3D printing

Figure 14: Two designed dimensions being shifted to match the closest layer

As CAM software automatically shifts a designed surface towards the nearest layer an inaccuracy of up to half a layer height can occur at the G-code generation stage before reaching the manufacturing equipment. This can be solved by reducing the layer height – raising the resolution – to reduce the error cause by misalignment or in extreme cases raise resolution to a point that layer heights are smaller than any desired increment; however this leads to expensive equipment and longer print times. As an alternative the Z-axis dimensional accuracy can be improved by a designer aligning each feature with the nearest layer multiple, reducing inaccuracies while using a lower resolution machine, producing parts with fewer layers and reducing print time. This also reduces CAM processing time as the software has fewer calculations to make in approximating the design.

The tool shown in Figure 15 is developed from layer design research to help align design features with layers and layers with machine capabilities. It also takes into account the simple mesostructure property of solid layer thickness for the layered shells of parts and suggests the closest alternative dimensions to the designer’s intent. For example, a 1.90mm feature height would produce an inaccuracy of 0.05mm so the alternatives of 1.80mm and 1.95mm are suggested based on an extrusion height of 0.15mm. The calculations that this tool is built on could be built into CAD software such as Solidworks to accelerate the design process.

Figure 15: Layering calculator

Thin-layers

The above tool helps reduce the number of required layers for a design by increasing the layer height to the maximum a design will allow. The tool also has an important use when designing parts that require thin layers to meet project requirements – high tolerances, mechanical, etc. The importance of layers being multiples of the machine’s resolution increases as the chosen resolution approaches the machines limit – as possible Z-axis errors can account for a larger percentage of desired movement.

Most of the research in thin layers in this thesis comes from experiments related to the guitar pick project. During this project it was found that multiple thin layers perform better than larger layers. These results align with composite literature into thin layers. When iterating a design’s layer height it is important to modify the number of solid layers the CAM profile uses for the top and bottom of the part as four layers are required at 0.1 mm layer height to equal two layers at 0.2 mm layer height – this effect increases drastically as layer height lower into the tens of micrometer range.

As discussed in detail Section 4.2.2, the process will lower print speed to allow enough time for each layer to cool and solidify before the next is printed. This increases print time particularly on parts created from a larger number of layers. As a greater number of layers produce a more reliable guitar pick, packs of 9 picks can be printed at once allowing the process to operate at full speed and allow enough time for each layer to solidify. These packs take a greater amount of time to print (~37 min) than an individual pick (5 min). Printing in packs allows for a timesaving’s per pick of ~1 min or ~9 min time savings over printing each pick individually.

Figure 16: Printing in packs for faster feed rate

Longer print times can also be reduced in manufacturing with the use of multiple print heads on a single machine – gang heads – which reduce individual part costs for production by creating multiple identical parts at once – parallel production. By adding a second print head and surface, the same pack of guitar picks can be printed in ~2minutes per unit, a third head allows for ~1.3minutes per unit, and so on. The benefit of adding more heads on unit cost is diminishing, making a large number of gang extrusion heads viable for longer prints or to increase production scale.

Figure 17: Diminishing returns of added gang heads

Integrating non-printed features

As the DAM process can build intricate internal features into parts - with minimal additional cost incurred by the added complexity – an opportunity arises to insert non-printed objects, partway through prints for added functionality. These added parts can allow for a stronger assembly by adding objects like bolts, or allow for added interaction by adding electronic parts such as LEDs or temperature sensors.

Again the use of layer multiples allows for easy integration of inserts as parts offset from layers can produce a cavity that is too small for the insert and create a collision between the part and hot-end during production, or it will produce a cavity which is too large, not holding the part securely.

Figure 18: Inserts and Layers

The process of inserting non-printed parts consists of modifying the G-code to pause the print at the top of the designed cavity, retract the filament – as to avoid oozing of molten plastic onto the part - and move out of the way. At this point the insert can be manually placed or an automated tool can be used to pick and place the inserts. The tool must then be moved back into place, the filament pushed back into the hot-end and the printing process continued (Figure 19).

Figure 19: Insert Process (top), paused print with inserted nut (middle), finished print with captive nut (bottom)

Layered Radii

The process creates all radii in the Z-axis by offsetting consecutive layers - Figure 20. Designed features must take this into account to reduce the staircase effect by creating larger radii - that are larger than a single layer. Smaller radii are created with the use and design for thinner layer height, however this will increase the production costs.

