecessor, wrote the constant as per psi, half of . So its flutter speeds were too high, about 41%, on the unsafe side (a lesson from Loft). Apogee Components’ Peak of Flight newsletter printed the same halved form in issues #291 (2011) and #411 (2016), and corrected it in #615 (2023). When fins flutter shows what it does to the guides’ rocket.
Worked examples
Martin’s two examples, then a rocket’s fins in a flight.
Martin’s worked examples
Martin gives two examples (pp. 6–7). The first is a wing with , 4% thick, untapered and ground-launched. He reads its off his figure 4 as “about 1.25 × 10⁶ psi” and judges the wing by material. The second picks titanium and holds figure 3’s ordinate to 0.8 × 10⁶ psi, well under titanium’s modulus, “to allow a reasonable margin of safety”. It then asks how thick the wing must be at each aspect ratio.
| Example | Martin | HPR Sim from eq. 19, psi or % | Difference |
|---|---|---|---|
| for , 4% thick | “about 1.25 × 10⁶” psi | 1.228 × 10⁶ psi | −1.8% |
| Titanium at 0.8 × 10⁶ psi, | 2.5% thick | 2.54% | +1.6% |
| Same, | 4.5% | 4.61% | +2.4% |
| Same, | “about 6.5”% | 6.43% | −1.1% |
Martin read these off a log-scale chart. He printed them on grids of 0.05 × 10⁶ psi and half a percent (his thicknesses all end in .5). Each of HPR Sim’s values rounds to his on that grid.
His verdicts on the first wing are the margin half of the example. The titanium row is his second example, at the ordinate he chose. HPR Sim checks them with the moduli Martin marks on figure 3’s axis, each a small box read off the same scan:
| Material | Martin’s mark, 10⁶ psi | Figure 3 ratio | Martin says | Against the band |
|---|---|---|---|---|
| Magnesium | 2.40 to 2.63 | 0.47 to 0.51 | “in the flutter region” | above |
| Aluminum | 3.82 to 4.28 | 0.29 to 0.32 | “marginal” | on it |
| Steel | 8.92 to 11.3 | 0.11 to 0.14 | “probably safe” | below |
| Titanium, his second example at 0.8 × 10⁶ psi | 5.78 to 6.34 | 0.13 to 0.14 | his design, with “a reasonable margin of safety” | below |
A rocket’s fins
The example program fin_flutter (in crates/hpr/examples/) takes the repository’s synthetic
54 mm (2.1 in) rocket. Its three fins are 3.175 mm (0.125 in) thick. Each has a 150 mm (5.9 in) root, a 60 mm (2.4 in) tip
and a 70 mm (2.8 in) span.
It flies the rocket on an I175 motor from a site 200 m (656 ft) up, and prints:
Panel: aspect ratio 0.667, taper ratio 0.400, thickness ratio 0.0212
Max q: 72455 Pa at 2.09 s, 432 m above the pad
Top speed: 355 m/s at 2.10 s; top Mach number: 1.05 at 2.10 s
Martin's band (figure 3): D/G_E from 0.25 to 0.31
material G (GPa) q_f (kPa) V_f 0 m (m/s) V_f 3 km (m/s) V_f/V at max q D/G_E
aluminum_6061 26.200 836.3 1169 1356 3.40 0.083
carbon_fiber 4.826 154.1 502 582 1.46 0.450
birch_plywood 0.750 23.9 198 229 0.57 2.893
basswood 0.511 16.3 163 189 0.47 4.246
balsa 0.138 4.4 85 99 0.25 15.680
The columns V_f 0 m and V_f 3 km are eq. 18’s flutter speed in standard air at sea level and
3 km (1.9 mi) above it. D/G_E is the figure 3 ratio at the launch site’s pressure.
- Aluminum passes both readings. Its figure 3 ratio is well below the band, and at max q the rocket flies at under a third of . A ratio below the band is no guarantee on its own: Langley’s D2 wing, its stiffness inboard, sat below it at 0.17 to 0.18 and fluttered from about Mach 1.69 (fin flutter tests, §2).
- Carbon fiber at 3.175 mm (0.125 in) has of 1.46. But its figure 3 ratio, 0.45, is above the band, where most of Martin’s wings fluttered or failed: not shown to be safe. Its modulus is a unidirectional ply’s of one aerospace prepreg. A ±45° layup of it would be stiffer, but wet-laid or woven hobby sheet may be softer.
