The Starting Point
Initial Research Concepts
Following my research on: Nervous System, Heather Dewey-Hagborg, Oliver van Herpt, and Mathieu Lehanneur.
Last year I used LiDAR data, which was interesting, but I'm keen to focus elsewhere. So I started out by creating a list of what data sources are personal and easily accessible to me.
Possible Data: Call logs, message logs, bank transactions, crypto transactions, stock transactions, social media interactions, internet traffic usage, car usage patterns, Apple Health data, Whoop data.
Since I'm using data, I will need to be careful which sources I choose and how I gathered the information due to law and regulations. Among the available options, Apple Health data and Whoop data stand out to me for that reason. Both platforms allow me to share my data with others, making it publicly accessible. Currently, I do share my information publicly within the Whoop app. While social media interactions are inherently public, they are unrelated to the common connection between Apple Health and Whoop.
Apple Health is fairly well known and is a simple way to store health data. I store mental wellbeing, mobility, and hearing as well as it is integrated with Whoop. Whoop is a fitness and health wearable that stores data such as heart rate, respiratory rate, blood oxygen, resting heart rate, heart rate variability, skin temperature, and more. The Whoop is never off my wrist for more than two minutes at a time and so has over two years’ worth of reliable data, making it the most sensible option for me to use.
written Concept 1
In the spirit of Oliver Van Herpt, I use my heart rate to control the oscillations through the object whilst it is being 3D printed. This could be done through a speaker, the same as Van Herpt, or through a different method.
This lacks originality and is just substituting in a different form of data. Regardless, I am just brainstorming.
written Concept 2
Taking the data of a single sleep, putting it into graph form, and then using the graphs’ frequency and amplitude to dictate the form of a 3D object.
Better than Concept 1, but unsure how well the amplitude and frequency would convert to a 3D vertical form. This requires further research.
written concept 3
Export my Mobility data from Apple Health. I have Double Support Time, Walking Asymmetry, Walking Speed, and Walking Step Length. To take this data and generate a dynamic form from it, essentially digitalising my walk.
What to do further, I'm not too sure. Again, further research into this will be required.
written Concept 4
Going back to my LiDAR scanning, they are points within a cloud. If I were to take those points and manipulate them according to different health metrics. For example, the size of the points is related to my heart rate, and the density of the point cloud is tied to sleep:restorative sleep.
The initial scan could be of my face, then manipulated through my Whoop data.
Reflection
Some of these ideas are really interesting, but the main question across all of them is: Why do I want to do this? What does it show?
Concepts 2, 3, and 4, I will do further research into.
Whoop Data
contents
To gain access to my Whoop’s raw data, I have to request the information, which will come within 24 hours to my email. Below is a screenshot from Whoop on how to export your data. It shows all the data that I will receive from them.
Choice of Data
Since the amount of data is significant, I have decided to select aspects of the data to use. After I downloaded the data, it gave me four .csv files: journal_entries, physiological_cycles, sleeps, and workouts. From these, the data within journal_entries and workouts aren't appealing to me. But the Heart Rate and Sleep Cycle information is.
concept development
Concept 2 Development
Upon receiving my data and looking at the selection of Heart Rate and Sleep Cycle information, Concept 2 seems like the better idea.
To start, I took the following Heart Rate Data: Resting heart rate (bpm), Heart rate variability (ms), Max HR (bpm), Average HR (bpm). Then I took an average across all of them. I am taking the RAW data as a RAW number, ignoring the measure. This data is per day, and I have 625 days.
I then used the Chart tool to chart the averages. I decided to use Line and Radar charts. Line is probably the most known chart, and radar looked like it would generate some interesting results.
Screenshot of some of the data and both charts
Once I had the graphs, I didn't know how to export their shape into a tool such as Fusion 360. Within Fusion 360, I found the option to import .dxf files, and I found an online tool that converted .pdf to .dxf. Following this, I copied both graphs as .pdfs and used the tool, opened it within Fusion, and refined each document.
Import
Select to delete
Export
Import
Select to delete
Export
Fusion experimentation
Line Chart Revolve
Line Chart Revolve
Line Chart Revolve
Line Chart Plane
Radar Chart Extrude
Following the successful import of the charts into Fusion, I made some simple objects. The first and final experiment reminded me of a phonograph cylinder. The idea of being able to play this was very interesting, but being able to print at that detail wouldn't be possible short-term, and the gaps between some data would probably be too large; however, it is an interesting idea. The second experiment was very basic and doesn't cause much inspiration.
