Optimizing 3D Prints: Results: Optimum Configurations for 3D Printing (Part 2)

Results: Optimum Configurations for 3D Printing

The summary of optimal settings obtained in the case of Experiment-1 and Experiment-2 can be found in Table 2 and Table 3 respectively. It shows the most ideal and least ideal configurations to use when it comes to choosing between the factors. It also shows which factors, or the factorial interactions contribute to the printed object in a statistically significant manner. This indicates the user to keep an eye on them.

Infill (Factor B)		Most Ideal Configuration		Least Ideal Configuration		Statistically Significant Factors
Low	High	Height	Base Area	Height	Base Area	Height	Base Area
Hollow	20%	Natural 20% None	Natural 20% None	Natural 20% Tapered	Natural Hollow Tapered	C	No Factors
Hollow	80%	Natural 80% None	Natural 80% None	Pink 80% Tapered	Natural 80% Tapered	C	B
Hollow	Solid	Natural Solid None	Pink Solid None	Pink Solid Tapered	Natural Solid Tapered	No Factors	No Factors
20%	80%	Pink 80% None	Pink 80% Tapered	Pink 80% Tapered	Natural 80% Tapered	A, C A*C A*B*C	A A*B, A*C A*B*C
20%	Solid	Natural Solid None	Pink 20% None	Pink 20% Tapered	Natural Solid Tapered	A, C B*C	B
80%	Solid	Natural 80% None	Natural 80% None	Pink 80% Tapered	Pink 80% Tapered	A, C	B, C A*C

Table 2 Summary of the results of the optimal configurations of Experiment-1

 

Infill (Factor B)		Most Ideal Configuration		Least Ideal Configuration		Statistically Significant Factors
Low	High	Height	Base Area	Height	Base Area	Height	Base Area
Hollow	20%	Natural Hollow None	Natural 20% None	Natural Hollow Tapered	Blue 20% Tapered	C A*C	B
Hollow	80%	Natural 80% None	Natural 80% None	Natural Hollow Tapered	Natural Hollow None	C	B, C B*C
Hollow	Solid	Natural Solid None	Natural Solid None	Natural Hollow Tapered	Blue Solid Tapered	C A*C	A A*B, B*C
20%	80%	Natural 80% None	Natural 80% None	Natural 20% Tapered	Blue 20% Tapered	C A*C A*B*C	C A*B
20%	Solid	Natural Solid None	Natural 20% None	Natural 20% Tapered	Blue Solid Tapered	C A*C	A, B, C
80%	Solid	Natural 80% None	Natural 80% None	Blue 80% Tapered	Blue Solid Tapered	A, C A*B, A*C A*B*C	B, C A*B

Table 3 Summary of the results of the optimal configurations of Experiment-2

 

When color pigments are added to natural PLA, some of its properties such as crystallinity is affected ^[25], leading to variation in the way the material prints. Hence, a similar experiment was performed with two different pigmentations of PLA. Factor A’s levels were changed to pink and blue. However, other factors and the settings were kept the same. The expected values also remained the same. Table 4 shows the optimal configurations for the Experiment-3.

 

Infill (Factor B)		Most Ideal Configuration		Least Ideal Configuration		Statistically Significant Factors
Low	High	Height	Base Area	Height	Base Area	Height	Base Area
Hollow	20%	Blue 20% None	Pink 20% None	Blue Hollow Tapered	Blue 20% Tapered	No Factors	No Factors
Hollow	80%	Blue 80% None	Pink 80% Tapered	Pink 80% Tapered	Blue Hollow Tapered	No Factors	B
Hollow	Solid	Blue Hollow None	Pink Solid None	Blue Solid Tapered	Blue Solid Tapered	No Factors	No Factors
20%	80%	Blue 80% None	Pink 80% Tapered	Pink 80% Tapered	Blue 20% Tapered	A, C A* C	A, B, C A* C
20%	Solid	Blue Solid None	Pink 20% None	Pink 20% Tapered	Pink Solid Tapered	A, C A* C	A, B, C
80%	Solid	Blue Solid None	Pink 80% Tapered	Pink 80% Tapered	Blue Solid Tapered	A, B, C A*C, B*C	A, B, C A*C A*B*C

