TUCer / Diploma thesis · Technical University of Crete

CFRP Composite Wheel.

A single-piece, tubeless carbon-fibre wheel replacing a two-piece glued rim — from RVE-based composite modelling through a three-mold manufacturing evolution to instrumented physical validation.

Photorealistic render of the finished single-piece carbon-fibre wheel with the tire fitted, showing the woven twill sidewall and the machined aluminium hub centre

FIG. 01 — CFRP composite wheel, TUC diploma thesis

Context: Diploma Thesis — TUCer
Assembly mass: -19.2%
FEA validation: Tsai–Wu RF 1.54
Recognition: Best Thesis Award

01

Design

The old wheel was a two-piece rim glued together, requiring an inner tube and concentrating load in the adhesive joint. The new design is single-piece and tubeless: Michelin's 3.00 B rim geometry (chosen over the alternative for a lower rolling-resistance coefficient at the competition's 35 km/h average speed) retains the tire at operating pressure with no tube at all.

The sidewall replaces spokes with a foam-core CFRP sandwich — evening load distribution, cutting aerodynamic drag, and resisting bending and buckling. The hub is deliberately simple: a two-piece triangular center bolts to a permanent adapter left on the vehicle, which carries the brake disc and bearings. Removing the wheel no longer disturbs the brake caliper.

Exploded CAD render of the sidewall, separated into three discs: an outer carbon-fibre skin, a foam core, and an inner carbon-fibre skin — FIG. 02 — Sidewall sandwich: CFRP skin / foam core / CFRP skin

02

Materials & FEA

META contour plot over the wheel showing the Tsai–Wu reserve factor, minimum of all layers, with the scale banded from red below 1.54 up to over 10 and the minimum element annotated at 1.611 — FIG. 03 — Ply-level Tsai–Wu reserve factor, min of all layers, subcase 4 (min element 1.611). The governing minimum across all five subcases is 1.54, in subcase 3 — the number quoted sitewide.

The laminate — twill 2×2 T300 carbon fibre in Resoltech 1050/1058s infusion epoxy — was characterized by RVE (Representative Volume Element) homogenization in ANSA/EPILYSIS: six strain-controlled load cases on a periodic microstructure model, yielding orthotropic properties (E1 ≈ E2 ≈ 52.4 GPa for the balanced weave, fibre volume fraction 0.454) fed into ply-level FEA with per-ply Tsai–Wu tracking.

Five load cases were simulated per candidate — four static (weight, cornering, braking, internal pressure) plus a transient pothole impact — and 9 candidate layups were scored on margin, mass and stiffness (full breakdown below). The winner, a 20 mm-core layup at 635 g, held a minimum Tsai–Wu reserve factor of 1.54 across all five subcases; thinner or coreless variants either failed outright or scored far lower.

Nine-candidate layup study

Any candidate with a reserve factor below 1 in any subcase is rejected outright and scores 0%; the rest are ranked by S = 3·max(RF) / (m³·max(u)^1.5), with mass m in grams and displacement u in millimetres. The exponents show the function was designed rather than fitted: a 10% mass increase costs about 27% of the score, and that same 27% needs roughly 14% more displacement.

Code	Description	Mass (g)	Min RF	Max disp. (mm)	Score
R6W6	20mm core wall	635	1.540	6.75	100%
R6W6	20mm core wall + 2mm core rim	776	2.479	5.96	93%
R5W6	20mm core wall	590	1.203	7.47	85%
R6W8	20mm core wall	659	1.211	6.45	76%
R6W6*	20mm core wall + LR	660	1.592	7.70	75%
R6W6	15mm core wall	596	1.521	10.67	61%
R6W6	10mm core wall	557	1.485	16.45	39%
R6W6	No core	478	0.147	363.60	0%
R4W6	20mm core wall	545	0.894	20.84	0%

Sorted by score, the function's own ranking. The bottom two failed outright — reserve factor below 1 in at least one subcase — and score zero by definition; they are published, not omitted. Five load cases per candidate: four static (weight, cornering, braking, internal pressure) plus a transient pothole impact, run as subcases 1–5. Subcase 5 is confirmed as the impact case; the source does not establish which static case is subcase 1, 2, 3 or 4 — TODO: Mike — confirm which subcase is which load case. FIG. 03 above plots subcase 4 (min element 1.611); the table's governing minimum, RF 1.540, falls in subcase 3. R6W6*'s LR suffix is undefined in the thesis text — TODO: Mike — what does LR expand to in the R6W6* row?