From this experimentation, to further develop this chart form, I need to find a different piece of software.
WASP App
Thinking about the making, I want to look at different machines, one of them being the Delta WASP 2040 Ceramic 3D Printer. I have previously used it in my second year, during which Geoff had mentioned that WASP has an app for owners to use to generate G-Code. I decided to check it out.
I discovered they allow you to import your own drawing in .dxf format, which was perfect. The tool gives you the ability to change the modelling dimensions of the Top, Middle Up, Middle Down, Bottom, Height, Twist, as well as changing the Slicing Settings (Nozzle size, Layer height, etc.). I immediately imported the .dxf file and it rendered, allowing me to change the dimensions.
Sleep
Following the WASP App exploration, I decided that I would use the Sleep data to manipulate the Modelling settings. Since the core of the design is my overall health, using the data from a single sleep shows the visual side of how sleep affects us. By making a range of different designs using different days, you could see how significant sleep actually is.
I decided on the following:
Top - Deep Duration
Middle Up - REM Duration
Middle Down - Light Duration
Bottom - Awake Duration
Height - Sleep efficiency %
Twist - Sleep Performance %
Reasoning:
The durations are ranked in order of importance.
Sleep efficiency is how effective the time in bed is used for sleep, which represents how well the ‘structure’ of sleep is utilised.
Sleep performance is an overall measure of sleep quality, which shapes how we feel when we are awake and so shapes the design through distortion.
Screenshot of a portion of the sleep data.
WASP App to Printing
Once I have the model, I need to adjust the slicing settings. If I were to use the Ceramic 3D Printer, I could simply select the appropriate nozzle size and layer height, export the file, and print it. However, I’m currently using the PLA 3D Printers, which this software isn’t optimised for. The software does offer an option for a 0.4 nozzle size, but when I try to export it, every piece of software reports that it’s corrupted. The workaround is to select a 1.2 nozzle size, which will compromise the quality of the print but will still work. Another issue is that the software is web-based, which limits its compute power. This causes problems when selecting the layer height. I’ve found that I can get away with a 0.4mm layer height, but this means I have to reduce the overall modelling height. But once the export is successful, I can then rescale the model in Cura.
Once in Cura, I can make any tweaks and send it off to the printer.
Beta Version
Modelling
I documented the process I went through to generate this beta model. During this print, I hadn't yet decided on which sleep factors would change the modelling settings, so the "Bottom" and "Height" do not correctly reflect any data. The slider bar (sometimes) moves up in sets of two, so I often had to go one above my data's numbers.
The print time was about nine hours, mainly due to the amount of support material.
In Cura, I had to ensure the wall thickness was 2mm to prevent it from collapsing or having a different issue arise.
The two machines the software is designed for
The dxf import and model settings
The slicing settings
Cura settings
Print
Since the print was 9 hours, I left it to print whilst I went home, but I have included some images from the first few hours.
It was a successful print, but the amount of support material and the difficulty in removing it mean that it isn't ideal. I need to change some Cura settings to reduce the amount of support or possibly remove it completely.
Beta object development
Version 1
Taking on board the feedback from Geoff, I removed the support entirely. I also ensured the data entered in the Wasp App was correct to the nearest selected digit. I did have to scale the height in Cura, but it exported without issue. After scaling the Z axis up, I then uniformly scaled it down to 75% to once again save material in case of failure.
Model Settings
Slicer Settings
Scale Z axis
Uniformly scale down to 75%
Cura Print Settings and Print Time
This was a larger model, but due to the lack of support, the print time was significantly lower.
Start of print
The opening up seems to be holding well
Continues to be successful
The inside could do with less spacing
Few stray pieces of plastic
Past the widest point but now it's going to overlay
Print pattern changes a few times
Successful print
Waiting for the bed to cool, lighting looks good.
Finished Print
Finished Print
Finished Print
Finished Print
When a light is past through, it looks incredible. the small gaps let light through like a lamp shade
Top view with light
Comparison to Beta
Successful print. I will now create a few different variations, print them ideally without supports.
Version 2,3,4,5
I decided to print multiple versions in one stint overnight. This was to understand how I could efficiently manage the time the printer was in use, to ensure I didn't hinder other people's progress.
Version 2
Version 3
Version 4
Version 5
I set them out in Cura to hopefully do within a single print session. I gave them adequate spacing and scaled them down to 75%.
The print started out fine with no issues. Except for Version 1, which I monitored and it looked like I recovered.