Table 4 Summary of the results of the optimal configurations of Experiment-3

 

The data from Table 2-4 show that tapered objects should have a lower priority while 3D printing. The bigger the shape, the more accurate the overall geometry ^[20], hence the cones can clearly be seen having the least ideal configuration in all cases, because the size of each successive layer reduces since they taper. Also, objects printed using natural PLA are consistently seen in the most ideal configuration columns in Table 2-3, making it preferable when compared to its colored counterparts. Choosing the settings available in Tables 2-4 could prove beneficial to reduce filament wastage while printing using PLA.

 

References

References can be found in the Introduction section.

Optimizing 3D Prints: Results: Optimum Configurations for 3D Printing (Part 1)

Results: Optimum Configurations for 3D Printing

When the factorial analysis was performed, each of the three experiments had six sub-experiments for every combination of the infill level. Each of these sub-experiments produced graphs and charts for further analyses to obtain the sets of optimal configurations. Fig. 3 shows one example with the pareto chart and cube plot in the case of the second experiment with an infill setting of 20% and 80%. Fig. 4 shows the main effect and interaction among factors with an infill setting of 20% and 80% in the case of Experiment-1. All the relevant graphs and charts can be found in the supplementary material.

The ideal configuration was determined by comparing the fitted means from the cube plot to the expected value. The expected value for the height was 30 mm and for base area was 706.8583 mm². The pareto chart shows the statistically significant factors in Fig. 3. The statistically significant factors indicate that the differences between these groups are not simply due to a chance and are real, this can be said with a confidence level of 95%, since all the obtained data was normally distributed ^[18,19].

 

Fig. 3 Pareto chart and cube plot of 20% and 80% infill for height (top) and base area (bottom) in Experiment-2.

 

The experimental factors can show effects such as main effect and interaction effects. The main effect is the effect of one factor on the experiment while ignoring the effects by all other factors. It shows how much the average performance of one level differs from the average performance of another level ^[18,19].

In different settings, different factors show main effect. For example, the plot in Fig. 3 indicates that the infill factor doesn’t show main effect, because the plot is nearly parallel to the central line of average. However, the color and the shape of the object show a significant level of main effect.

In the case of the interaction plot, which shows whether one factor affects another ^[18,19], the graphs make it very clear that the tendency to interact is high when the lines are intersecting. Henceforth, if two or more factors interact, they are indicated using an asterisk between them.

 

Fig. 4 Main effects and interaction of 20% and 80% infill for height (top) and base area (bottom) in Experiment-1.

 

References

References can be found in the Introduction section.

Optimizing 3D Prints- Experiments: X-ray Tomography

Experiments: X-ray Tomography

A powerful technique in determining the internal structure of closed objects is X-ray computed tomography (CT) or simply, tomography or CT scan. A beam of X-rays is projected on the desired specimen. The radiation transmitted by the specimen is captured by an optical receiver. Images are captured for each discrete rotation of the specimen, and they are reconstructed into a 3-dimensional density map. This map can be used to observe the internal structure and determine its flaws and structural inhomogeneity ^[5,12,16]. It can also be used to determine minute morphological variations. A Bruker Skyscan 1172 Micro-CT scanner was used to perform this experiment. The specimen here is the 3D printed PLA object using the FFF printer. This is placed using a mount and a paraffin film, so that there is no movement. This film is transparent to X-rays.

It was important to choose an appropriate voltage to keep the power supplied very close to 10 W, so that the X-ray source and the transmitted X-rays from the specimen remains high. A low voltage would result in an inefficient capture of images by the receiver ^[5,16]. The voltage chosen here was 44 kV. The corresponding current was 222 mA.

When objects were scanned, the captured dataset had 631 horizontal cross section and 1000 vertical cross section slices of images. When it came to the scanning the outer shell, 641 images were captured for pink cones and cylinders, whereas 901 images were captured for the transparent cylinders. The resolution of the obtained images was set to 1K (1000 x 666 ppi for the outer shell, 1000 x 632 ppi for the vertical cross section, and 1000 x 1000 ppi for the horizontal cross sections). The images were captured for all 4 infill levels for natural cylinders, and for both pink cones and cylinders.

