Quick Study Involving Falls on Static Materials
|
While bored over the last few days, I decided to create a quick, nonchalant study contrasting static falls on different types of material. The testing parameters were simple: I took 10 11 falls on four different slings; a Metolius PAS, a 1 nylon sling, a polyethylene sling and a piece of dynamic rope; and I then repeated the test with a rigid mass, dropped at a greater height. However, the primary function of this study is in relation to the first test and first two graphs showing the relationship between different sling types. Test one: - 160 lb flexible mass - Fall factor 0.22 - Fall distance, 5 5.5, adjusted to ensure appropriate fall factor Test two: - 11.5 lb rigid mass - Fall factor 1.05 - Fall distance 24 26 First, a quick rundown of the testing methods. In the photo below, you will notice three pieces of material: a 9.1mm yellow rope, a green rope (the test sample) and a piece of green webbing with tape on it. In order to execute the test I would use the GriGri, with my weight held by the yellow rope, to position properly myself as indicated by the tape on the green sling. Then, when ready, I would activate a quick release, which would drop me directly onto the sample, 5 5.5. Last, I would stand back up and repeat the test another nine times. After running through four testing samples, I repeated the test with a solid-steel hoop, which I referenced above as test two. In test two, I did not use a quick release, but rater I picked the hoop up and carefully dropped it directly in line with the load cell. |
|
"Part of the reason why the standard deviation is so high with the ridged-mass trials, relative to the flexible-mass trials, is because of the lack of a load-absorbing component of a flexible human. Of course, the other reason why the standard deviation is rather high is because this was an informal test and therefore slight deviations in the exact placement of the hoop before I dropped it adversely affected the readings, as well as technicalities involving the mass of the load cell." |
|
Lets see... 500 lbf is what.. 2.4ish KN..... right? With the small fall distances that is alarming if extrapolated out to what happens at the cliff. |
|
All I am getting from this is that you used two different weights, two different heights, and two different FF. Umm, great...I guess. |
|
robrobrobrob wrote:Lets see... 500 lbf is what.. 2.4ish KN..... right? With the small fall distances that is alarming if extrapolated out to what happens at the cliff. Second question, what did you use to make the grey/orange graphs, those are nice.Microsoft Excel. 500 lbf is 2.224kN, and yes you could count on these types of forces at the cliff. I have measured forces as high as 3.1kN while hard bounce testing an aid piece (tested on a cliff). |
|
Ray Pinpillage wrote:All I am getting from this is that you used two different weights, two different heights, and two different FF. Umm, great...I guess.You are looking at two different tests in one combined thread. The value is in the first and second graphs, specifically the force differences of near-equivalent falls on different slings. The part I found most interesting was that the PAS produced a higher mean force than a mostly-pure Dyneema sling even though the PAS is only about 28% Dyneema, although I do have a theory as to why that is. Of course, to truly call it one way or another, that is to say one sling will produce a higher impact force than another, I would have to conduct more testing at all fall factors which would be hard considering the probably injury involved. None the less, I thought it might be interesting to show the differences between the most common slings used to tie in to the belay with, even if I can only display one fall factor level. |
|
GLD wrote:"Part of the reason why the standard deviation is so high with the ridged-mass trials, relative to the flexible-mass trials, is because of the lack of a load-absorbing component of a flexible human. Of course, the other reason why the standard deviation is rather high is because this was an informal test and therefore slight deviations in the exact placement of the hoop before I dropped it adversely affected the readings, as well as technicalities involving the mass of the load cell." None of the first, lots of the the last.In order to understand why the standard deviation varies, we would need to look at all of the data, and measure before and after properties of the materials (did the slings stretch a couple millimeters, for example?, and was there a trend in the data). The second time you dropped the load, you dropped it on a different sling than you did the first time. What was your sampling rate? And is your load cell and DAQ made to sample dynamic loads where the force goes from zero to peak back to zero in a few milliseconds? For a *edit* drop test on a static rope *-edit, the peak load is very hard to measure, because it happens so fast. Even with a 1000 Hz sampling rate, you're bound to sample on either side of the peak most of the time. You might nail it periodically, seeing random higher peak values. Would you post the full results of at least one of the drop tests? Also, load on a bolt-in-rock anchor is probably greater, because the rock will strain less than the combined system of the load cell, the cable above it, and whatever you have it bolted into (wood truss?) |