Figure 20: Layered Radii

This is an area of rapid advancement, as older version of slic3r – from 2 years ago (2012) - would create a hole and a partial part failure. However the newer versions fill in the area below the radii with a solid layer and ignore smaller radii. This is an example of when design suggestions to avoid failure can be automated into software tools. Beyond changing the dimensions of layers or radii splines can be used to produce curved surfaces that balance design intent and process capabilities. This is discussed in greater detail in the next section.

Support Design

Key to designing support material is an understanding of how the process is going to build your part and which areas will potentially fail. A printed parts surface is either internally supported with infill, self-supported with solid layers or perimeters, externally supported with removable material, or a combination of. Understanding these methods and iterating with prototypes allows for smoother and stronger surfaces while reducing production costs.

Infill

A parts’ infill is an internal support structure for solid layers above it and provides strength to the part. A denser infill provides greater support for solid layers while lower densities such as less than 10% infill do not provide adequate bridging support. These parts must rely on the machine’s ability to bridge gaps – discussed in section 4.2.2- and on the self-supporting nature of multiple solid layers.

Solid Layer

Solid layers are capable of bridging wide gaps with little to no support – Figure 22. Each successive layer adding support to the next increases the surface quality from an unacceptable level to a smooth finish. To better communicate this style of support a construction metaphor is used. When a floor is built, floor joists, which are equivalent to the 1^st layer type – Figure 21 - are used which are spaced apart and span large distances while providing support for a floorboard. The floorboards or 2^nd layer type placed above in turn provide support for a decorative 3^rd layer.

Figure 21: Layer support classification

Just as floor joists are not considered a desirable ceiling feature in housing an exposed under side of a 1^st layer type is often inconsistent and should be avoided on parts or limited to non-aesthetic sections of a part (as done in Figure 22). A designer can increase the surface quality by increasing the layers support by other means such as greater infill - for non-overhanging surfaces -, self-supporting perimeters or external supports.

Figure 22: Surface defects from extended bridging

Self-Supporting Perimeters

When a surface is not located above its internal structure but rather overhanging, support strategies such as solid layers or infill are not practical or possible. The next design consideration for creating supported surfaces is in using the perimeter as an integral support structure. Similar to bridging gaps, the limit on self-supporting perimeters is dependent on a machine’s ability to rapidly cool molten extrusions. A greater cooling ability allows for a higher surface quality at a larger step over - a smaller overlap with the previous layer. Greater step over abilities can produce curves that are closer to the horizontal plane as shown in Figure 23.

Figure 23: Varied curves and resulting overlap percentages

Figure 24: Failed step-over rate (e) and one solution (f) to obtain a similar curve

The limit of 0% overlap can be overcome with self-supporting perimeters by modifying the parts intended extrusion height and width. Shown in Figure 24 the same angle can be produced with a greater overlap between layers by using a smaller extrusion height. The same effect is achieved by extending the extrusion width within the capabilities of the machine. These options can be explored by using the calculator and supporting material shown in Figure 25 which is based on the equations [Tan θ = (Extrusion Width x (%Overlap/100)) / Extrusion Height] and [%Stepover = (1 – (%Overlap/100)) x100] developed during this thesis.

Figure 25: Self-supporting perimeters calculator

A redesign using both solid layers and perimeters can create an exposed 1^st layer type – Figure 21 – that is consistent. This has been done on the hot mount shown in Figure 26. In which the self-supporting perimeter was used to reduce the distance of unsupported extrusions.

Figure 26: GoPro video camera printed mount version 1 (top), iterative design process from version 1 on the left to the self-supporting surface on version 3(bottom)

External

The use of external support should be minimized through the already described methods above as its use increases the amount of material used, print time, and post-processing labor. The design of external support material is a complex task that involves designing many temporary parts. Limited design intent can be given to automatically generated support material in the CAM phase or designed for with the part in the CAD phase of design.

The support structures created must not be stronger than the features it supports as this increases part damage during removal. Each of these temporary parts ideally must not fuse with the intended part but provide support for it. Creating supports that have undesirable part properties such as thin extrusion widths, low density, few solid layers, and little inter layer adhesion can do this. If the support material is made to weak via the mentioned methods it will provide inadequate support.

Figure 27: Biologically informed support material, version 1 (left), version 2 (right)

Applying knowledge of support design to manually written G-code can produce novel support structures that mimic nature. Shown in Figure 27 are two prototype support structures that mimic a spider’s dragline that reduces both print time and material required to support the design. These biologically informed support structures are created with the help of the tool shown in Figure 28 that calculates extrusion rates. As this support material is not automatically designed or coded the use of this support style is limited due to the greater design and coding time/skills required in its creation.

