Key Takeaways
- FEA works best alongside engineering fundamentals, not instead of them
- PLA is brittle with no plastic plateau, so linear elastic modeling is appropriate
- Hand calculations and FEA should match. If not, something's wrong
- Beam theory identified the critical load path before I opened the FEA software
- Thickness matters way more than width for bending stiffness (cubic vs linear)
Arm executing movements with real-time inverse kinematics. Simulated in Webots, mirrored on hardware.
Overview
Designed from scratch in SolidWorks, 3D printed, controlled via Arduino with real-time inverse kinematics. Simulation pipeline couples Webots with ROS2 to test trajectories before running on hardware. FEA in ANSYS drove gripper redesign, validated against classical beam theory.
Exploded View
FEA Study: Finding the Weak Link
Small deflections at the gripper become big positioning errors at the payload. Minimizing compliance matters a lot here.
Why Linear Elastic for PLA?
PLA is basically brittle. It fails around 2.9% strain and 45 MPa with no real plastic plateau. It doesn't yield and strain-harden like metals. Since I'm looking at stiffness and deflection (not failure), a linear elastic model with the measured modulus is the right call.
| Property | Value |
|---|---|
| Young's Modulus | 2.13 GPa |
| Poisson's Ratio | 0.36 |
| Density | 1240 kg/m³ |
| Tensile Strength | ~45 MPa |
Mesh and Loading
Imported CAD into ANSYS Discovery, simplified geometry (removed cosmetic features, kept load-bearing stuff). MultiZone meshing for hex-dominant elements, target >80% element quality. Fixed the wrist mount, applied 2N to each finger (4N total, ~400g payload).
Original CAD
Simplified for FEA
Initial Results
Strain concentrated in a short linkage member, not the fingertips. That linkage was the compliance bottleneck, so improving fingers would've been pointless.
Peak strain in the motion-transmission linkage, not where I expected
Beam Theory Analysis
Treated the linkage as a cantilever. Second moment of area goes with thickness cubed but width linear:
Thickness increase is way more effective than width for bending stiffness. But real constraints: the linkage rotates on pins, limiting thickness. So I pushed both where I could.
| Dimension | Original | Optimized |
|---|---|---|
| Length | 37 mm | 37 mm |
| Width | 6.0 mm | 9.3 mm |
| Thickness | 2.0 mm | 3.0 mm |
Validation
Before rerunning FEA, I calculated expected strain analytically. Cantilever with 2N load, optimized dimensions:
After redesign: strain reduced, load distributes more evenly
Bottom Line
- Used FEA to find the problem, beam theory to understand it, then validated the fix with both
- Simulation should support first-principles reasoning, not replace it