|
20 kN wrote: You are looking at two different tests in one combined thread. The value is in the first and second graphs, specifically the force differences of near-equivalent falls on different slings. The part I found most interesting was that the PAS produced a higher mean force than a mostly-pure Dyneema sling even though the PAS is only about 28% Dyneema, although I do have a theory as to why that is. Of course, to truly call it one way or another, that is to say one sling will produce a higher impact force than another, I would have to conduct more testing at all fall factors which would be hard considering the probably injury involved. None the less, I thought it might be interesting to show the differences between the most common slings used to tie in to the belay with, even if I can only display one fall factor level.The valuable side of the "test" would have been comparing the peak forces between a rigid mass and a flexible mass falling on static slings. All the rest has been done, more or less. Is this for a class or certification? |
|
Eric Krantz wrote: In order to understand why the standard deviation varies, we would need to look at all of the data, and measure before and after properties of the materials (did the slings stretch a couple millimeters, for example?, and was there a trend in the data). The second time you dropped the load, you dropped it on a different sling than you did the first time. What was your sampling rate? And is your load cell and DAQ made to sample dynamic loads where the force goes from zero to peak back to zero in a few milliseconds? For a *edit* drop test on a static rope *-edit, the peak load is very hard to measure, because it happens so fast. Even with a 1000 Hz sampling rate, you're bound to sample on either side of the peak most of the time. You might nail it periodically, seeing random higher peak values. Would you post the full results of at least one of the drop tests? Also, load on a bolt-in-rock anchor is probably greater, because the rock will strain less than the combined system of the load cell, the cable above it, and whatever you have it bolted into (wood truss?)Okay, here is my time-to-go-to-sleep reply (so ignore spelling/ grammar): The slings did stretch a fraction, but only after the first drop. Repeated drops did not increase the length of the sling. The sample rate was about 560 Hz, which while not quite as fast as I would like, but it is fast enough to get quite close. Both of my conditioners (I own two identical ones) are able to provide a selectable sample rate, ranging from 1 Hz to 500. I have done quite a bit of testing involving sample rate where I will put two cells in series with each other, select two different sample rates, and drop a load or take a fall on both cells and compare the data. Normally what I would do is I would select one cell to scan at 500 Hz. and the other to scan at 100 or 10 Hz. Then I would take drop a weight or take a fall on the cells and compare the results. Interestingly enough, I am always able to get the 100 Hz. unit to read within about 5% of the 500 Hz. unit. For 10 Hz, the reading is off by about 20-25%. As far as the cell goes, my cell is a standard NTEP class III cell, nothing special. I have spoken with a number of manufacturers with regard to dynamic loading on class III cells and they have all told me that dynamic loading in the mS range (1,000 Hz and under) is not an issue. They tell me as long as the peak is not occurring in the low uS range, I am fine. If by DAQ you mean my conditioner's A/D converter, well, it can scan at 500 Hz., so yes it is designed to capture dynamic loads. :) I still think the load on a bolted anchor is going to be the same because in this test the bolt in the ceiling was attached to pure concrete and I used a load cell, made out of 1" thick alloy steel and an MBS of 15,000 lbf< attached directly to the sling via a steel carabiner. There really was not anything in the system stretching except me, and to a limited extent, the sling. Here is the data set for the very first Dyneema sling drop shown in the graph above (160 lbs flexible mass). Each reading is about 1.78mS apart and expressed in lbf. 0 1 1 1 2 3 6 6 8 8 8 7 9 11 13 15 17 18 20 22 24 27 29 29 32 37 38 41 45 49 53 58 62 67 73 79 84 87 98 109 112 124 137 146 159 170 181 194 205 217 230 242 254 266 278 290 301 313 323 332 342 351 360 367 375 381 387 392 396 400 402 403 405 405 404 402 399 394 388 380 373 363 353 342 332 319 309 299 288 279 270 260 253 245 238 232 225 220 215 208 203 198 195 190 186 