View full per-subcase data

Code	Description	Mass (g)	Rejected	Reserve factor (Tsai–Wu) by subcase					Max displacement (mm) by subcase					Score
Code	Description	Mass (g)	Rejected	S1	S2	S3	S4	S5	S1	S2	S3	S4	S5	Score
R6W6	20mm core wall	635	No	2.033	1.863	1.540	1.611	2.036	2.166	2.289	6.748	5.667	2.152	100%
R6W6	20mm core wall + 2mm core rim	776	No	2.869	2.841	2.479	2.752	2.875	1.446	1.568	5.963	4.253	1.440	93%
R5W6	20mm core wall	590	No	1.613	1.466	1.203	1.313	1.614	2.685	2.863	7.467	6.272	2.686	85%
R6W8	20mm core wall	659	No	1.618	1.470	1.211	1.502	1.620	2.489	2.663	6.446	5.074	2.485	76%
R6W6*	20mm core wall + LR	660	No	2.084	1.911	1.592	1.632	2.087	1.883	1.968	7.704	6.780	1.868	75%
R6W6	15mm core wall	596	No	2.026	1.857	1.521	1.594	2.027	2.445	2.576	10.670	8.862	2.471	61%
R6W6	10mm core wall	557	No	2.014	1.848	1.485	1.548	2.019	3.048	3.200	16.450	13.130	3.053	39%
R6W6	No core	478	Yes	1.566	1.379	0.147	0.180	0.899	39.330	43.850	363.60	280.10	77.50	0%
R4W6	20mm core wall	545	Yes	1.222	1.101	0.894	1.063	1.223	10.434	12.039	20.837	12.792	3.586	0%

Source: analysis.tex, TUCer diploma thesis, table tab:results_table and its scoring-function block, verbatim.

03

Manufacturing

Tooling went through three iterations. The first mold, machined from MDF, proved the process was feasible in the lab but suffered from moisture sensitivity, heat deformation, and resin waste. The second, machined from UHMWPE on a CNC router, released cleanly without release agents but couldn't be fully sealed for the rim's more complex geometry. The third and final rim mold — a two-piece gelcoat–fiberglass–vinylester composite cast from a CNC-machined foam positive — produced three rims with no deformation or wear.

Parts were built by vacuum infusion (VARTM): 18 fabric pieces for the rim and 6 plus a core for the sidewall, laid up with the mold's smooth face at the tire interface and its peel-ply rough face used as the bonding surface, needing no sanding. Each part cured for 16 hours at 60°C in a purpose-built oven — see the curing oven case study — then the two halves were bonded with epoxy and post-cured.

Carbon twill fabric stacked inside the orange-and-green two-piece composite rim mold, the weave visible around the full circumference — FIG. 04 — Carbon twill laid up in the rim mold

04

Results & validation

The old wire-spoked wheel with an inner tube standing beside the new single-piece tubeless carbon-fibre wheel, both tires fitted — FIG. 05 — Old (A) vs. new (B) wheel

The finished wheel assembly weighs 2841 g against the old wheel's 3515 g — a 19.2% reduction (674 g), with the wheel structure alone down 37.9% and the hub down 48.6%. At vehicle level, that's roughly 2.7 kg saved across four wheels.

Physical validation ran on a custom-built instrumented test bench: a pressure test held +20% over operating pressure with no cracks, a fatigue test ran 40 kg at max speed for a duration equal to three race lengths with no damage, and comparative energy testing measured 44–59% lower specific energy (Wh/km) for the new wheel under load, with rotational inertia consistently 11–47% lower across all four test scenarios.

Awarded Best Thesis — Technical University of Crete