How We Cut Multi-Stage Rotor Simulation Times from 6.5 Hours to 13 Minutes

If you work on turbomachinery simulations, you know how painful full 360-degree rotor models can be. They drain your workstation's hardware, chew through RAM, and force you to spend hours waiting for a single solve just to capture accurate local stresses and vibrational characteristics.

While working on the fan rotor for the Türkiye’s first national turbofan engine TEI-TF6000, we wanted to see if we could completely bypass this computational bottleneck. By using the Multi-Stage Cyclic Symmetry feature in ANSYS Mechanical 24.2, we managed to drop our solve times from 6.5 hours down to just 13 minutes without losing structural accuracy.

Here is a breakdown of how the methodology works and the actual data from our validation.

The Problem with Non-Matching Stages

Standard cyclic symmetry is straightforward. For a single rotor disk, you mesh one sector, apply cyclic boundary conditions, and the solver handles the rest. But multi-stage assemblies are a different story because blade counts rarely match across adjacent stages.

The TEI-TF6000 fan rotor is a perfect example of this:

• Stage 1: Has 17 blades, requiring a 21.17° sector.

• Stage 2: Has 36 blades, requiring a 10° sector.

Because 17 and 36 do not share a common geometric denominator, the sector boundaries do not line up. Normally, this leaves you with two bad options: write complex, heavy constraint equations at the interface to force them together, or bite the bullet and mesh the full 360-degree assembly. To ensure we captured the correct multi-axial stress fields and pre-stressed modal behaviors, we typically had to go with the massive 360-degree route.

Connecting the Interfaces Mathematically

The multi-stage cyclic symmetry feature solves this non-matching mesh problem mathematically. Instead of forcing a hard geometric mesh match at the stage interface, the solver transforms the displacement and force vectors into a cyclic symmetry coordinate space using a real-valued Fourier matrix approach.

To verify the accuracy of this tool, we ran a direct comparison between two models:

1. The Full 360° Model: The baseline assembly including the full bladed rotors, bolts, and nuts.

2. The Multi-Stage Sector Model: A streamlined model where each component was reduced to its minimum sector angle (21.17° for Stage 1, 10° for Stage 2, and down to 1° for axial-symmetric zones).

We used HEX8 elements for the disk sections and TET10 elements for the complex blade geometries. The materials were nickel-based superalloys for the rotating parts and stainless steel for the bolts, with frictional contact defined at the joints.

Both models ran through the exact same three-step loading sequence: bolt pre-tension, centrifugal loads from operational speed, and finally combined gas pressures and thermal gradients, followed by a pre-stressed modal analysis.

The Numbers: 360° vs. Multi-Stage Sector

The resource savings from dropping the 360-degree mesh are substantial:

• Node Count: The full 360° model ballooned to 2,023,390 nodes. The sector model needed only 88,774 nodes—nearly a 23x reduction.

• Solve Time: The full assembly took 6.5 hours to complete on our workstation. The multi-stage sector model finished the exact same analysis in 13 minutes.

• File Size: The 360° results devoured 57.5 GB of disk space, while the sector model required just 6.7 GB.

Is it Accurate?

Speed means nothing if the solver misses critical hot spots. When we compared the equivalent (Von Mises) stresses, total deformations, and natural frequencies side-by-side, the correlation was excellent.

• Stress & Deformation: The stress concentrations across the blades, disk fillets, and bolted joints matched up closely. Localized variance in equivalent stress stayed within a tight 2% to 5% margin, and total deformation deviations hovered between 1% and 11%.

• Vibrational Dynamics: For the first natural frequency and modeshape, the full model resolved at 267.2 Hz compared to 270.67 Hz on the sector model. Across the primary modeshapes, the overall frequency variance stayed reliably between 1% and 11%.

What This Means for the Design Loop

Full 360-degree simulations will always have their place for final structural verification and certification. But during the iterative design phase—when you are constantly tweaking geometries, shifting blade counts, or running optimization loops—waiting 6.5 hours per run is a massive bottleneck.

Bringing that turnaround time down to 13 minutes completely changes how we work. It allows a design team to evaluate dozens of structural iterations in a single day rather than a single week, bridging the gap between high-fidelity physics and rapid design execution.

Reference: Kortağ, U., Akdaş, M., & Kutlu, O. (2025). Fan Rotorun Çok Kademeli Döngüsel Simetri Yaklaşımıyla Gerilme ve Titreşim Analizi. XIII. Ulusal Uçak, Havacılık ve Uzay Mühendisliği Kurultayı, Eskişehir, Türkiye.