181 177 174 171 168 164 161 160 157 155 153 152 150 149 148 147 146 145 145 143 142 142 141 140 139 139 138 137 137 136 134 134 133 133 132 132 131 130 130 129 129 128 128 127 126 126 126 125 125 124 124 124 123 124 123 122 122 122 121 121 121 120 120 120 121 120 120 119 119 119 119 118 119 119 119 119 119 119 120 120 120 120 120 121 121 121 121 120 121 121 121 121 122 122 122 123 122 123 123 124 124 123 124 125 125 125 125 126 127 127 128 128 129 129 130 130 131 131 132 And here is the dataset from the first Dyneema sample with a rigid mass: 1 3 7 8 9 20 45 92 205 417 544 562 591 576 493 282 126 53 9 -1 Obviously, it is a MUCH faster fall without a flexible mass. Interestingly enough, if I were to set the conditioner sampling the first trial (shown above) at 50 Hz, the readings would look similar to see what you see above, but the peak load would still be fairly close to what is seen at the 500 Hz. sample rate. I spoke with the manufacturer of the conditioner for quite some time about this and they tell me to establish the percentage difference between the peak load and the values immediately before and after. Whatever that percentage value is represents the approximate scan-rate-limited error rate. So far I have found that statement to be reasonably quite accurate. But I agree that the peak load of the rigid mass tests is probably marginally more than what I recorded, but likely not by enough to really matter for my purpose. Ray Pinpillage wrote: All the rest has been done, more or less.Do you have any links? I have never seen someone take a fall on a static sling and record the load. That was the main motivation to explore this topic a bit. Every related test I have seen was in regard to steel weights being dropped on a sling, I was unable to find a test involving a real person. The test was for fun, it was neither for a certification or a class. If it was I would have been for a cert or class I would have performed a far more rigorous and formal study. :) |
|
20 kN wrote: Do you have any links? I have never seen someone take a fall on a static sling and record the load. That was the main motivation to explore this topic a bit. Every related test I have seen was in regard to steel weights being dropped on a sling, I was unable to find a test involving a real person. The test was for fun, it was neither for a certification or a class. If it was I would have been for a cert or class I would have performed a far more rigorous and formal study. :)Even though your study isn't comprehensive, I still think its interesting. This guy has done similar testing (Check out myth #7): geir.com/mythbuster.html |
|
Thanks for the data. Based on what you wrote, it looks like you have covered many of the bases to eliminate sources of error, and you are thinking about sampling rate, and missing the peak due to too slow sampling rate. Great! Also, I like the comparison you spoke of - using different sampling rates on two different load cells to estimate the required sampling rate that will pick up the peak. |
|
Eric Krantz wrote:Thanks for the data. Based on what you wrote, it looks like you have covered many of the bases to eliminate sources of error, and you are thinking about sampling rate, and missing the peak due to too slow sampling rate. Great! Also, I like the comparison you spoke of - using different sampling rates on two different load cells to estimate the required sampling rate that will pick up the peak. I'm curious what happens around the peak load area of the rigid drop test (circled in orange). I trust the peak values of your flesh-and-meat drop test, because there is a nice relatively flat peak that indicates sampling rate was fast enough. However, funny stuff around the peak of the rigid.... almost as if you are picking up points where the load is fluctuating wildly due to vibration. Is it possible that high frequency vibration in the circled area makes the measured load fluctuate, and this is why the data aren't linear there? As in the second chart below. Have you read about Nyquist sampling rate? en.wikipedia.org/wiki/Nyqui… Basically, it says that the highest rate of vibration dictates the sampling frequency necessary to characterize the vibration. You need to sample at twice the frequency of the vibration. You should get a CR9000x. You can sample at 100,000 Hz with this baby. campbellsci.com/cr9000xIt is interesting that you brought up vibration. A fellow climbing engineer recently expressed that he was having issues involving the peak load capture of a multi-point trad anchor whereby he subjected the anchor to high impact falls and tried to record the load. He mentioned the vibration of the load cell synthesized a higher-than-expected load. I took a more careful look at the rigid-mass data and it appears you are correct, the vibration of the cell is messing with the data. This chart represents samples 2, 4, 6, 7 and 8 of the Dyneema sling rigid mass test: This chart represents samples 1 - 6 of the dynamic