- Plywood, basswood and balsa fins of this size fail both: at max q the rocket flies at 1.75 times plywood’s .
CI checks that the program still prints exactly this.
In code, with a FlutterPanel (its API page
has a worked example that CI runs):
// `fin_set` is a design's `FinSet`; `summary` a flight's `FlightSummary` (Flight metrics).
let panel = FlutterPanel::of_fins(&fin_set)?;
let g = materials::shear_modulus_of(&fin_set.material).unwrap().shear_modulus_pa;
let chart = panel.figure_3_ratio(g, launch_pressure_pa)?; // against FIGURE_3_BAND
let margin = panel.margin(g, &summary)?; // V_f/V at max q; `None` if it never flew
FlutterPanel::new takes , and directly. Martin’s figure 4 covers from 0.5 to 3
and from 1% to 10%. The example’s 0.667 is inside. Outside that range the numbers are an
extrapolation.
Readings and materials
Readings and shear moduli.
Readings where the source leaves room
Most of these give the lower of the flutter speeds the source allows. Two can go the other way: an airfoiled fin’s modulus, and a booster’s max q.
- Thickness ratio at the root. Martin’s wings keep one thickness ratio from root to tip. A hobby fin keeps one thickness, so its ratio grows toward the tip. The simulator takes the root’s, the smallest.
- A solid fin’s is its material’s . Martin says a solid wing of aluminum plots at aluminum’s modulus (p. 6), and the simulator does the same. His definition, (eq. 12), assumes a thin airfoil’s torsion constant (eq. 10). A flat plate’s is , which would double and raise the flutter speed by . The simulator keeps the lower reading. The exception runs the other way. For a fin with a NACA four-digit section, eq. 12 gives . The simulator’s for such a fin is then up to 2.9% high (rounded up from ).
- Martin’s taper factor. His derivation carries a factor . Here corrects the twisting frequency for taper (his eq. 8) and gives the chord three-quarters of the way out (his eq. 14). He replaces the factor with . The two agree at and are 3% apart at . Between, is up to 47% larger (at ), which lowers there by up to 17.5%. The simulator keeps his form, since his figure 3 was drawn with it.
- Every fin set sees the whole flight’s max q. A booster’s fins leave at the separation, so the flight’s max q can come after they are gone. Their true is then at least the one given. That holds as long as the booster’s own dynamic pressure after the separation stays below the flight’s peak. The simulator doesn’t check that.
Shear moduli
The criterion needs the fin’s shear modulus in its own plane. hpr_design::materials::SHEAR_MODULI
gives it for 14 built-in materials, each with its source, page and the web address it was read
from. shear_modulus_of finds a design’s copy of a built-in material by its name and density. Where a source gives a
range, the lower value is kept. Metals are given in ksi (1,000 psi) and Msi (10⁶ psi); 1 Msi is
6.895 GPa.
| Material | shear modulus (GPa) | Source | Basis |
|---|---|---|---|
| Aluminum 6061 | 26.2 | MIL-HDBK-5J, Table 3.6.2.0(b1): 3.8 × 10³ ksi | stated |
| Aluminum 7075 | 26.9 | MIL-HDBK-5J, Table 3.7.6.0(b1): 3.9 × 10³ ksi | stated |
| Steel (plain carbon) | 75.8 | MIL-HDBK-5J, Table 2.2.1.0(b): 11.0 × 10³ ksi | stated |
| Titanium Ti-6Al-4V | 42.7 | MIL-HDBK-5J, Table 5.4.1.0(b): 6.2 × 10³ ksi (TIMET: 6.2 and 6.66 Msi) | stated |
| Carbon fiber/epoxy | 4.83 | NCAMP (a composite-material qualification program) report on Hexcel 8552 AS4 tape: in-plane 0.70 Msi, room temperature | stated |
| Birch plywood | 0.750 | Riga Wood Plywood Handbook, Table 4.11: panel shear | stated |
| Birch (yellow) | 1.04 | Wood Handbook ratio × bending modulus × 1.10 | derived |
| Oak (northern red) | 1.11 | same | derived |
| Maple (sugar) | 0.873 | same | derived |
| Spruce (Sitka) | 0.725 | same | derived |
| Basswood | 0.511 | same | derived |
| Balsa | 0.138 | same | derived |
| Acetal (Delrin) | 1.06 | data sheet | derived |
| Nylon 6/6 | 0.490 | data sheet , conditioned | derived |
How the derived values are made:
- Woods: is the shear modulus in the plane along the grain and along the growth rings. The simulator takes the Wood Handbook’s ratio (Table 5-1) and multiplies it by an estimate of , the stiffness along the grain. As the table’s footnote says, that estimate is the wood’s bending modulus at 12% moisture raised by 10%. is the smaller of the two in-plane ratios for every wood listed. Eastern white pine is not in Table 5-1, so it has none. Martin marks solid wood at 0.070 to 0.120 × 10⁶ psi (0.48 to 0.83 GPa). The handbook’s birch and oak are above that.
