A challenge. Finding the length of a potato strip

No trivial trains and tunnels this time!

This is an actual problem that was solved about 20 years ago by one of our engineers. The goal is to measure the length of a potato strip rapidly. The potato strips would free fall one at a time vertically and they were oriented vertically before falling. The potato strip would break two very fast photo eyes on the way down through a ‘scanner’ or length grader. There is a top photo eye and a bottom photo eye so the bottom tip of the potato strip would first block the top eye then the bottom eye. The top eye will be unblocked when the top tip of the fry uncovers the top photo eye. The sequencing depends on the distance between the photo eyes and the length of the fries.

I was asked if measuring the length of potato strips could be done. I immediately said yes if the potato strips could be oriented vertically reliably. That was a food handling problem. I was asked how because the speed of the fry is not known since the fry does not free fall until the top tip of the potato strip is released. Longer strips are released with the bottom tip closer to the top photo eye so long strips are not moving as fast since they have less distance to free fall.

I told one of our engineers how I would solve the problem in a general way and then let him solve it. He did. I did not ask how. I knew how I would solve this problem.

There are two very high-speed counters, one for each photo eye. The counter used a many megahertz clock so there was no problem with time resolution. The counter for each photo eye would start counting when the photo eye was blocked and stop counting when the photo eye was unblocked. The counts could be converted to time.

What is the formula for determining the length of the potato strip?

This is not trivial but not that difficult. It will take some scratch paper and head scratching. There are probably a couple of ways of looking at this problem. If the counters are configured to measure the time between blocking the top and then blocking the bottom, and unblocking the top and unblocking the bottom then there is a different formula. I think this second method of measuring the time between the blocking on the top and bottom then the time between unblocking on the top and bottom may be easier.

PM me your solutions, if any.

The reason for measuring the length is that the French fry producers want the length to be about 2.5 to 3.5 inches so they fit in one of those red McDonald’s French fry holders or similar. There were a lot of good reasons for measuring the length of fries this way.

Nice one Peter, reminds me of a project I did some years ago. Counting sausages into a can, it would also measure the length. the sausages were fed into a bowl feeder, fed by belts into the diverter & can, this worked at very high speed (we used a high speed camera to actually see what was happening). The process counted the sausages into the can and also measured the length, if the length was out of spec an air jet rejected it into a bin. when the last sausage entered the can the diverter flipped backward to divert the next sausage into the following can, as this can moved forward the diverter then moved forward with it. This was on a Siemens S5 PLC, to get the speed the count/length blocks were called many times in the organisation block (task), I have no idea why I/O interrupts were not used just I/O refresh (perhaps not available at the time).
Cannot remember the way we did the calculations this was in the early 80's but it worked a treat. By the way this was either 4 large sausages or 8 mini ones in a can, however, the mini ones were not checked for length as these would rarely break, you have probably can guess this was for beans & sausages in a can. Sorry my brain is dead so will not be attempting to find a formula.

If you get a short french fry, what do you do?

I would just use 3D laser scanners... They do make them.

Had to interface with sorting machines that used them, very impressed although those were colour sorters, the plant we put in would sort peas, beans and various pulses like lentils, I had the pleasure of putting 10 pieces of metal painted slightly darker colour than the average pea in a hopper of 5 tons of peas on the commissioning (they were metal just in case they missed any and could be picked up in metal detection at a later date), it picked every single one out on 5 trials.

Yes, this technique can be used for other things besides potato strips. The lengths were used to build a histogram. The data was used to allocate potatoes by size to different plants. The end goal is to have most fries in about the 2.5 to 4 inch size. We cut strips longer than 6 inches in two so they fit in the little red fry holders that McDonalds uses.

There shouldn't be strips less than one inch. They get grade out up stream. The length grader doesn't care about the length of strips except the longer ones tend not to start to tumble. The algorithm works.

3D scanners are expensive. If the PLC has fast enough counters and many do, then this can be done in a PLC or a arduino. We didn't use a motion controller for this. Most of the cost was programming and reporting the length histogram.

We do color scanning and more.
https://deltamotion.com/peter/Videos/Delta Fry Cutting Machine Demo.mp4
but this topic is about a simple length grader and how to solve a problem.

This length grader was cheap and simple.

