Nose cones, transitions and solids of revolution
See the nose cone and transition shapes HPR Sim supports, and how it finds each one’s volume, center of mass, inertia and surface area.
In short
- What it models: the outer shape of nose cones and transitions (conical, ogive, elliptical, power, parabolic and Haack series). For each, the volume, center of mass, inertia and surface areas, solid or as a shell of given wall thickness.
- Sources: G. A. Crowell Sr., The Descriptive Geometry of Nose Cones (1996), and appendix A of the published OpenRocket technical documentation v13.05 (2013).
- How well it is validated: by analytic tests, the first of four kinds of evidence, and against another code. Filled shapes match closed forms to 1e-10 (relative). They match independent high-precision integrals to 1e-12 on 22 noses and transitions, and 20 walls match to 1e-10. HPR Sim weighs three walled probe parts within 3.93e-5 (relative) of OpenRocket’s mass (the probes); the ogive’s, the largest, has no traced cause. The parts catalog’s noses and transitions are compared too (the masses against OpenRocket). Not compared with weighed parts or a real flight.
- What it leaves out: OpenRocket measures a wall across each station, not square to the surface. The two agree on a cone; on three blunt elliptical catalog noses this model weighs up to 0.48% more. HPR Sim also does not cut a steep end square to the axis, as OpenRocket cuts its probe transition. A square cut would add 1.28% and 2.29% of the wall in this page’s examples (Solids of revolution).
Code and sources
Code: hpr_design::shapes (profiles) and hpr_design::solids (volume, centroid, moments, areas).
Sources:
- [CR] G. A. Crowell Sr., The Descriptive Geometry of Nose Cones (1996), pp. 1–6 and 12–14. Cited, not pinned: the only copy found is on a plain-http university mirror, and the reference library pins over https. [TD] gives the same curves.
- [TD] S. Niskanen, OpenRocket technical documentation v13.05 (2013), appendix A, pp. 102–106,
pinned as
openrocket-techdoc-13.05. This is the published document, not the program’s source.
Profiles
A profile gives the outer radius at distance aft of its forward end, . Each shape is a normalized curve with at the tip and at the base:
| shape | or | parameter | source |
|---|---|---|---|
| conical | none | [CR] p. 1; [TD] A.1 | |
| ogive | , below | [CR] p. 4 | |
| elliptical | none | [TD] A.6 | |
| power series | [CR] p. 2; [TD] A.8 | ||
| parabolic series | [CR] p. 5; [TD] A.7 | ||
| Haack series | , below | [CR] p. 6; [TD] A.9–A.10 |
The ogive’s height and the Haack series are:
-
Ogive. The shape is Crowell’s secant ogive, an arc of radius through the tip and the base rim. Its radius is given as a multiple of the tangent radius .
- 1 is the tangent ogive, whose slope is zero at the base.
- Values above 1 meet the base at an angle.
- Values below 1 bulge past before the base.
- The arc reaches the tip only while its center is not above the axis, which needs .
The center is computed on the chord’s perpendicular bisector and the height as
which avoids the cancellation near the tip of a slender ogive.
-
OpenRocket’s ogive parameter is not adopted, because [TD] contradicts itself about it:
- A.3 defines it as .
- A.4–A.5 describe the first of a tangent ogive of length . At , , that has , not .
- The
.orkimporter reads it as A.3 defines it (the format guide). No probe design has run a secant ogive through the OpenRocket jar.
-
Haack. Monotone for , because
is the von Kármán (LD-Haack) ogive and the LV-Haack. [TD] limits to in the program.
-
A Haack nose or widening transition above has no drag data faster than sound. Nor does a bulged secant ogive (arc radius below the tangent ogive’s). So the simulator’s drag buildup has refused them since its extension through Mach 1. Their shapes, mass properties and center of pressure are unaffected, and a drag table stands in for the drag (Aerodynamics).
-
Elliptical, power-series () and Haack tips are blunt: their slope is infinite. (a flat cylinder) is rejected; model it as a tube and a bulkhead.
-
[CR] has errors. Its prose says “greater than twice the length” for the bulged secant ogive, against its own formula. Its ogive and ellipsoid areas are wrong, and the author marked them so. Its elliptical formula measures from the base, and its ellipse CP ratio should read . HPR Sim uses none of those.
Transitions
A transition runs from fore radius to aft radius over .
-
The shape’s tip lies at the smaller end. Growing aft, . A boattail is the mirror image,
[TD] doesn’t say how a shrinking transition is oriented. This choice keeps both ends’ radii exact and the profile monotone (a lesson from Loft).
-
A clipped transition ([TD] §A.7) is cut from a whole nose cone of base radius , where its radius is . The nose length is chosen so the piece is long.
- Shapes other than the ogive invert by bisection.
- The ogive’s curve depends on its fineness, so the nose length is found by bisection too.
- Conical and tangent-ogive transitions are the same clipped or not; the test checks this to 1e-12.
- A clipped ogive with is rejected: its profile isn’t monotone, so the cut is ambiguous.
Solids of revolution
Per unit density, with inner radius (zero when filled):
Each slice is an annulus: about the axis and about its diameter, moved to the reference plane by the parallel-axis theorem. excludes the end faces.
-
Each half of the profile is integrated from its own end in , where is the normalized distance from that end. The substitution removes the singularity of a blunt tip’s surface integrand, and measuring from the end keeps the tip exact. The integrals use
hpr_core::quadrature(Quadrature) at a relative tolerance of 1e-12. -
A wall of thickness is the part of the solid within of the outer surface (component geometry and mass properties). So is measured normal to the surface, which is how molded and laid-up shells are made. Its inner radius is the lower envelope of circles of radius on the profile, over :
with the minimum taken over the where . The surface is the profile over its own length, ends included, with no extension past a cut end.
