Skip to content

Mechanical Engineering.CAE Pre-processing.FEA Meshing.

I prepare finite element models and meshes for durability analyses in the automotive industry, from clean geometry in HyperMesh to a simulation-ready model.

01 · About me

I'm in the final stage of a Mechanical Engineering degree at UNIFEG and, since May 2025, I've been working with CAE at TC Projetos, preparing finite element models for major automotive brands. My focus is pre-processing: cleaning up geometry, building 2D/3D meshes in Altair HyperMesh and making sure every element meets quality standards. From the first geometry cut to the last quality check, I treat every mesh as the foundation of a reliable result.

projects in the portfolio
3
automotive brands served
3
tools I use
7
graduation year
2026

Mechanical Engineering

Bachelor's degree at UNIFEG (2022–2026), in its final stage, focused on CAE pre-processing and finite element meshing.

Mechanical Engineering: Bachelor's degree at UNIFEG (2022–2026), in its final stage, focused on CAE pre-processing and finite element meshing.

Journey

  1. UNIFEG

    Started the Mechanical Engineering degree.

  2. TC Projetosnow

    CAE pre-processing internship since May 2025.

  3. Graduation

    Mechanical Engineering, with a thesis on an AI-generated swingarm evaluated by finite elements against the Yamaha YZF-R1 one.

✓ assembly complete · 3 parts, 2 bolts

02 · Process

From sketch to validated part, each stage sets up the next. Drag the control and follow the part through each stage; in the simulation, pull the tip of the beam to apply load.

resultsshown in the Simulation stage

F
— kN
δ = FL³/3EI
— mm
σ = Mc/I
— MPa
SF = σy/σ
—

ASTM A36 steel · E = 200 GPa · σy = 250 MPa · bar 20×40 mm · L = 500 mm · deformation ×20

01 · Idea

It all starts with a sketch

A real problem, a notebook and lots of questions. The points are still loose: time to understand what needs to exist.

  • Sketching
  • Requirements
  • Research

03 · Mesh Quality

In pre-processing, every element goes through quality checks before reaching the simulation. One of the nodes is misplaced: drag the inner nodes (or use the arrow keys) and watch the criteria react.

E1E2E3E4E5E6E7E8E9

click an element to inspect it

6/9 passed

1 failed element: this model can't go to simulation.

Element E3 · QUAD4

failed
CriterionValueLimit
Aspect ratio1.35< 5
Skew20.8°< 60°
Jacobian0.59> 0.6
Min angle65.4°> 45°
Max angle138.4°< 135°

idealcould be betterwarningfailedWarpage does not apply here: the patch is flat (2D). Typical pre-processing limits; each project defines its own.

drag to rotate · hover the nodes

Element types

CHEXA (hex8) · 3D solid

The most accurate and efficient solid, but it requires splitting the geometry into blocks to mesh by sweeping or mapping. More work, better result.

1st Straight edges: displacement varies linearly between nodes.

nodes
8
DOF per node
3
element DOF
24
accuracy
ease of meshing
computational cost

qualitative comparison

04 · Thesis

To what extent can generative AI, combined with HyperMesh and finite element analysis, deliver a lighter swingarm without compromising its stiffness and structural strength?

Mechanical Engineering · UNIFEG · 2026 · co-authored with Gustavo Henrique Rezenda da Rocha · advisor: Prof. Valdoel da Silva Junior

drag to compare
AI version: Displacement
Yamaha reference: Displacement
Yamaha referenceAI version

Total displacement under 15 kN at the eyelet, at 60°: 0.231 mm for the reference and 0.262 mm for the AI version.

−9.3%

mass

19.27 → 17.48 kg

−11.6%

eyelet stiffness

64.8 → 57.3 kN/mm

−7.9%

peak von Mises

104 → 96 MPa

1.6 → 5.2

safety factor

yield, driven by the alloy

Workflow

  1. 1Meshy3D shape from a text prompt
  2. 2Fittingpivot, axle and eyelet matched to the Yamaha
  3. 3HyperMeshclean-up of the AI mesh and C3D4 solid
  4. 4Abaqus15 kN at the suspension eyelet, at 60°
  5. 5Verificationindependent linear solver in Python
Test boundary conditions: wheel axle clamped, pivot free and 15 kN at the link eyelet at 60°

Mesh quality

MetricCriterionYamahaAI
C3D4 elements261,684252,507
Nodes54,25153,412
Median edge6.1 mm6.0 mm
Aspect ratio≤ 5.099.995%99.98%
Tet collapse> 0.2100%100%

Same size and same quality in both models: that is what makes the comparison fair. In C3D4 the Jacobian is constant and equal to 1.

Conclusion

Generative AI quickly delivered a lighter swingarm with the same mounting points, but no efficiency gain: it removed material and lost stiffness in the same proportion. The limit is stiffness, not strength. To truly optimize, the AI shape should feed a load-driven topology optimization.

Next steps: all five load cases, C3D10 with a convergence study, fatigue analysis and a bench prototype.

05 · Portfolio

CAE pre-processing and finite element meshing, from my thesis to the automotive industry.

06 · Contact

Guaxupé, MG · open to opportunities in CAE and pre-processing

Subject

Email

contato@kauasales.com
Prefer your email app?

Kauã Sales: Choose where to start

01 / CAE

CAE pre-processing and finite element meshing for the automotive industry.

02 / DEV

Websites and web apps with Next.js, React and TypeScript, from layout to deploy.