Figure 28: Extrusion Calculator for Biologically informed support

Profile Design

Like when building a house the foundation or first layer profile is important to the print’s success. To succeed the profile must adhere to the print surface and allow for removal of the part once complete. The extrusion ratio, print speed, and surface area of the first layer profile all influence adhesion and part removal. The effect of surface features on adhesion is discussed in section 4.2.3.

If greater adhesion is required and the equipment/design allows for it, a shorter and or wider extrusion can be used. Extreme extrusion widths also affect inter layer bounding as thin extrusions have a weaker bond and can make a part flexible, thicker extrusions have a stronger bond but at an extreme cause collisions and nozzle wear from the excess material oozing upwards.

Figure 29: Extrusion Ratio

These ratios also apply to part design as a short wide part has a greater surface area in contact with the bed, creating a stronger bond. A sort part also reduced the torque applied to the bond between the part and bed – torque = length * force – reducing print failure when the extruder bumps the part, while making the part harder to remove when complete. A part can be made artificially wider with the use of a brim that adds extra material around the first layers profile to increase its surface area. The use of brims should be avoided as they add extra material, print time, and add a step in post processing.

Parts with thin sections like the earring displayed in Figure 30 are difficult to remove without causing fractures in the thinner profile sections. Such parts should be modified to have an even surface adhesion. This can be done by cutting out the center of each solid section for the first layer, creating a hollow void under the raised section of the earrings.

Figure 30: Algae inspired earrings

More important than the extrusion width or extrusion ratio is the speed at which the layer is placed. If too slow, time is wasted, but if too quick the nozzle collides with previous laid extrusions causing defects. Increased rigidity in the positioning system, can allow for faster first layer movements. Faster movements can also be made through G-code modifications made manually to the acceleration (M204 S5000) and travel move feed rates (F6000) for all first layer moves. This is required to reduce travel moves to the speed of extrusion moves, as a fast travel rate on the first layer causes previously laid extrusions to become dislodged from the surface.

Thin features

Every feature’s profile is made up of a perimeter section with a core of infill at varied densities. The width of the perimeter section is dependent on the extrusion width multiplied by the number of perimeters used, while the infill fills the remaining area to the selected density - 0% to 100% - using the selected pattern – rectilinear, honeycomb, etc. - these two simple structures are what create the profile of every printed object.

When creating thin profiles at the limit of two times the extrusion width - Figure 27 - errors occur in both processing the design and creating it. Through design and CAM iteration a part can be created thinner with the use of only perimeters or the use of smaller extrusion widths. A design that uses two perimeters without an infill allow for the narrowest sections but are susceptible to delamination under minimal force. If the machine is capable of producing thinner extrusions the infill and/or perimeter can be made thinner than the other allowing for the creation of both structures. A narrower perimeter increases strength but narrower areas remain weak, as the infill does not fit. A narrow infill illuminates this problem but results in excessive infill print time for thicker parts of a profile. In practice these options should be avoided by redesigning the part to allow for thicker sections or closer widths that can use a smaller nozzle/extrusion width effectively.

Mandatory radii for part accuracy

To accurately design for printed parts each edge that is to be profiled must have a radius equal to or greater than the extrusion width. This is the lower limit that edges can be created at due to the nozzles shape Figure 31.

Figure 31: Profile radii

The inner edge of a designed sharp corner also becomes rounded and requires post processing or a design modification to allow assembly with sharp cornered parts – such as angle iron. This can be done by adding a circle cut along the inside edge and filleting the resulting corners – Figure 32 and Figure 33.

Figure 32: Exaggerated sharp inner corner correction

Figure 33: Failed part on top and successful compensation on bottom

Radii for increased profiling speed

Much as the design of a road influences how quick a car can maneuver its corners the design of parts and their tool paths also depicts how fast a printer head can move, along with the forces it will experience when doing so. Just like when driving if you come to a sharp corner like one at an intersection, you must first nearly stop in the direction you are traveling before you can start traveling in another. If the road has a ramp than your car – or extruder – can move along at a higher speed with minimal deceleration, in the optimal case for both situations the curve is such that no deceleration is required and just a change in direction can be achieved.

Figure 34: Corner action while profiling

This smoothing out of acceleration and deceleration allows the addition of radii to have a large effect on profiling time by reducing the amount of material required – as radii get larger – and by superimposing one axis’s deceleration with another axis’s acceleration. The profiling speed increase via the addition of radii is limited by the equipment’s maximum print speed and by the cooling time per layer - which are discussed in section 4.2.5 and section 4.2.2 – both limits can be reduced through equipment design to allow faster printing.