BTW, there are two ways to use the photo eyes I realized that that when I made the first post. I solved the problem both ways. I do not know exactly how it was done 20 years ago.

I assume now with scan times what they are, you could get away with just one sensor and recording the time it is on for, thus determining the length.

I'm guessing the speed of the fry / chip falling doesn't fluctuate too much depending on it's weight.

What an interesting project that must of been though.

janner_10 said:

I assume now with scan times what they are, you could get away with just one sensor and recording the time it is on for, thus determining the length.

I'm guessing the speed of the fry / chip falling doesn't fluctuate too much depending on it's weight.

What an interesting project that must of been though.

Click to expand...

Bad assumption. The fry accelerates due to gravity and that isn't dependent or weight or mass. PLC scan times are too slow and irregular. However, PLCs do have counter cards. We used an off the shelf stand alone computer board but I don't know which one. It had the two counters and the gate inputs to start and stop counting.

Using

s1=Ut+at*t/2

s1=distance between photo cells
t=time taken from photocell 1 to photocell 2 from counter module

re-arrange to determine the U, the speed of the strip when it crosses the first photocell.

If t1 is the time the first photo cell is covered,

then length of strip is given by

s2=Ut1 + at1*t1/2

I did something similar more than 15 years ago, I used several photocells to measure steel pipes length from 6 meter min. up to 12 meter max.
When pipe cover two photoeyes initially , you capture that moment, then when pipe starts abandoning (conveyor is moving) first photeye the counter start counting until pipe leaves second photoeye, distance between photoeyes is known, then add that distanced from what you got from counter (pulses) converted to distance.
Result is pipe length.
In my case I used 6 or 7 photoeyes because 6 to 12 meters is a huge difference.

L D[AR2 said:

Using

s1=Ut+at*t/2

s1=distance between photo cells
t=time taken from photocell 1 to photocell 2 from counter module

re-arrange to determine the U, the speed of the strip when it crosses the first photocell.

If t1 is the time the first photo cell is covered,

then length of strip is given by

s2=Ut1 + at1*t1/2

Click to expand...

You got it! At least the general idea.
You found the simplest solution. There are 3 that I can think of depending on you you use the photo eyes.
I was hoping to post an evil emoji but you ruined it.

Peter Nachtwey said:

but you ruined it.

Click to expand...

Ruined it? Really?


I would expect there are more than a few here who remember Physics 101; more than likely it is why they are here.

Here is a link to 3 solutions. Each requires a different configuration of the two timer/counters.
https://deltamotion.com/peter/Mathcad/Mathcad - length_grader.pdf

L D[AR2,P#0.0] worked out how to solve the problem the simplest way even though he didn't calculate the equation but that is just busy work. This first method has the advantage that as second strip can enter the scanning area while the previous strip is exiting.

The second method really doesn't allow a second strip to enter the viewing area until the first has exited.

The third method uses the simplest way to configure the timers but the most complicated calculation not that is can't be done more than quickly enough.

I have no idea which method was used 20 years ago. I just know it worked.
Since we uses a single board computer, there were many options for configuring the timers/counters but a PLC may be limited to method 3 since it is the simplest hardware configuration.

I use Mathcad to solve multiple equations and unknowns. I don't do the busy work by hand.

I am also from the potato world. I remember seeing a post you made a little while ago about high speed cameras and food and I thought to myself I have probably worked on your equipment in the past.

Did you work on equipment by Kiremko or KEY or Tomra? They seem to be the big three (in Aus anyway) and I have worked with them all. My answer to the length is like other smart arses... get a strip sorter and it will tell you!

mp91 said:

I am also from the potato world. I remember seeing a post you made a little while ago about high speed cameras and food and I thought to myself I have probably worked on your equipment in the past.

Did you work on equipment by Kiremko or KEY or Tomra? They seem to be the big three (in Aus anyway) and I have worked with them all. My answer to the length is like other smart arses... get a strip sorter and it will tell you!

Click to expand...

I don't know anything about Kiremko or Tomra, Key is a competitor.

We are actually the outside R&D for the major potato processing company. This potato processing company actually owns the technology since they funded all the development. I don't know of any installations outside the US. I am not involved with the development of the potato scanning, grading and defect removal any more. There are other people that do that. I am mostly involved in motion control.