rope rigid mass test: I also took a look at the flexible mass test, and with no surprise the graphs for all tests were linear without aliasing or vibration. So that got me thinking. I distantly remember the load cell swinging around more during the rigid mass tests. During the flexible mass tests the load cell barely moved because I dropped straight down onto it. So I decided to do some more testing to try to determine how much of an effect the vibration has on the readings. To accomplish this I put two large load cells in series with each other, a 5000 and a 10000 lb cell. I then repeated the rigid mass test. The idea this time is that the load cell on bottom would vibrate much more than the one on top, and therefore the top load cell would capture the actual peak load whereas the bottom cell would capture the peak load with induced vibration. Here is what I got. I staggered the graphs so that you can see them easier, but the peaks occurred at exactly the same time. In this test the 10k cell was on bottom and the 5k cell was on top. Start by looking on the right side of the graph. The first red and blue spikes represent the same simultaneous test. Then the next set represents the next test and so on. What we see is that the load cell on bottom (which was the subject of more vibration) always reads higher than the load cell above it (which was the subject of less vibration). Even though I know my cells are accurate and calibrated properly, I decided to flip the cells around and repeat the test just to be sure. This time the 10k cell was on bottom and the 5k was on top. Again, start from the right of the graph and look left. Sure enough, the results were the same. Next, I replicated the flexible mass test because I recalled that the load cells did not vibrate very much in the previous flexible mass tests. Sure enough the cells both produced the same peak readings, within 3 lbf for every trial. Oh, and BTW, being subjected to a 900 lbf shockload was not fun. Now I have an idea of what a FF2 might feel like. Therefore, this brings me to my questions. Are the abnormal curves we are seeing in my graphs a result directly of the load cell moving around, or from signal aliasing? You mentioned Nyquist rate. I am familiar with it as it applies to cameras. Any pro cinematographer knows that if someone is going to film a screen, the camera capture rate has to be at least twice the refresh rate of the monitor they are filming or else they will get aliasing in the video. However, I dont directly see how aliasing would occur in the A/D converter of my dyno. So rather, it seems if the graph abnormalities are a function of the vibration of the cell as you expected. However, if true then I dont see how scanning at twice the vibration frequency would fix anything. I see how it would increase the confidence of the results and the accuracy of the graph, but if the vibration is inducing synthetic load increases, then the load increases (i.e. the vibration) would appear regardless of how fast the dyno was, no? |
|
Aliasing is always possible when a continuous signal is sampled. (That's exactly what the A/D converter does.) For signals with bounded bandwidth it may be avoided by sampling fast enough (Nyquist's Theorem), but in practice, a little aliasing is always there, which means that it is not posible to perfectly reconstruct the original signal from its samples. |
|
Hey 20kN, very cool stuff. Keep posting it as it is very interesting material. |
|
brenta wrote: {sound of an airplane flying over most people's heads}Thanks for that, Brenta. Always amazed how many engineers I've run into through climbing, and how few combined their professional experience with their hobby. No idea how many here fully understand the implications of your post, but I for one appreciate your taking the time. On a side note, this has always been a pet peeve of mine and why I went with Daytronic 4077's... On top of the 1000Hz scan rate they have a separate pure analog channel to capture the true peak. I'd be interested to see if there's a difference between the results given by one of those and Sayar's equipment, but I sent all my extra 4077's off to the folks at the Rope Test Lab FB group to help with their rope access testing. |
|
