Shock needs faster data than you would need for normal robot accelerations and angle sensing. I love the peak capture idea proposed above!
This work by team 900 may be of some use to you. They also used an ADXL375.
1 Like
Hereās an update from us over at Team 4160ā¦
We have a working v1.0 of a crash sensor and we would love to get crash data from teams.
Itās around $80 in parts and we based it in part on the work of team 900 the Zebracorns.
CAD, 3D print Files, schematic, parts list, and code all up on github:
Testing Video:
4 Likes
Iām actually working on doing something similar to this for a college course. Weāre designing an engineering experiment to test solid vs hollow noodle bumpers, different types of wood backing, and the material covering it. How are you simulating the robot impact? Weāre planning on something simple like dropping a weight from a height onto the bumper with the sensor.
(Iām primarily an electrical engineer so let me know if Iām misremembering my physics and making the mechanical engineers wince.)
2 Likes
Gravity is nice and reliable! The weight will give you a known energy and velocity input. You will need to pick your official bumper surface position to standardize the results. Maybe normalize assuming the noodles compress completely. A quick high speed video will answer that question.
I would have a bit of guidance on your weight, and plan to keep it from sneaking out and smooshing anything important.
Magnetic door locks arenāt a terrible way to release the weight that can be electronically controlled. They also make quick release shackles for drop testing, if you want a hand pull. Iāve used those up to 30 or 40 tons 
1 Like
I wish I could do this fast enough to run it at the Robot Remix this weekend 
And San Diego is a bit far of a drive to borrow ours 
We have built a bumper tester but have not fully tested it, and given the energy it has at impactā¦ We want to get a little more safety testing before releasing it to the wild. Itās an approx 6 ft drop and even minimally loaded (mass) it reaches approx 18 F/s.
Give us 2 more weeks of testingā¦ Weāll publish the CAD more widely then.
What falls faster, a pound of bricks or a pound of feathers?
I doubt you are reaching terminal velocity in your setup, or that your setup is a sail and density would change your Vfinal, so adding mass shouldnāt change your final speed?? Just the energy available on impact.
Agreed, there is a lot of room for approx in what we have designed, and what we achieveā¦ Hence holding back a week or two till we have some numbers.
Itās been an interesting teaching moment for the kids on the team that havent gotten to that point in Physics (Energy, Collisions, etcā¦).
Showing them:
KE = PE
1/2 m v^2 = mgh
1/2 v^2 = gh
along with F=MA
and discussing how regardless of the mass the velocity at impact should be pretty similar regardless of the mass on the impactor. but adding weight to the impactor gives more force that has to be dissipated in deceleration.
For now just putting the impact sensor design for anyone to build and test. Do with it as you willā¦ Please fork the github and contribute back if you make changes or improvements. Iāll take more photos of the build as I solder up another one today.
A pound of bricks, because bricks are heavier than feathers:
/s
10 Likes
Inaccurate experiment. Thatās testing weight. The original clearly asked which would fall faster. Since feathers have greater drag, they would reach the ground after the bricks.
/s
A word of caution on drop testing, it can very repeatable if your test setup is good, but it can give very inconsistent results if you donāt have a good setup. The weight needs to hit the target in the same place at the same speed and in the same orientation every time. Depending on the weight and target small changes in position or angle could have a big impact on results. If you are trying to get data about a type of collision you need your impactor to simulate the shape, squishyness and orientation of the real impactor in addition to mass and speed. I used to perform ANSI drop tests on chairs, and we used a padded bag of lead shot to simulate squishy human back sides because a hard object wouldnāt conform to the seat and would throw off results. To get good impact data on a bumper to bumper collision you need to actually hit your sample with a bumper in the proper orientation.
In automotive collision tests you propel the test sample (car) into a standard target (wall). Arguably driving a robot into a wall is the easiest test to set up, but it might not be the most repeatable. One common test for materials is the Charpy impact test where a pendulum is released from given position to collide with a sample with a very precise position and velocity. In a Charpy test you measure how far the pendulum swings after breaking the sample as a measure of how much energy it has retained.
If I had the time to devote to robot collision testing I would use a pendulum similar to the Charpy test. It would be pretty easy to make a pendulum with a variable weight and swapable impact heads with a bumper corner, bumper side and metal bar (truss). You could have different release points for a range of robot speeds. Then you just need to mount sample bumpers or frame sections and release the pendulum. It would be very repeatable, easy to configure to different collision types, easy to add sensors for data collection or it could be run without any electronics to test durability.
2 Likes
We just need to form the FIRST Bumper and Frame Collision Administration. FBFCA for short.
3 Likes
Seconding what Patrick recommended above.
Mark Koors once built a pendulum impact tester using a 2x4 that swung from a hefty bolt, with a cinder block fixed to the swinging end. Work area had high ceiling so he was able to use a pivot point that was about 11 feet from the floor, on a steel truss. The item being tested was on a table. Impact load was adjusted the way that Patrick illustrated, by changing the initial pendulum angle. I recall seeing video so probably someone at AM has it.
2 Likes
Hereās the tester a few iterations ago:
Weāve been improving it bit by bit, making the release more consistent (gate latch), adding rigidity and consistency in the release and swing.
Itās pivoting on the giant elevator bearings from andymark from a few years ago.
We are testing and improving the apparatus and getting some good results. We are open to Teams coming on by to test their ideas (San Diego, Mission Bay). We will write up and publish our results.
This is the start, and itās getting more repeatable. We need to make a few more adjustments, but I hope that we will be able to at least provide Comparative dataā¦ So that there is something to backup choices we make with bumper design, and to maybe give FIRST some ideas about how bumper rules could/should change (maybe next year)!
In the graph, no data is collected when it is idle, data is collected when a shock event occurs, so they are all evenly spaced out.
The first 4 events we were testing and making adjustments. The last 8 are reasonably valid tests (we are still planning a few more adjustments to the release). The video was from the day before. Todays videos are all out of focus.
What instrument is measuring the data shown on the vertical scale? ADXL375 mentioned in your earlier post?
A peek at how you attach the instrument would be helpful. In the video it looks like the pendulum strikes the unit under test at point near where the instrument wire runs from.
The sensor measures the sqrt((x^2)+(y^2)+(z^2)) of the accelerometer data.
The sensor design and info are here:
Itās attached to the Aluminum 1x2 on the back in the middle.