Something catastrophic went wrong through the night
Version 3 was the only salvageable piece.
As you can see from the images above, the print failed. Since this was done overnight, I don't know at what point it went wrong, but Version 2 was (for the most part) successful. In the first image, Version 1 does appear to have an issue, but I monitored it, and it looked like it recovered and wouldn't be an issue.
Upon a debrief with Geoff, he said a possibility for the failure was that the spacing between the pieces was too small for the print head to fit in, causing it to knock the pieces around, which would cause issues when it came to printing on the previous layers.
Remodel
Fusion 360
Following the tutorial with Geoff (02/02/2025), I enquired about SLS Printing at Print City. The previous versions were all too thin, being 0.4mm, this meant I had to thicken my piece up to at least 1mm thick. Since using the WASP APP was becoming a pain to work with, outside of the WASP printer, I decided to model the piece within Fusion 360.
Import .dxf
Select extrude
Extrude to the height (Sleep efficiency %) and then set the number of faces to 3
Select the bottom face and scale up according to the numbers document
Select the lower middle face and scale up according to the numbers document
Select the upper middle face and scale up according to the numbers document
Select the top face and scale up according to the numbers document
Select the top two faces and twist (Sleep performance %)
Select the whole object and fill in, set to "Collapse" and "Keep Creases"
The piece
The scaling for each section comes from my Sleep.numbers file. The radius of the .dxf file is 29mm so I created a table that takes the data from the Sleep and gives me the scale for the forms that's based on the 29mm .dxf file.
This is not all the data. I have 625 rows in total.
Version 6
Taking the steps I came up with in the exploration stage, I took real data from my sleep data and printed it.
Within Cura, I adjusted it to have a 2mm wall thickness with no top or bottom thickness, just to save a bit of time. The total time to print was 14 hours, 44 minutes, with a rough cost of £4.25.
Cura Settings
I left the print to go overnight, and I kept it at 100% scale. It turned out really well; Fusion is the way to go.
ON the 30cm squared print bed for scale
Side upper view
Above view
Base
Top
Twists
Twists
Compared to the WASP App 75% scale model
Compared to the WASP App 75% scale model
From the images above, you can see that there are no longer any gaps between the layers, which results in a much smoother and elegant look, especially as the lines wrap and twist around the side. There are some stray pieces of plastic that can be trimmed and cleaned up, but overall, it is a significant improvement over the previous one. At 100% scale, it looks incredible, but I would like to see it larger.
The stray pieces of plastic are likely due to the resolution. Within Cura, you can see I set it to be Fast. Overall, it does a good job, but this is not good enough for a final piece, so when appropriate during this research phase, I need to print two objects, one at fast and one at normal resolution, to compare.
Version 7
Following the success of the previous version, I decided to use the feedback given in the group tutorial (13/02/2025) and look at a really bad night’s sleep data.
By ordering the list in ascending order of sleep performance, I quickly found the worst. Sleep performance is the best measure of overall quality. The worst result was 26%, with Sleep Efficiency also being at its lowest of 30%, and each stage's duration was: Deep - 29 mins, Rem - 9 mins, Light - 112 mins, Awake - 371 mins. Fortunately, I keep a journal and I was able to trace the reasoning for this back to the data - illness.
When the model was complete, I was worried that the print wouldn't succeed without support because of how gradual the print would be. Regardless, it was quite a short print so I decided to try without supports and monitor it.
The Cura settings. You can see how short the piece is. I had to scale it down to 60% size just to fit on the bed; it was a very wide object initially.
The slight gradient the printer has to build layers on
Finished print on the bed
I took the skirt off the piece. Side profile with the twist viewable
Flat on a desk
Top side view
Comparison to V6 which was a randomly selected sleep
Comparison to V6 which was a randomly selected sleep
The drastic difference in shape between V7 and V6 clearly shows just how much sleep affects you. Such a huge difference between illness and average.
Since this was an illness, I will take the next worst sleep performance model and print that. Following that, I will do the same with the best sleep performance. Create a set of four objects, the two worst and two best sleeps. While still trying to refine the print quality.
Version 8 and quality refinement
Returning to refining my workflow, now understanding how capable the print is, I printed the second worst sleep this time without illness. I decided that I would tweak the settings for this print, resulting in multiples.
Full Scale on the Print Bed
Some of the lines have printed but others seem to have collapsed
This issue of collapse seems to be a recurring issue. I presumed that this was due to my selection of "Fast" for the resolution settings. To help me understand the problem, I took the same object, scaled it down to 75% to save time and filament, and used the "Normal Resolution".
