Droppin' Stuff!

Now that the International Congress on Medieval Studies in Kalamazoo, MI has come and gone, I can talk about some tests I did for a good friend's paper at the conference: penetration testing on a reconstruction of the Lubbek Wappenrock. This marks the first real outing of Da Towa, and so this post will also discuss some of the changes I have made and will make in the future.

The post-drop carnage.

DAQ update

The first major change was that I bought a DAQ with a higher time resolution: the DLP IO-14, which has the capability of recording at up to a 20 kHz rate in bursts of 8192 samples. All that was required was swapping out the smaller DLP IO-8 I previously used in the electronics takeout box and modifying the datalogging script, then recalibrating the system. I didn't expect the calibration to change much, and it really didn't. But you know what they say about assumptions.

I did a few tests with the chisel penetrator, but knew I'd want to verify against the EN13567 penetration test, so...

New Penetrator

I decided to try to make a penetrator like the one called for in the penetration tests detailed in EN13567: specifically, a 3mm square penetrator with a pyramidal point with an included point angle of 120\(^o\). it started quite innocently as a 3/16" round pin punch, but through judicious use of a sharp file, some homebrew jigs and a set of trusty digital calipers it became this:

When you don't have a machine shop, and mechanics calls, you do what needs doing.

To get the square profile, I started by clamping the flat sides of the pin punch ensuring the whole thing was square to the vise. I then used the vice as a reference plane to start removing material from the pin punch section. Periodically I'd check the diameter until I'd removed half of the total (~1.8 mm total). Then, I'd flip the piece and repeat the process. Finally, I used a set of soft jaws to enable me to make the final set of faces, which corresponded to an edge on the body of the punch (it is a hexagonal section), using a similar process.

The real fun was ensuring I got the point correct: 4 planes, with a specified angle. It took me about two tries to get a point I really liked. To do it, I ended up fixing the newly squared punch in a similar manner to when I made the faces. I then clamped a wooden guide to the vice to ensure the file stayed in the correct plane to create the desired angle. After a surprisingly short period of time (maybe an hour or so start to finish), I had a reasonable facsimile to the penetrator described in EN13567.

Final measurements and closeup of my homemade EN13567-like penetrator.

Impact Zone Modifications

Once I actually got the new penetrator out to Da Towa, I realized that it wasn't the same length as the chisel, so I used a bolt to make up some of the length difference (so that the point was well above the PVC holder while still contacting the load cell in some way). This adds the potential for some strange behavior during impact if the two aren't in good contact, and is something I plan to address in the future.

I also realized I'd have to make some adjustments to the 'landing area' for the sample holder to ensure that it didn't try to ride too far down the penetrator (ideally, it'd only contact on the 3mm section, not riding up to the tapered section). I ended up settling on some 'pig mats' (spill cleanup mats. Think: really butch paper towels) and later adding a layer of rubber hose to help dissipate excess energy after full penetration was achieved. The advantage to this approach is that it is adjustable to how much hose I layer. The disadvantage is that it takes some trial and error to get right without cutting penetration short - really I need to make a longer penetrator (this one is about 2/3 the length called for in EN13567, and lacks the distinct shoulder of that penetrator). Also, for extremely flexible fabrics this wouldn't really be a good solution, but I thought it would likely work for the stuff I was testing

The pig mats are the yellow material below the penetrator.

Preliminary Tests!

In EN13567, the verification process for the penetrator is called out as using a sample of a cotton canvas as described in EN388: essentially two layers of a 16 oz cotton canvas. Unfortunately, I had none on hand. K-zoo was fast approaching and I still wanted to verify that everything was in working order before testing the samples I'd been sent. So I snagged some scrap of the heaviest plain woven linen I had on hand: probably about 8 oz. From this, I cut a pile of circular samples using the holding rings of the sample holder as a pattern. Each of the samples I then marked so that I could offset the angle of the warp fibers as called out by EN13567 (45\(^o\) offset). I also set up the DAQ to record 8192 samples at 4 kHz: a roughly 0.25 ms resolution.

I carried out a series of drops at 36", 72" and 81" (max height) with a 1.540 kg sample holder (the samples were only about 2g a piece: a negligible increase over the empty holder). For each drop height, I performed at least 3 repetitions with fresh samples:

Linen preliminary samples before and after. Note the large round holes from the shoulder of the penetrator.

If you notice the size of the holes, I was running into an 'over penetration' condition. The 'over penetration' occurred when the sample rode down onto the taper because of kinetic energy that wasn't dissipated during penetration and the stop not being in the right place. This was more an annoyance than a real issue as the forces were still being measured at a high enough rate that I could see distinct penetration events. It did make sample postmortems more difficult, and led to more noise in the penetration signal:

Preliminary 72" drop test with 2 ply white linen samples. Note the two sets of large peaks (penetration events) and the signal noise after penetration.

Summary of preliminary drop tests of 2 layers of coarse, heavy linen (~8 oz).