 

References

References can be found in the Introduction section.

Optimizing 3D Prints- Experiments: Statistical Conformity

Experiments: Statistical Conformity

While determining the quality of the 3D printed objects, it was important to make sure that all the prints conformed to the specification. The two measurements of interests were the height of the object and the diameter, as they are required to determine surface area of the objects. When the right-circular cone and right-circular cylinder were modeled, they were designed to have a height and base diameter of 30 mm each. The statistical software, Minitab was used to analyze the data. Three factors were considered.

The first was pigmentation. It is important because it has an influence on the final print-shape. The user always has an option to use colored PLA filament. Adding colored pigments to natural PLA will give it different properties ^[25]. For an extruder temperature above 493 K (219.85 °C), the pigmentation becomes relevant as it affects the roughness of the final print ^[23], hence a lower optimal temperature was chosen. The temperature can also alter the printing material’s crystallinity, depending on its color ^[25], which will, in turn, affect the appearance of the printed object. In this experiment, the colors used were natural (translucent), pink, and blue.

The second factor was the amount of infill. The amount of infill has a range of 0 to 100%. An infill level of 20% and 80% are commonly used ^[23], and on top of that, the two extreme levels of 0 (hollow) and 100% (solid) were chosen along with them. For a shorter print duration, a lower infill is chosen, for better stability, a higher infill is chosen ^[23].

Finally, the third was the type of shape/deformation of the object, i.e., whether it gradually tapered (cone), or didn’t (cylinder). There are several other factors which can contribute to the quality of print such as the temperature of the nozzle, size of the nozzle, rate of filament retraction and many more ^[9,20,21,26]. They are not considered in this experiment.

Once the prints were complete, they were measured using a digital Vernier caliper for accurate measurements, and the grand averages of the values of the height and base diameter were obtained. The area of the base was also calculated. Each experiment was replicated to check for bias. The test used to perform the analysis was Anderson-Darling test using Minitab. The collected data was confirmed to be normal with a 95% confidence interval, as shown in Fig. 2 for cones of 80% infill. Similar data for other infill levels of all objects also exists and is made available in the supplementary material.

Fig. 2 Graphs showing normality of the measurement data in the case of 80% infill level.

 

In Fig. 2, the factor that determines the significance of the result of the normality test is p-value. Since the p-values were higher than 0.05, the data collected is normal and does not have false-positives. The closer the p-value to 1, the more normally distributed is the data ^[18,19]. It also shows the standard deviation of the data.

Experiment-1 was a 2^k factorial design, where k is the number of factors; each factor has two levels. The pigmentation factors were selected to be natural (translucent) and pink, and the shape was tapered or none. Each time the experiment was performed, only a pair of the infill factors was chosen to keep the levels consistent. i.e., a combination of two among hollow, 20%, 80%, and solid were chosen. Depending upon the combination, an appropriate level was chosen to be the lower infill and higher infill level. This was done to determine the optimum levels of infill if the choice were among the combinations. The shape/deformation factor were tapered and none.

Experiment-2 was the same as the first, except the pigmentation was changed from pink to blue. Finally, Experiment-3 factored the colored PLAs (pink and blue) for analysis. Table 1 shows the factors and their associated levels. For simplicity, they are labelled as factors A, B, and C in the table and henceforth.

 

Factor	Level 1	Level 2
Factor A (Pigmentation)	Color 1	Color 2
Factor B (Infill)	Lower Infill	Higher Infill
Factor C (Shape)	None	Tapered

Table 1. Factors and their levels

 

References

References can be found in the Introduction section.

Optimizing 3D Prints- Experiments

Experiments

To determine the optimal configuration, two kinds of analyses were performed on 3D printed objects of select shapes or deformation. The deformation here indicates whether the overall shape of the object is tapering or not. The shapes chosen were right circular cylinder and right circular cones. These objects were printed at different levels of infill and with different colors. The colors are referred to as pigmentation, especially in the figures.