Figure 35: Radii speed test

When the corners of a four-sided hollow part - Figure 35 - (10mm x 10mm x 4.95mm) are given a radius of 0.5mm a timesaving is realized. Larger radii reduce the time further to a maximum timesaving of 12 seconds - 12.6% reduction at 5mm radii – by also lowering the amount of material used. When numerically analyzed based on machine capabilities the addition of radii has a larger effect on average speed when the acceleration rate is lower - 9% average speed difference at 3000mm/s^2 acceleration, 90mm/s max speed – or when the maximum speed is higher – 8% at 9000mm/s^2 and 150mm/s – this is compared to a normal speed change of 3% - at 9000mm/s^2 and 90mm/s. these results show that the addition of radii is beneficial to both lower accelerating machines and to sturdier machines that are capable of high speeds and acceleration, as radii increase the amount of time spent at the desired speed.

Mesostructure Design

The observations from the testing of seven different parameters has generated a greater understanding of structural and commercial changes caused by mesostructure design decisions. The collected data is compared against the standard 30% test specimen – Appendix C - and other infill data collected during an engineering directed study on the effect of infill on tensile properties of materials. The study is conducted using the ASTM D638-10 standard with a single specimen per case. Each specimen design is a modification of the 30% infill specimen’s mesostructures. The following modifications where made to each case:

• Infill every 2 layers: The infill for this specimen is printed every second layer creating a core made up of 0.3mm layers and a perimeter of 0.15mm.

• Infill +45 Degrees: The infill pattern was rotated 45 degrees switching its +45/-45 layup to a 0/90 layup with the applied force.

• TB 0.45: An extra solid layer – 0.15mm - was added to both the top and bottom of this specimen. 2x 0.3mm solid thickness to 2x 0.45mm solid thickness.

• 2 perimeter: An extra perimeter has been added to this specimen

• EW0.3 EWP0.5: The perimeters extrusion width is kept at 0.5mm, while all other extrusion widths are modified to 0.3mm.

• Honeycomb: This specimen’s infill style has been modified from a -45/45 layup to a honeycomb pattern.

• Side: This specimen’s orientation has been rotated 90 degrees so that it is printed along its long side.

A large variation in material properties is capable via mesostructure design modifications as shown in Table 3. The extended data and observations are discussed further in the rest of this section.

Table 3: Material property range via mesostructure variation

As a major part of commercial design, how much material and how long a process takes, influence the cost of production and the products success or failure. A typical pricing structure is shown in Figure 1 and each mesostructures influence on operational and material costs are ranked in Table 4.

Table 4: Mesostructure resource comparison (percentage change)

As the data in Table 4 shows the most extreme variation in resources occurs when a part is printed on its side - as shown in Figure 36. This change in orientation requires 22% more material and takes 93% more time then the reference specimen. This time increase makes print orientation the largest factor for costing. Comparing this to printing an infill for every two perimeters – 0% change in material, 18% less time – shows that design decisions related to how the part is created (its mesostructures) can raise or lower the amount of time required printing the same amount of material.

Figure 36: Print orientation has a greater effect on structure and time than any other mesostructure

Part orientation also effects the creation of radii – profiled vs. layered - and need for support, both discussed earlier in this chapter. As such a balance must be found between factors affecting the mesostructures design, layer design, support design, and profile design. This balance should be considered from the beginning of the design phase as to minimize the number of layers, which reduces costs and maximize desired part qualities such as surface quality and strength.

Figure 37: Correct orientation allows plastic parts to cheaply connect steel structures

Table 5 adds yield data to the comparison of resources and are ranked based on an equal weighting of each of the five variables shown. In practice this weighting would not be equal but rather vary based on project goals. Again print orientation is least desirable, as its structural performance does not make up for the extra resources it takes to create.

Table 5: Equally weighted comparison of mesostructures (percentage change)

Shell design

Adding an extra top and bottom layer and/or adding another perimeter can strengthen a part by increasing the thickness of its shell. Both options add similar strength to the test specimen with modest additions to time and material consumption. Due to solid layers taking less time and delamination issues with multiple perimeters it is recommended to add solid layers to your design.

The use of extra solid layers more evenly balances the shells thickness (0.5mm walls and 0.3mm top/bottom to 0.45mm top/bottom) then the addition of an extra perimeter that increases the wall thickness to 1mm. This balancing should help more evenly distribute the forces placed on the part and could be the reason for the higher yield specific strength.

Adding another perimeter to the shell mostly increases strength when parallel to the applied force. When force is applied on other angles delamination can occur between the two perimeters. This failure most likely starts as small inter-perimeter fusing errors, as shown in Figure 38. Even though an additional perimeter adds 54.29% to the break strength of a tensile specimen the extra layer causes a small part such as a printed c-clamp to fail much easier then one printed with a single perimeter. Each extrusion in the bellow pictures is 0.6mm wide.