While we're here, it's worth making some observations based on this preliminary test data. First, the main point of the test was to verify that I could get some useful penetration data. Mission accomplished on that one: my time and force resolution were good enough to get something of use for a basic comparison. But notice the difference in the penetration forces at the different drop heights (only 36" and 72" shown above): there's about a 34% increase in penetration force when I drop the speed by about 30% (if I include the outliers. The difference is greater without them). The kinetic energy drops by half as a result of the height change, but as is obvious from the samples, there is sufficient energy to penetrate at both heights.  So, the speed affects the force necessary: something discussed in papers I've read on needle penetration forces in sewing operations, but something I need to understand better. That said, we can see significant outliers in both cases: for each drop height one sample has a dramatically reduced penetration force. This is likely a product of variation in the samples themselves, and to some extent in the experimental setup. This highlights the need for a fairly large sample size when doing this sort of test: EN13567 calls for a minimum of 10 samples.

One final thing I noticed was that my voltage signal seemed to oscillate quite a bit more than it did when I did my thrust tests, even at zero load. So the overpenetration issue wasn't the only source of signal noise. More on this in the next steps.

Overall, I was able to get some reasonable confidence in my procedure and equipment, at least sufficient confidence to try out the samples from Jessica. Oh, and her presentation was only 5 days away. So like any good research engineer on a rapidly approaching deadline, I had a beer and strode forth boldly to my doo… I mean testing.

Wappenrock & "350N" Fencing Bib vs Da Towa

Because Jessica did a much better job describing the construction of the wappenrock than I ever could, I'll just summarize it here and point folks to her work (not yet published), from inside to outside:

• 20/20 thread count twill weave fustian layer
• 8/8-9/9 irregular tabby weave heavy linen canvas layer
• cotton fiber fill 
• 8/8-9/9 irregular tabby weave heavy linen canvas layer
• 20/20 thread count twill weave fustian layer, painted with a linseed oil and carbon black paint

The whole mess is then quilted through all of the layers, a process which required a vice and some serious muscle based on Jessica's stories. Cutting it with a pair of good fabric scissors certainly was tough. The sample 'swatch' she sent over was sufficient for me to mark off 6 samples, though I only tested 2 in the preliminary study. Each sample weighed about 8 g.

Sample swatch of the wappenrock. Painted black side down.

For comparison to something a bit easier for folks to relate to, I used a bib from one of the masks I received from my good friends at the Chicago Swordplay Guild: a used by good condition Fencing.net "350 N" mask. The mask bib was noticeably easier to cut than the wappenrock, mostly because most of its layers were much less stout (though it was also quilted). Each sample only weighed in at 5 g:

• thin and light synthetic fabric layer
• low-density open cell foam padding
• thin and light synthetic fabric layer
• more dense open cell foam padding
• coarse and stout synthetic layer

350 N Bib samples

Again, I only ran two samples from one bib, though I was able to cut 4 good samples out of the bib. The bib samples were ~8mm thick, with 25 mm quilting, compared to the ~8.5 mm thickness of the wappenrock and two quilting widths: 25mm & 15 mm.

Test Procedure

The procedure I used for the tests was fairly straight forward:

1. Samples cut and maintained at room temperature prior to test
1. Wappenrock samples cut moments before testing began
2. Bib samples cut several days before but no material loss was noted while they sat
2. Sample loaded into sample holder and tightened with pipe wrench (as tight as I could make it)
3. Sample holder loaded into bottom of tower and door secured closed
4. Sample raised to desired height from point of penetrator
5. Datalogger started and sample dropped after ~1 s
6. After impact and data collection finished, door reopened and sample holder removed
7. Resulting damage inspected before and after removing sample from holder

Each test was made from a drop height of roughly 1.83 m (6 ft as measured, because the tape measure wasn't metric), with an estimated speed of ~6 m/s on impact with penetrator point. The sample holder weighed in at 1.54 kg as in the earlier linen tests. Datalogging was carried out at 4 kHz (.25 ms/sample) and had approximately a 6 N resolution with an estimated error of +/- ~30 N (due to ground variation mentioned earlier).

Results!

The results of the two sets of tests were quite interesting. Firstly, all four samples penetrated fully - so the bumpers weren't causing any issues. Further, it illustrates that the energy I was using was sufficient to cause penetration in these samples with my EN13567-like penetrator. The kinetic energy in my tests was only ~28 J, compared to the EN13567's 90 J, while being at roughly the same speed range. So, in short, if I were to have added mass to the penetrator the additional energy would have just been 'excess' to be dissipated by the bumpers at the bottom of the tube. I'd expect this to change with different penetrator sizes though, particularly for much larger profiles. The reason for this is fairly straightforward: to penetrate a sample with a given penetrator, a certain amount of mechanical work (think: energy manifested as mechanical action) must be applied to the penetrator (in this case, by the sample falling onto it). This work is the minimal energy required to deform the sample, causing the yarns to migrate or fracture, ending in the penetration of the sample. But there'll also be heat being generated due to friction, and some other secondary forms of energy dissipation that occurs (noise, for example - though possibly too low volume to hear). The greater the amount of deformation that must occur, such as if we had a penetrator that must fracture more yarns, the greater we can expect the minimum energy to become. This gets a bit muddled however if you change penetrator shapes such that the dominant energy dissipation mechanisms change: say from a yarn migration mode with a sharp penetrator to a yarn fracture mode with a very blunt penetrator.