The first was a statistical experiment, performed to determine whether the shapes conformed to the specifications. The second was a tomographic scan to determine the variation of structure and the layers of the printed objects.

The printed layer thickness was set to be approximately 200 µm. The printer used was Ultimaker 2+ Extended. It was set to the following settings: nozzle size of 0.4 mm, nozzle temperature set at 210 °C, default build plate temperature of 60 °C, and PLA filament thickness of 2.85 mm. Fig. 1 shows a selection of print samples of different colors, shapes, and infill.

Fig. 1 Print samples from one set of experiment

Optimizing 3D Prints- An Introduction

Introduction

Many 3D printers provide a high print resolution, suitable for developing high-fidelity prototypes from a computer aided design model. One of the most widely available printing processes is Fused Filament Fabrication (FFF) type, also known by its trademarked term, Fused Deposition Modeling (FDM) in common literature.  Intricate shapes can be printed through FDM printers such as airfoil and Moebius strips ^[7]. FFF prototype surfaces can be enhanced on a millimeter scale even when they have geometric textures ^[4]. However, it is common practice among engineers and designers to build the low-fidelity versions first, as a proof of concept. There are several low-fidelity FFF printers available in the market. They can be used with a wide range of materials. But the most frequently used materials are spools of polylactic acid (PLA) as they are less toxic ^[8,9,13,23]. Due to the ease in their operation, portability, and abundant materials, 3D printers are designed to have fairly good environmental features, making them practical in educational institutions ^[6]. However, they can be made more sustainable ^[11] and economical ^[17] through material waste reduction.

Experimental studies have showcased the properties of different materials or different colors ^[23] while investigating effects of individual factors on the printed object. Each study focuses on at least one parameter and one material to show its impact on the quality of the final product. Poor surface finish is often caused by tessellation of the computer aided design file and slicing processes. However, the surface roughness can be reduced by modeling a design through optimizing the parameters before fabrication ^[14,22,26].

Printing of the first layer is crucial, as uneven material deposition on the first layer can change the specimen height of other layers ^[14]. Surfaces of the printed objects, especially ones which are textured, tend to show the staircase effect, where each printed layer is distinctly visible and looks like a staircase ^[10]. It is an undesired side effect in low fidelity 3D printers.

Parameters such as build direction, temperature of the extruder, and layer height play a major role in showing dimensional accuracy when compared with infill pattern ^[3]. The quality of geometry of the product also depends on print speed and layer height ^[20]. The surface roughness is also affected by the wall thickness of the printed object ^[21]. Although part build orientation affects mechanical properties such as tensile fatigue of the PLA material ^[2], this study focuses on the surface quality and dimensions of the objects.

Using appropriate design rules while building prototypes can save the hassle of wasted material, time, and costs associated with them. Current design rules exist only for certain boundary conditions and does not include all types of printing processes ^[1]. Statistical and engineering process control can be used to detect and correct the variation in the fabrication process ^[15]. The cost benefits of 3D printing are industry specific. However, material costs make up to 12% of the total costs in additive manufacturing. On top of that quality assurance costs need to be considered ^[17].

PLA is inexpensive, but wasting it should not be encouraged. Because of poor design choices, material type, amount of infill, and several other factors, many prints fail, and many do not appear as expected by the user. In other words, they do not have good quality of print. Hence, hundreds of printed objects are discarded and can easily affect the environment, making the process less sustainable, unless properly recycled ^[8]. But recycling can affect the material, which could, in turn, affect the print quality made using the recycled material ^[8,25]. So, this study shows a way for carefully planning the 3D printing process by using the most favorable settings, to obtain the best possible results without unnecessarily wasting filaments.

Existing approaches use Analytical modeling ^[22], Taguchi method ^[3] and factorial designs ^[4,14,21] to determine dimensional flaws, and X-ray tomography ^[5,12] or scanning electron microscopy ^[24] to determine internal and morphological flaws. In the current investigation, the print material was chosen as PLA because it has consistently been proven to print with ease ^[13] and is not toxic. To reduce the waste from rejected prints, this study uses a 2^k factorial design to obtain a range of optimal print settings. An X-ray tomography is also performed to determine and analyze the unevenness of the print layers and surface quality.