Here is a side-by-side comparison between the two. As you can see, the issue is not resolved, meaning that it's a slicing issue, and I need to change my Cura settings.
We can see within the "Prepare" section that it is purposely not printing these sections. I obviously didn't do the correct checks before printing.
Following this discovery, I played around with the Cura settings to try and force the software to detect it. I believe the issue is that the walls are very thin at the edges, and the software refuses to slice these parts because the printer can't print that thin. However, within Cura, there are several settings that allow me to force the printer to print them regardless.
I created a custom profile called "ThinWallPrint" with alterations from the "Balanced - Normal - 0.15mm" settings. Above are the alterations.
Although not perfect, the result is far better. This was the best I managed to figure out on my own. I will contact Gary at Print City to see if he has any recommendations.
50% Scale
50% Scale
It's now printing most of the walls, but it is still lacking detail, which is reflected within Cura. I have done the best I can to force the printer to print every wall regardless of width. Upon more research, Cura used to have a dedicated 'Vase' mode, which seems to be what I'm looking for, but it doesn't appear to be an option. However, there will be a way around this issue; I just need to find it.
I have discovered the equivalent setting - "Surface Mode" and set to "Surface". This mode prints a single wall tracing the mesh surface; it has no infill and doesn't allow a base or top to be printed, which may possibly cause some issues for me later on.
My only concern is that a single wall with no support feels quite risky, but it's the only method I can see that will work, so I must give it a go.
50% Scale
50% Scale
This was very successful. At 50% scale and at "Fast" resolution, it prints the walls correctly and is near perfect. The next step is to do a 100% scale model at "Normal" resolution to see if I can get it to be perfect.
The wall thickness and no support turned out to be no issue at all, but I’m still worried about 100% so another good reason to go to 100%, for future peace of mind.
The flexibility of TPU95A
100% scale next to 50%
Compared to the first 100% model
The family of V8 showing the refinement development
The refinement process order
The family in order of refinement
We ran out of PLA filament, so I used a spare roll of TPU-95A. I did a bit of research and knew it would be elastic, but my main goal was to see if at 100% scale the print retained the quality it had at 50%.
The result was a pleasant surprise; I didn't expect it to be as elastic as it was. The print quality was perfect, which now means I’m happy with the Cura settings and can proceed with more prints. Using PLA or plastic as a final material doesn't appeal to me, but using the TPU-95A has changed that; its unique properties justify it, so I may consider using it in the future.
V8 - Nylon
Following the tutorial with Geoff (06/02/2025), I did some research into SLS printing and specifically the artist Dr Lynne MacLachlan. Printing in nylon would give a great final material whilst still allowing for colouring. I contacted Gary at Print City for a quote to print two of them. Both to experiment with dyes. Gary did say the nylon 12 is dark, and so I will be limited to dark dyes, but I will have to take a look when I receive them.
The two identical prints
The two identical prints
Side profiles
Leading lines
Some of the print came out damaged which may just be because of how the pieces were laid out on the print bed
Overall, I am very happy with the result, especially for the price. Together, these cost £65 at 100% scale. As you can see in the last couple of photos, there are some small printing issues, but I think that can be fixed by using a different layout in the printer. However, I need to ask Gary what he suggests.
DYEING
During the formative review with Geoff (27/02/2025), I did research into the appropriate dyes. He mentioned a couple of brands: RitDye and Dylon. Dylon seemed to be for use with washing machines only, which wouldn't work because of how delicate my objects are, so I went with RitDye in a powdered form.
Gary warned me I was limited to darker dyes, so I chose: Dark Green, Navy Blue, and Purple as these were the top three darkest colours available. I bought them through the RitDye Store, and at £3.89 per box, it was a good price - they recommend two boxes per 1KG of fabric, my print weighed 60 grams. I bought two boxes per colour which totalled to £28.33 including shipping.
Boiling Water
Measure out correct amount of power. 2 sachets (64g) = 1KG of fabric
Stir the power into the boiling water
Check the dye is strong enough
Place object in
Move object around
After 30 minutes I took it out
Left upside down to dry
Pour liquid into container for reuse later on.
Above are the steps I took for the first dye test - found online here. I chose Navy Blue because the sachet was slightly torn and leaking. During this process, I wore PPE including: rubber gloves, safety glasses, an apron, and a face mask. I did this in a kitchen, so all surfaces and equipment were cleaned thoroughly afterwards.