- Plastics: an unfilled plastic taken as isotropic (the same in every direction), from its data sheet’s tensile modulus and Poisson’s ratio . Nylon soaks up water. Its conditioned value, measured after it has, is less than half the dry one.
- Carbon fiber: a unidirectional ply’s in-plane . That is a 0/90 laminate’s in-plane shear modulus. Plies at ±45° raise it several times, so for such a laminate this reads low.
No source found states a shear modulus for G10/FR-4: its data sheets give flexural moduli only. There is none for PLA, ABS, PETG, polycarbonate or acrylic either. For those, pass your own from a measurement or your supplier. A woven glass/epoxy laminate is not isotropic, so doesn’t apply to it.
Tests and references
The tests that pin the criterion, and the sources it comes from.
Tests
In crates/hpr-sim/src/flutter.rs, unless named otherwise:
| Test | What it pins |
|---|---|
flutter_denominator_matches_tn_4197_eq_18 | The constant against Martin’s 39.3 psi, and against eq. 18 on three panels at two pressures, to his rounding; twice Loft’s constant (its lesson) |
martins_worked_examples | Both of Martin’s examples on his grid, and his three verdicts and his titanium design against the band, through figure_3_ratio (the tables above) |
scaling_laws_in_thickness_shear_modulus_and_pressure | as , , and , to 1e-12; the same in three atmospheres |
the_halved_constant_reads_a_higher_flutter_speed | The halved constant printed in Peak of Flight #291 and #411, and Loft’s, reads above eq. 18’s flutter speed on three panels in two airs, and higher by the same factor: the flattering side, which When fins flutter marks |
taper_factor_against_the_frequency_factors | against : 3% at , at most 47% at |
the_margin_is_the_flights_least_at_its_peak_dynamic_pressure | On the flight of Valetudo (a RocketPy example rocket), the margin is at max q and at or below every row of a 1 ms record; from each row’s pressure and speed of sound agrees with , to 1e-12; no peak gives None, a peak that isn’t a number is refused |
a_trapezoidal_fin_set_gives_its_panel, out_of_range_inputs_are_refused | A fin set’s , and ; elliptical and reverse-tapered fins, a taper outside 0 to 1 and inputs that aren’t positive are refused by name, from code and from JSON |
hpr_design::materials::tests::shear_moduli_reproduce_the_sources | Each built-in modulus against its source’s numbers |
References
- D. J. Martin, Summary of Flutter Experiences as a Guide to the Preliminary Design of Lifting Surfaces on Missiles, NACA TN 4197, 1958. Appendix, eqs. 1 to 19, pp. 11–15; examples, pp. 6–7; figures 3 and 4, p. 19.
- T. Theodorsen and I. E. Garrick, Mechanism of Flutter: A Theoretical and Experimental Investigation of the Flutter Problem, NACA Report 685, 1940. Martin’s ref. 6, the source of his eq. 1; not read for this project.
- MIL-HDBK-5J, Metallic Materials and Elements for Aerospace Vehicle Structures, 2003.
- Forest Products Laboratory, Wood Handbook, FPL-GTR-190, 2010, Tables 5-1, 5-3a and 5-5a.
- NCAMP, Hexcel 8552 AS4 Unidirectional Prepreg Qualification Statistical Analysis Report, NCP-RP-2010-008 Rev D, 2011, Table 3-3.
- Riga Wood, Plywood Handbook, 2022, Table 4.11.
- Celanese, Zytel 101L NC010 data sheet, 2023; Delrin 100P NC010 data sheet.
The decision behind these choices is ADR-078, from the M1.10b milestone.