 

Optimizing 3D Prints- Brief

Due to the plethora of things made using 3D printers, a large amount of waste is produced in the form of failed prints and wasted filaments to obtain prints of the best quality. It is important to ensure that the printing material wastage is minimal, even when it is inexpensive, for a more sustainable additive manufacturing. To keep a printed object the closest in appearance to its computer aided design, it is ideal to test the parameters that make for its surface quality. With the appropriate settings for these parameters, it is possible to reduce material waste and print failures. This paper shows that, it is possible to determine the optimal settings for different levels of infill, so that the user specifications are met. It also presents the statistical experiments performed on the printed objects of specific shapes, color and infill level, the tomographic images of the outer shell and the internal structure of their infill, to obtain the favorable configurations for optimal print quality.

***

This was supposed to be a journal paper titled Determining Favorable Configurations for Low-fidelity Filament Freeform Fabrication 3D Printers to Attain Optimal Print Quality and Reduce Wastage, but I think I will post it in my blog instead.

Why? Because this is the best course of action. Enjoy my months of research which I will post occasionally.

Heat Maps

It’s 3.14! Happy Pi Day!

A heat map is a graphical representation of collected data, where large data points are plotted in such a way that it represents the concentration of those points through colors. The color scheme depends upon the choice of the user. Normally, a darker color represents higher density and a lighter color, lower density of the data points.

Using heat maps often help identify the flaws within physical objects (if one knows what to do and how to use it), and movements of mouse cursor, or density of visual concentration while eye tracking in interactive displays.

This makes them very useful in user experience and usability studies to understand why people choose certain parts of a website or a software, and where they have their eyes fixed while using it.

Below is an blurred image of a website (left) and its heat map generated (right) while I was testing it to improve its usability.

Below is a time lapse video of heat maps generated by scanning hundreds of layers of a 3D printed object using an X-ray CT scanner for one of my projects, which has something to do with optimizing 3D prints. More on this another time.

 

Optimizing 3D Prints

This is going to be a new series of posts which do an elaborate research on optimizing 3D printed parts.

Why am I doing this?

Because I don’t like wasting print filament, even when it is dirt cheap… You can get 1 kilogram of poly lactic acid, perhaps the safest material to print, for as little as $20.

For something so cheap, why bother about wasting or not?

No. That’s a bad attitude. It is not sustainable to waste plastics, knowing how they are made.

I once reloaded a new filament on an ultimaker but forgot to stop the process. Do you know what happened? The hit extruder kept releasing the filament.

Well, you might say, just end the process when you see the extruder releasing the material

And I’d do exactly that. But I got caught up in another work and completely forgot about this. The bigger issue was I had kept the 3D printer on… overnight… The new filament (an entire kilogram of it) was turned into a thin string of plastic when I realized what had happened.

What a waste!

I don’t like it when waste happens for unnecessary reasons, be it food, or in this case printing material.

This series will show– mostly through technical means, but also influenced by good design practices– how to curb wasting the 3D print filament in general, by playing with the printer to get the best quality prints.

Emergence

Ideas are more or less a mental representation of an object. What thought provoking idea do you have today, I thought some time ago. Then it hit me like a light bulb popping out of thin air in cartoons when the character gets an idea.

Why not make that?

So meta!

Anyway, I had a long desire to make something out of wood, but it had to be 3D. Without any tools in my hand to make a 3D object out of wood, I had to make it 2D. Neither I, nor my tools would budge, so we compromised and decided to make it 2.5D.

Why is it 2.5D? Because I only had access to X,Y and half of Z axis.

In the end, it came out as I had planned. Just like a light bulb popping out of thin air in a cartoon. but this time, out of a block of wood.

This slideshow requires JavaScript.

Well, it is called emergence because it emerged like an idea. But it also true that it looks like the idea is still forming while it reveals itself. Enjoy this video, it is a bit longer than usual since this will be my last post on using CNC… For now.

Thank you for reading!