top of page
  • LinkedIn
Screenshot 2026-01-08 at 13.11.28.png

FDM Variation in Bearing Hole Diameters

Anchor 1
Objectives
Identify the best printer settings to minimize print time and maximize bearing hole quality and consistency.
Outcomes and Contributions
Bearing holes must be a specific diameter to ensure bearings are installed correctly. Bearing holes that are too small can cause the bearings to compress, preventing the axles from sliding smoothly through. Bearing holes that are too large can cause the bearing to slip out. As part of a team, I devised a fractional factorial experiment to test 4 factors to find the ideal 3D printer settings. In addition, I collected data by measuring 3 batches of 50 3D printed bearing holes to determine if the printer we were using is in control. The control chart generated from this experiment indicated that the printer was, in fact, out of control. It also helped identify a significant mean shift that occurred when a window in the room where the 3D printer was in was left open, causing the temperature of the room to decrease. The fractional factorial experiment generated an equation that was used to find the optimal settings. Reprinting samples with these settings resulted in reduced variance, proving that the fractional factorial experiment resulted in improved quality.
Technical Details

All samples were printed using a Bambu A1. The fractional factorial experiment was based on a 24 full factorial experiment. The factors were XY acceleration, resolution (defined in the Bambu slicer as the number of points defining an arc), external cooling, and filament material. Making the defining factor I = ABCD, we found aliased factors and reduced the experiment to a fractional factorial 24-1 experiment. Since we had 5 replicates for each run, totaling 45 minutes of print time, the fractional factorial experiment significantly reduced the print time to complete all required experiments. Using this setup, we created a linear regression. After eliminating insignificant terms, the resulting linear regression was used to find the optimal print settings. Further details can be found in the final presentation slide deck.

Fig. 1 - 1 Batch of Parts

Fig. 1 - 1 Batch of Parts

Fig. 2 - Control Chart Indicating Mean Shift

Fig. 2 - Control Chart Indicating Mean Shift

Fig. 3 - Response Surfaces Generated from Experiment

Fig. 3 - Response Surfaces Generated from Experiment

bottom of page