The navy blue looks great, but anyone who walks past, including myself, thinks it is purple rather than navy blue. When next to the dye box, it matches the colour with a slight dark tint, and compared with the grey, it looks very different, so the dyeing process absolutely works, and I've got the correct method. I think I will leave this for a few days and wait to see how the green looks, and maybe I'll redeye it purple to understand if relying works and how much of a difference there is between this navy blue and the purple.
Dark Green Dye
Following the successful dyeing of the navy blue, I did the other this time using the green dye. I followed the same steps and the result was just as great. This time the colour was hard to see initially, but after 24 hours, it's very clearly a dark green or forest green.
The tissue test
Drying, a small line looks like it wasn't properly submerged or the dye didn't take.
A fully dyed side
Next to the Navy Blue
In the natural sunlight
In the natural sunlight with the Navy Blue
Side by side on a table
Side by side next to the dye boxes
Purple/ReDye Test
Following the success of both dyes, I decided to re-dye the navy one because of how often people thought it was purple, and this would give me an understanding of how well objects of this thickness re-dye.
Power weighted out
Tissue colour test
Purple and Green
Purple and Green
Purple and Green
Purple and Green
The camera doesn't pick up the colour very well, so I have done a bit of correction to try and make it the same vibrancy as it is in person.
Version 9
Initial Print
After spending the time to get the correct settings within Cura for Version 8, I thought the rest of the objects would be easy to print; unfortunately, that was not the case.
At first glance, it looks like it printed well, but when you look closer, the print had an issue quite early into the print and somehow managed to recover. I think this was because within Cura, I set the filament to the Ultimaker brand rather than generic. Possibly, the filament wasn't hot enough, and it didn't stick to the previous layer, and as a result, it missed a large section until it finally caught. Another option was that the print nozzle actually hit and shifted the object, which is unlikely but still an option. I wasn't there to monitor it, so next time, I will start it a few hours before I leave so I can monitor it.
second print
This time I used PETG filament for a transparent look and used the generic setting for it within Cura.
This print turned out much better, but I hadn't considered an option that was blatantly obvious in the first failure. The fins failed to print, which at the time I just associated with the print failing lower down. But looking at the result of this print, it's because of the angle of that section of the print. It's trying to print virtually side by side rather than diagonally on top, which is a problem I can't solve without using supports. The third print I will use supports to see if this fully rectifies the issue.
Third print
Taking a look at Cura, I have added supports and blocked everywhere except the part that failed, just to save time and filament. Below is a screenshot demonstrating that. I will start this early on in the day so I can stay to make sure it starts the print well and doesn't ruin itself.
Cura settings with supports
At this point in the print, I noticed a small issue - the print was flexing up which caused the extruder and casing to slightly catch on the side however it hadn't done any damage so I left it
Finished Print
The support's couldn't be fully removed without casing damage
There were still sections it failed to print but considering supports didn't help, there isn't much I can do
This confirmed that the issue lay with the angle of the fins. The sections where support was added printed; however, because of how thin they were, I couldn't remove the supports. I did cut them, but I'd have to remove the support layer by layer or sand it down. Otherwise, the print was successful. The fin issue might not be as big an issue as it seems after the talk with Jonathan Boyd - since the fins are extreme because of missing data, isn't it appropriate that they actually aren't perfect, giving a more accurate depiction of the data... the same as the nylon? I do agree, but if that is the case, it can't look like a failure, which the nylon doesn't.
Mounting
During this refinement process of getting a consistently high-quality print, the idea of how I display these pieces has been looming in my mind.
Concept 1
Since the group tutorial (13/02/2025) gave me feedback that the sideways viewing is more appropriate, I have been thinking about the best method to display them in that orientation.
Concept 1 involves creating a chain where the deep sleep of object 1 matches the deep sleep of object 2, and then the awake of object 2 matches the awake of object 3, and then object 3's deep matches object 4's, etc., for the longest possible chain. Since I have over 600 days’ worth of data, I fed all the data into Grok 3 (Think enabled) and asked for it to find the longest possible chain. It returned this result: Days 1, 124, 56, 102, 3, 82, 205, 14, 20, 233, 447, 4, 150, 6, 220, 263, 13, 153, 67, 74, 25, 39, 200, 202, 206
These would then be put onto a thin metal bar that would mount onto a wall or onto a plinth - the bar could have an LED strip running under it to illuminate the inside of the object, possibly creating an interesting lighting effect.