brenta wrote:Aliasing is always possible when a continuous signal is sampled. (That's exactly what the A/D converter does.) For signals with bounded bandwidth it may be avoided by sampling fast enough (Nyquist's Theorem), but in practice, a little aliasing is always there, which means that it is not posible to perfectly reconstruct the original signal from its samples. Of course, the artifacts due to aliasing (often simply called aliasing) depend on the type of signal involved, but they all result in some kind of distortion of the reconstructed signal. Wikipedia's article on aliasing in signal processing has a nice image showing how two sine waves may be mistaken for one another if not enough samples are taken. As a special case, a sine wave at the Nyquist frequency (half the sampling frequency) will masquerade as a constant zero because it is sampled when it crosses the t axis. (This assumes that the first sample is taken at t=0. A cosine, then, will not vanish, because it is sampled at the peaks and troughs.)The Wikipedia graph is very useful, and I was looking at it earlier. I understand what you are saying, and you do bring up some important points. It is silly because I understand aliasing in audio, video, and 3-D graphics, but I did not think too much about how it relates to this application. The main reason why I never thought too much about it is because a drop test is a single positive rise cycle followed by a return to reference, and that is it (except in some cases). So I asked myself, "If an A/D converter samples the signal at 560 Hz., then what is the minimum number of samples required to capture a half wave that wont result in undersampling?" Consider the first blue waveform in Dynemma sling rigid mass test. It has a duration of approximately 12 samples. Since it is a half wave we have to double that to get the full-wave frequency which is 24 samples. My dyno was scanning at about 560 Hz. Accordingly, at 560 Hz a waveform that travels from reference, + peak, reference, - peak and back to reference in 24 samples would have a frequency of 23.33 Hz. We know that if we have a capture rate of 560 Hz, then we cannot capture waveforms faster than 280 Hz., which would be the Nyquist frequency. However, that does not make sense to me because a shockload that has a rate of 280 Hz would only cross two samples per half wave which is not even remotely close enough to capture a shockload unless the shockload repeats over and over (such as in the case of a constant vibration). Maybe my midnight calculations are off but that is where I am failing to see the role of the Nyquist frequency in this application. Thoughts? :) |
|
I may add some detail later, but in short summary, the signal sampled by your A/D converter is not a sine wave. Its spectrum is not a single line and is nonzero at higher frequencies than the one you computed, especially if it exhibits some "ringing" as suggested by Eric Krantz. |
|
20 kN wrote:The Wikipedia graph is very useful, and I was looking at it earlier. I understand what you are saying, and you do bring up some important points. It is silly because I understand aliasing in audio, video, and 3-D graphics, but I did not think too much about how it relates to this application. The main reason why I never thought too much about it is because a drop test is a single positive rise cycle followed by a return to reference, and that is it (except in some cases). So I asked myself, "If an A/D converter samples the signal at 560 Hz., then what is the minimum number of samples required to capture a half wave that wont result in undersampling?" Consider the first blue waveform in Dynemma sling rigid mass test. It has a duration of approximately 12 samples. Since it is a half wave we have to double that to get the full-wave frequency which is 24 samples. My dyno was scanning at about 560 Hz. Accordingly, at 560 Hz a waveform that travels from reference, + peak, reference, - peak and back to reference in 24 samples would have a frequency of 23.33 Hz. We know that if we have a capture rate of 560 Hz, then we cannot capture waveforms faster than 280 Hz., which would be the Nyquist frequency. However, that does not make sense to me because a shockload that has a rate of 280 Hz would only cross two samples per half wave which is not even remotely close enough to capture a shockload unless the shockload repeats over and over (such as in the case of a constant vibration). Maybe my midnight calculations are off but that is where I am failing to see the role of the Nyquist frequency in this application. Thoughts? :)This is all speculation, but my thoughts: The Nyquist frequency comes into play during that millisecond of peak load. It's not a continuous linear increase to a maximum peak, then a linear drop off. Instead, a longitudinal wave is propagating through the sling or rope, and the load on the anchor would be vibrating with the frequency of that wave. You can get a close approximation of the frequency of this wave using Young's Modulus and the density of the material, then you'll know the Nyquist frequency. It appears that won't be affected much by the tension in rope, so you should be able to calculate a general value and use it. How is maximum load during the 1-millisecond affected by the S-wave? I don't know, but this may also come into play because the S-wave may cause the sling/rope to slightly elongate, increasing and decreasing the tension at the anchor as a function of the frequency. Might be easier to calculate the frequency for an S-wave, given a rough estimate of maximum tension from your tests, length of sling and mass. There's a great longitudinal wave animation here: animations.physics.unsw.edu… |