SUMMARY
Novel fiber-reinforced composite materials enable the potential for high performance flywheel energy storage systems. Designing flywheel components capable of realizing this potential is a challenge that heavily relies on simulation driven design. The work presents the Abaqus finite element analysis (FEA) techniques used to successfully design a flywheel assembly capable of operating at more than 35,000 rpm. Strength and compliance behaviors had to be balanced under the rotational inertial loads. This was accomplished by applying a multi-level design optimization approach using the Tosca framework. Rotordynamic analyses of the optimized system were then conducted to determine critical speeds of the system so performance problems associated with these resonances could be avoided when subsequent spin testing is performed.
Rotordynamics Design Analysis of a High-Speed Composite Flywheel Hub
Fiber-reinforced composites can outperform metals in flywheel energy storage — higher energy density, better margin of safety. The catch is the part nobody photographs for the brochure: the hub that joins a composite rotor to a steel shaft and holds the whole thing together past 35,000 rpm.
Spin a disk that fast and physics stops being polite. Below, we walk through how Quartus designed that hub using simulation-driven design in Abaqus FEA — work presented at the 2026 Simulia Americas Users Conference and sponsored by the Office of Naval Research.
Why composite flywheels are worth the trouble
A flywheel stores kinetic energy while it spins. Both the energy it holds and the stress it carries scale the same way — with density, radius squared, and speed squared (E ∝ ρr²ω²). That single relationship is the whole design tension in one line: the features that make a flywheel store more energy are the same ones trying to tear it apart.
So realizing the composite’s potential isn’t a materials win — it’s an optimization problem with two requirements pulling against each other:
- Strength under steady spin — managing quasi-static stresses and deflections
- Rotordynamics — managing structural vibrations and resonances across the operating range
You don’t get to optimize one without paying attention to the other.
The hard part is the hub
The rotor grows as it spins. The hub has to grow with it — matching radial expansion at the rotor inner diameter without letting go and without becoming the weak link.
For a 140 mm ID rotor, that means roughly 1.5 mm of radial displacement capability, which translates to an elastic strain capacity ranging from zero to about 2%. On top of that, the hub has to survive high rotational stress and be engineered so the rotor fails before the hub does — a deliberate rotor-fails-first hierarchy for safety and reliability.
Hitting a displacement target that large, at that size and speed, with a metal part, is not a detail. It’s the design.
One model, every analysis
Rotating-body problems usually get split into separate worlds. Rotordynamics gets a simplified “stick model.” Spin stresses get an axisymmetric disk model. Two tools, two teams, two sets of assumptions that don’t quite talk to each other.
We didn’t do that. A single high-fidelity 3D model handled both the full stress state and the rotordynamics of the complete assembly — gyroscopic effects, complex mode shapes, multi-material contact, all of it. One source of truth instead of two approximations you have to reconcile later.
Analysis-driven hub optimization
Feasibility studies pointed to titanium (Ti-6Al-4V) as the best practical choice, joined to the rotor in a bonded assembly. Trade studies pointed to a spoked hub — good for managing radial mass and hoop stress. What they didn’t hand us was the spoke geometry. That stayed stubbornly elusive.
So we optimized for it, using Tosca to drive an Abaqus FEA of every candidate design, with three objectives that had to hold simultaneously: stress under 1 GPa, manufacturability, and acceptable stiffness and dynamics. Rather than one monolithic solve, we applied optimization techniques in sequence so each step built insight for the next.
The starting-point design balanced inertial load against compliance — and was badly over-stressed. A useful failure: it told us exactly where the geometry had to work harder.
Sizing optimization tuned spoke thickness first — a fast, robust way to get the bulk cross-section right. Shape optimization then reshaped the spoke surfaces, and the result was genuinely non-intuitive: features no one would have drawn by hand, balancing flexural compliance against stress in a way only the solver found.
Finally, interactive DFM turned the optimized shape into something a shop can actually cut. Smaller features mean tighter tolerances, higher cost, and higher stress — the difference between EDM and conventional CNC turning. This is where performance meets a purchase order, and where a good engineer trades a little of one for a lot of the other.
Does it actually hold together at speed?
Verification, not vibes.
Under quasi-static spin at full speed, the integrated flywheel met both its at-speed displacement requirement and its strength requirement — the hub grows with the rotor, and nothing goes past its limit.
For rotordynamics, we built a Campbell diagram transformed into the fixed reference frame — where the whirl, axial, and torsion behavior is actually intuitive to read — and swept it across the full operating range. The verdict: acceptable dynamics, with no critical speeds of concern inside the operating envelope.
What we’d tell the next team
A few things this project reinforced, in case you’re staring down a rotating-machinery problem of your own:
- Geometry dictates the reference frame. A symmetric rotor can be solved in the fixed frame; an unsymmetric one forces the rotating frame. That choice ripples through everything.
- Symmetry discipline extends to the mesh. Get sloppy meshing a symmetric part and you’ll manufacture asymmetry the physics never had.
- Fixed-frame results are the ones humans can read. Rotating-frame output has to be transformed before it means anything to a reviewer — so we modified the Abaqus/CAE Campbell diagram plug-in to report fixed-frame behavior with damping, and to track and classify modes cleanly.
The takeaway
Rapid hub design was possible because we stacked several optimization approaches in the right order — and because the physics of rotating bodies threw enough unique curveballs that guessing was never an option. A successful design here simply could not have happened without reliable FEA.
That’s the through-line of how we work: model the full system to the point of trust, then build. If you’ve got a high-speed rotating system — or any structural problem where the “seemingly impossible” requirement is the whole point — that’s exactly the kind of work we like.
This work was sponsored by the Office of Naval Research through contract No. N00014-23-C-1010. The views expressed are those of the authors and do not reflect the official policy or position of the Department of Defense or the U.S. Government. Distribution Statement A — approved for public release; distribution unlimited.