- The envelope is exact for any continuous profile. Take a point above the lower half of some surface point’s circle. The profile crosses its height closer than , so it is in the wall anyway.
- Where the surface meets the end plane at an obtuse angle inside the wall, the rim’s circle
rounds the wall’s inner corner. That is the small end of a transition, or the base of a
bulged ogive.
- A square cut would add a sliver of of section per unit rim length, with the surface’s angle to the axis. That is 3.1e-4 at 7°.
- On steep ends it matters, here with . Cut square, a transition of radii 27 to 49 mm (1.1 to 1.9 in) over 15 mm (0.59 in, 56°) gains 1.28% of its wall. One of radii 20 to 37.3 mm (0.79 to 1.47 in) over 10 mm (0.39 in, 60°) gains 2.29%. Both are the sliver integrated around the rim, over this model’s wall; no test pins them.
- OpenRocket weighs its one probe transition, a shallow cone, heavier than this model by 2.85e-6 (relative). That is the sliver a square cut adds (the probes). No steep end or real part has been compared.
- The sliver grows without bound only as the end turns vertical. There, a square cut (made by extending the surface along its tangent) closes the end with a disc of thickness .
- The minimum comes from a 32-point scan, a golden-section search, and the two end points, which the search only approaches from inside.
- The hollow is integrated separately. It is split where reaches zero and where the nearest surface point moves between the lateral surface and a rim, since both are kinks.
- [TD] doesn’t say how OpenRocket measures thickness, and [CR] measures it radially. A radial wall differs from this one by a factor in wall volume, 1.4% for a cone three calibers (base diameters) long.
- OpenRocket’s masses follow a wall measured across each station, inner radius . On a straight profile that is this model’s wall: a probe cone of that shape weighs the same within 5.1e-7 (the probes). On a curved one they differ: this model weighs up to 0.48% more on three short elliptical catalog noses (the masses against OpenRocket). Whether that is the probe ogive’s 3.93e-5 is not traced.
Verification
- Closed forms (
shapes::tests::nose_volumes_match_closed_forms, 1e-10 relative):-
Every shape’s filled volume and centroid.
-
Cone and ogive: , , and from arc integrals.
-
Half spheroid: , and for prolate and oblate cases.
-
Power series: , , and the paraboloid’s .
-
Power series and parabolic series: and too; parabolic series by polynomial integrals.
-
Haack: and
integrated in .
-
Loft’s tangent-ogive value, , gives 6.7509e-4 m³ (a lesson from Loft).
-
- mpmath references (
solids::tests::filled_solids_match_the_mpmath_references):- 22 noses and transitions of every family, in both directions, clipped and not.
- All seven quantities, to 1e-12 relative (worst measured 2.4e-14).
- Reference:
validation/fixtures/design/shape-integrals.json, fromvalidation/oracles/design/shapes.py(40-digit tanh-sinh on the defining formulas, Haack in , with no code shared with Rust). - These references alone check the filled shapes’ moments of inertia, both Haack areas, and the wetted areas of two series: power () and parabolic.
- Walls against mpmath (
solids::tests::walls_match_the_mpmath_references):- 20 walls: 11 noses of every family, and 9 transitions both ways, including unclipped blunt ends and clipped ones.
- Volume, centroid and both moments to 1e-10 relative (worst measured 5.9e-12).
- Reference:
validation/fixtures/design/wall-integrals.json, fromvalidation/oracles/design/walls.py, which shares no code with Rust.- It finds the envelope from the roots of its derivative, bracketed from the window edges and solved by bisection.
- It splits the integrals at the kinks it finds, and requires every hollow integral’s error estimate below 1e-18.
- Reviews found three faults this test now pins:
- Blunt transition ends failed to converge, or were 5e-6 low, until the end points became candidates.
- The first version’s extensions along finite end tangents made wall mass jump by 8.3% between shapes whose end slopes rounded to finite and to infinite.
- The oracle itself first missed minima next to the window edge and left kinks unsplit.
- Walls by hand:
- A conical wall is the cone minus the same cone moved aft by : volume, centroid and
both moments by hand, to 1e-9. A cone so thick that its hollow is 3.8 mm (0.15 in) long matches the same
formula to 1e-12 (
a_nearly_filled_cone_matches_the_offset_cone). - A conical transition’s wall is the square-cut frustum shell less the fore rim’s sliver, in
polar coordinates about the rim. Its mass, centroid and sliver section match to 1e-10
(
mass::tests::hollow_transition_and_freeform_fin_cg_are_exact_centroids). - A tangent-ogive wall is bounded by the concentric arc of radius : volume by hand, to 1e-10.
- A tube matches the hollow-cylinder formulas.
- A wall thicker than the body fills it, and a thin wall’s volume tends to .
- A conical wall is the cone minus the same cone moved aft by : volume, centroid and
both moments by hand, to 1e-9. A cone so thick that its hollow is 3.8 mm (0.15 in) long matches the same
formula to 1e-12 (
- Each profile ends at 0 and , slopes match central differences, and parameters out of
range are errors (a lesson from Loft).
- Haack tips use and a Taylor series for below , so the tip slope is , never NaN.
- Ogive radius ratios up to 1e12 give the cone. A power series at the minimum exponent, 0.05,
matches its closed-form volume as a nose and as transitions both ways
(
extreme_parameters_stay_accurate_or_fail_loudly). - Unknown fields in a shape or wall are rejected.
- Transitions hit both radii and are monotone both ways, clipped or not (a lesson from Loft). Bulged ogives are excluded, because their profile is deliberately not monotone.
- Power-series exponents below 0.05 are rejected. Blunter profiles approach a flat face the integrals can’t resolve, and unclipped transitions below about 0.038 fail to converge.