A Comparison between Two Fabrication Methods in LUCIA Model Production

Group members: Antonija, Katherine, Xinyin (Sing), Wen-Yi

LUCIA is a low-cost tool for training cervical cancer screening, diagnosis, and treatment [1]. We aim to produce two sets of four cervixes using two methods and assess them based on quality, cost, efficiency, and scalability. One method is directly making cervixes via 3D printing (simplified as 3D printing), and the other is creating a positive mold via 3D printing, a negative mold using silicon, and casting using polyurethane (simplified as molding and casting). We expected 3D printing to produce consistent products with good qualities because 3D printing is straightforward with more predictability. Timewise, 3D printing would be less favorable in mass production. In molding and casting, the initial creation of the positive and negative mold takes time. Once they are made, the polyurethane casting is fast. We chose vasculature models 1I, 1J, 2I, and 2J. 

We downloaded STL files from Canvas. Then, we imported them into TinkerLab, duplicating the four models to create a positive mold, and exporting the four separate models and the positive mold into one STL file. We processed the file in Bambu Slicer and printed them. The initial layer height was 0.2 mm. However, another group reported that 0.2 mm created artificial details on the surface. A spaghetti defect occurred minutes after we started printing. Therefore, we adjusted the height to 0.09 mm and reprinted. Printing took 9 hours and 28 minutes (Fig. 1).

Figure 1. File processed by Bambu Slicer.

The four separate cervixes were successful. However, the positive mold had a few problems. First, we designed slanted walls for the ease of demolding. However, there was a gap between a slant and a wall (Fig. 2). Second, the dimension of the box (85mm x 85mm x 35mm) was small, so the cervixes were close to each other, creating challenges for demolding (Fig. 2). 

Figure 2. Four separate cervixes and the initial version of the positive mold.

Consequently, we printed version 2 of the positive mold with 90mm x 90mm x 35mm. We accidentally enlarged the cervixes (Fig. 3a). After correcting the mistake, we printed version 3. Interestingly, those cervixes had artificial lines (Fig. 3c). To improve the quality and ensure a better chance of success, we lowered the printing speed and adjusted the layer height to 0.08mm. Version 4 was successful (Fig. 3d). One slant is shorter than the wall, but it does not affect the silicon mold.

Figure 3. Positive mold version 2 (a), 3 (b, c), and 4 (d).

Creating silicone-negative mold

We combined 110 mL silicon part A with 110 mL silicon part B, whisked them, and poured them along box edge (Fig. 4). We let it set for 7.5 hours and demolded using flathead tools to introduce air (Video 1).

Figure 4. Creating silicone-negative mold.

IMG_1748 2

Video 1. Demolding.

Polyurethane Casting 

We used Smooth-Cast 300 with a 3-minute working time. We mixed 40ml each of part A and part B and poured them into the silicon mold (Video 2), let it solidify for 10 minutes, and demolded it. 

a1023035-5a18-44ce-bd32-006608effc8e

Video 2. Timelapse video of polyurethane casting.

We observed traces of bubbles on 1I and nonintact numbers and letters (Fig. 5).

Figure 5. Bubbles and missing bottom of 2 and J.

To eliminate bubbles, we tried to apply the mold spray, which did not alleviate the problem (Fig. 6).

Figure 6. Products from using mold spray.

In the final attempt, we used a vibration machine during solidifying (Fig. 7). The labels came out intact, but 1I still had dents on top (Fig. 8).

Figure 7. Vibration machine.

Figure 8. Polyurethane product.

Post-processing

We sanded the polyurethane products, colored all the cervixes with acrylic colors (on different days), drew details of blood vessels using thin brushes or a red pen (Fig. 9), and applied a clear coat. Interestingly, the acrylic colors darkened, creating challenges for color matching. 

Figure 9. Draw fine details.

Final Products

The final products resemble the originals (Fig. 10c). There are slight color deviations, and the stokes and traces of 3D printing are visible. Spray paints could alleviate the problem, but spray paints with colors matching the originals were not available. We also compared pen-drawing versus brushing. 1J and 2I used a red pen, and 1I, 2J used a brush. The pen was easy to control and produced consistent thin lines. Brushes produced thicker lines, but the acrylic color is more vivid (Fig. 10a, b).

Figure 10. Final products. (a) from 3D printing. (b) from molding and casting. (c) Group picture with originals in the middle.

The acrylic paints filled in most of the dents of 1I from molding and casting (Fig. 11a). “2J” from the molding and casting is a bit rough (Fig. 11b).

Figure 11. Detailed images of 1I (a) and 2J (b).

We did paired t-tests and concluded that the 3D-printed cervixes are closer in dimensions to the originals (Table 1).

Finally, we assessed the two methods based on cost, ease of production, quality, and scalability. Please refer to our technical report for more details. To summarize the key points:

  1. Molding and casting costs more for the initial set, but both methods have similar costs in subsequent sets (around $75-76 each).
  2. Molding and casting has a cheaper material costs of around $8 each, while 3D printing costs $12 per set plus the cost of the 3D-printer.
  3. Labor cost accounts for around 89% of the total cost for molding and casting and 83% for 3D printing.
  4. 3D printing is straightforward, requiring fewer skills.
  5. The rate-limiting factor for 3D printing is printing time, and that for molding and casting is post-processing. The production rates based on current assumptions are similar (counter to our expectations). 
  6. With improved post-processing workflow, molding and casting may be more suitable for mass production.
  7. 3D-printed cervixes have consistent and better qualities.

The decision on methods depends on various factors, such as the desired qualities, the local cost of labor, and material properties.

Print Friendly, PDF & Email