I went through the first 3 and confirmed they aligned and then made them in Fusion 360 to make sure this concept could work.
I was not happy with this result because the lines didn't flow and they just looked wrong, and it doesn't show the data well. So I decided to try a slightly different chain where Object 1's awake aligned with Object 2's deep, etc. This proved much harder for the AI to do, even though I kept the prompt and the wording visually the same, subbing out awake and deep appropriately. As a result, I used a different AI model, DeepSeek (DeepThink R1 enabled), which gave me the longest valid chain of five days: 72, 132, 125, 309, 240.
This looked much better; however, it still felt wrong and removed the individual aspect of single-night sleeps, especially when presented with random dates just to make it look seamless.
Concept 2
Using the research I did on the Aureole collection by Darklight Design, I am really drawn to wall mounting and even including lighting.
Concept 2 involves designing a mount for my pieces for them to be attached to a wall. The pieces themselves can't have a mounting system built in because I can't hollow them out. So this concept needs to hold my pieces to the wall and then also have housing for a light and diffuser.
Concept 2 development
Version 1
Cura settings
The printed piece
Piece inside the object
Doesn't hold it straight
The large extruded sections have enough flexibility to bend without breaking, allowing them to fit inside the object, but when they are inside, they don’t hold it straight because one side slides down. The object has some flexibility itself, so the next iteration I want to try two options. One where the "L's" are much closer to stop the sliding and let the flexibility of the plastic let it fit through the hole. The second would be a direct iteration from this version, where I would add an extra part below the existing extruded section to also stop the slipping.
Version 2 and 3
Less tall and used the backing to stop slides
Cura Settings for both
Support block and part broken off
Support block and part broken off
These didn't work because I added supports which blocked the pieces, and I couldn't remove the blockages. I tried with so much force that I broke both pieces. A new design is required.
Version 4
Cura settings
The two parts
I didn't account for tolerance
I took inspiration from an old light we had lying around from when the house was built. I noticed it has a screw mechanism and decided to try it. Unfortunately, I didn’t account for tolerance, so they didn't fit, but I liked this idea much better.
version 5
3% tolerance applied
Added screw holes
the two parts
Still not enough tolerance
Much better than the previous version, but it still hadn't got quite enough tolerance.
Version 6
Cura settings
Fit together but slightly too small tolerance
Separate parts
Rescaled for tolerance
Fit together perfectly
Following the group tutorial (13/03/2025), I looked at creating a twist lock mechanism. This worked quite well, but I printed these at 50% scale to save time and filament before knowing it would work. Since this did work, I will print at 100% scale and test on my piece.
version 7
Cura settings
The two separate pieces
The combined piece
The gap between for the object to go into
Inside fit
How well it holds the object
Slightly too large diameter which caused it to break my object
Printing it at full scale worked well, but the diameter was too large, and the tabs that stick out for the locking were causing some issues. I had to use so much force that I broke my object. In the next print, I want to revisit the screw mechanism as I all try and print a thinner version and in TPU to try and make the tabs flexible so I can squish them in whilst they go into the object.
Version 8
PETG top part and the 20% infill base
Holds together well
Very hard to move the tabs inwards
The 10% infill base
Doesn't hold together quite as well
Still very hard to move the tabs inwards
Version 8 is a thinner iteration of Version 7 and printed in TPU to attempt to give flexibility so it can get inside the object without breaking it or without extreme force.
I printed the same object twice, first with infill at 20% and the second with infill at 10%. I reprinted it because, as seen in the photograph, it was much harder to manipulate than I had hoped, so by lowering the infill I hoped it would soften it, which it did but not as much as it needed to. Another slight concern was that on the second print, the PETG top section didn't fit as well as it did on the higher infilled print. Following this, I decided to retreat from TPU and reconsider the screw mechanism.
version 9 and 10
Cura layout for both prints
Version 9 two pieces
Holds well but with just a bit of force they come apart very easily
Version 10 two pieces
Much stronger connection than Version 9
Ignoring TPU, I went back to PETG and reprinted this time, bringing back the screwing mechanism. The tab lock version (9) has the problem that with force it comes away; to fix this, I'd have to increase the size of the tabs, making it even harder for it to go inside the object. The screw mechanism worked perfectly and held together really well. Following this, I decided that my mounting system will use a screw mechanism rather than tab locking. The next step will be to use this design with nylon to make sure this will work, but that will be done under refinement.