Now that we have some understanding of the energy aspect of the penetration, let's look at the forces. For all four samples, we see two regimes: the initial impact and penetration regime (0-5 ms), and the secondary impact regime (8-18 ms). In the first regime, the penetrator makes first contact and begins deforming and penetrating the layers.  At initial impact, these samples are traveling at roughly 6 m/s or 1 mm in about 0.17 ms. However, as penetration occurs the kinetic energy at initial impact is being dissipated as work on the sample and thus the speed will decrease (no major mass change).

Preliminary results for 350 N bib and Wappenrock test samples.

In the initial region, we see the bib samples exhibiting a peak forces of 991 N and 859 N for samples 1 and 2 respectively. Bib sample 2 actually exhibits a marginally higher peak force in the secondary impact regime, which may be caused by impact with the broad shoulder of the penetrator after sliding down the 3 mm stem area. These forces are well more than double the rated value for the bibs (350 N), which is likely a product of a significant factor of safety being taken into account in their design and production - not an unusual practice in safety equipment. In both bib samples, the failure mode was yarn migration and breakage with little extrusion of the padding. During the initial impact, it seems likely that the thick synthetic outer layer offered the most resistance to penetration with subsequent layers offering progressively less resistance, as illustrated by the peak early in the initial impact region that then quickly decays. In bib sample 2, the penetrator actually penetrated through a line of quilting stitches, and I recorded a lower peak force (even taking into account the voltage swing) than bib sample 1, which penetrated between quilting lines. This is something that I'd like to repeat, to see if this trend holds.

Bib samples (left) and Wappenrock samples (right) after penetration.

 The Wappenrock samples, on the other hand, exhibit a longer initial penetration region and substantially higher forces: 1789 N and 1846 N, for samples 1 and 2 respectively. The forces are nearly twice those of the 350 N bib samples, and the initial penetration region is also nearly twice the duration. In Wappenrock sample 1, we see a 2-peak form in the initial penetration region: a peak of roughly  834 N and a second peak of 1789 N. However, in Wappenrock sample 2, we don't see the two peak form, instead we see a slope change in the load curve. From inspecting the samples, we see that the failure in sample 1 included the extrusion of some of the padding through the exit hole. So it is possible that the first portion of the force curve (from 0-2 ms) is the point at which the outer layers are being penetrated, while the period from 2-3 ms corresponds to the penetration of the padding material (the drop indicating that the padding is easier to penetrate than the outer layers) and finally 3-4 ms corresponds to the penetration of the inner layers. In sample 1, it's easy to attribute the large rise to the force needed to both penetrate the liner layers as well as extrude the padding. Meanwhile in sample 2, it appears that the penetration of the outer layers included some extrusion of the coarse inner layer through the outer, though not to the extent of the padding in sample 1. The main difference between the two Wappenrock samples is the density of the quilting: sample 1 had wider quilting, providing more opportunity for the padding to move, so one possibility is that the similarity in the peak penetration forces between these samples comes from similarity between the forces needed to extrude the padding and to penetrate the tightly quilted padding. Since both were penetrated, it's difficult to say which failure mode is 'better', though perhaps an argument based on wound cleanliness could be made. However, a much larger sample size of each quilting width is needed before drawing any final conclusions.

Another interesting finding is that the slopes of the force curve from about 0-1.9 ms are very close for both the Wappenrock and bib samples: this most likely corresponds to the deformation and onset of penetration of the sample. One of the unfortunate drawbacks to my setup is that I can't video the moment of impact: that would lend a great deal of additional information.

Next Steps

After a bit of poking around I found out that the voltage swing issue was actually the outlet I was using: in the thrust case, I'd used a fully grounded outlet while the outlet next to the tower had an open ground. Apparently this affected the voltage regulator's ability to maintain a consistent voltage: it was still within the USB spec, but more variable than I wanted for measurement purposes. The solution was simple: use a different outlet.

For the shoulder/over-penetration issue, I decided to slide some washers over the end of the penetrator and glue them together to mimic an abrupt shoulder that would hopefully reduce over-penetration issues.

By the end of the testing, when I finally got the 15oz cotton duck, I'd noticed that my tower seemed to wobble more and the clearance to the load cell was smaller than expected. The bottom support had begun to pull out from the wall, and so the entire base had shifted. I fixed this by remounting the entire base more securely and adding a set of support feet (that I really should have had in the first place… oops). Sadly I didn't notice this until after doing 10 of the canvas samples… good thing I have lots of it.

The final issue I have to address is the connection between the force transducer, the ground and the penetrator. Some of the noise in the post-impact period likely comes from these bits not being fully integrated. Part of this solution may end up being buying a 200 kg-f button load cell, which would be roughly the force equivalent to the system I have currently set up.

With the above issues solved, or at least addressed, the next big task will be to perform the drop tests of the 15 oz duck to compare with the values cited in EN13567 for penetrator verification - that way I'll know if I'm producing results that would be in the right ballpark for future comparative testing.