I built a machine that counts screws
Вставка
- Опубліковано 15 бер 2023
- My website and store: indystry.cc/
More about the project: indystry.cc/screw-counting-ma...
Patreon: / nikodembartnik
You can find all the files that you need to replicate this project on GitHub, it's open source so you can build it for free: github.com/Indystrycc/OpenScr...
Instagram: nikodembartnik
My gear:
Camera: amzn.to/3VdbfGH
Lens: amzn.to/44nY2z0
Better lens: amzn.to/424iuCX
Tripod: amzn.to/3Nh2Cca
3D printer: bit.ly/43SS7S5
Label printer: amzn.to/3oW55yn
Headphones: amzn.to/3LDTXz6
Speakers: amzn.to/3nl34eJ
Oscilloscope: amzn.to/3HmGMjN
Lab bench power supply: amzn.to/3HpotdD
Soldering station: amzn.to/3Vd39xH - Наука та технологія
That's interesting. Generally in these cases the parts are counted by their weight.
This would be great as a secundary check. Let the machine count, add some clips so that the screws are ejected in a baggie. check weight and ship
Yeah after doing inventory for things like this, Weighting them make it so much easier.
I came to comment pn weight, but was also thinking Displacement?
@@moxl_ exactly, and reject and under/over weight bags based on a sepcific tolerance to account for variance in the weight of each part. just how they do in food production factories. my parents both work in a factory that produced chocolate cake products and each finished pack of cakes had to be within tolerance otherwise they would be rejected off the production belt automatically.
Most scales have a piece counter. Hit the button, place one on the scale, press the button, and now it will do some simple division and count how many you place on it. I think it was designed for jewelers. Counting these by hand is nuts haha
Have you considered using a weigh scale to count the screws? Our local hardware store does this and its usually pretty much spot on
LOT OF PARAMETERS DEPENDS ON LOAD CELL
they work great for electronics too, I think the smallest weights you can reliably count without a cover is about 0.02 grams but dirty/dusty parts can mess it up
also things that are in a plastic package get counted wrong all the time
@@JobinPThomas Huh? Just get a scale that is accurate enough, and a print the optimal scoop.
Most places uses scales for this stuff, as they weigh enough but are consistent enough to count using them. If you know each screww is 22g+-1g. 264g is 12 screws. And you know that if the weight is between 252-276g will be 12 screws
@@JobinPThomas we used it for a magic trick where a mentalist speed counted buttons. And they are even lighter than screws, works flawlessly and fast.
My 2 cents, from my experience - Hobbyist grade inductive sensors have a hard time sensing small parts, but since you are only counting 1x type of part you might get away with it.
Optical sensors, you need to restrict the beam width as much as possible (small holes on both sides of the sensing area). Parts WILL find some orientation that bypasses the sensor when falling. Unless you can perfectly align the part.
I wanted to count different types of parts, some plastic parts weighing less than a gram, and I didn't have much luck with cheap load cell. The solution I found was simply dropping the part onto a suspended piezoelectric disc, starting off by simply hanging the disc in the path of the falling part using the wires soldered to it.
Here's what you do, you have the machine count the screws, verify the first batch, weigh the batch and then just weigh each batch as they come out.
If the weight is off then add the missing screws and if it's not you can move to the next one.
I would also suggest having it depositing the screws in the final bag directly.
On the other side of things you could have the machine drop screws onto a scale until the weight is correct within a tolerance that way you'll have 0% error rate.
Yeah. The difference in the weight is far smaller than with the actual count!
why not add this part in to the machine?
Adding a little scale in to the output and let the machine count the weight.
@@MARkoe93 Which is one of the things I mentioned in my comment no?
adds more sensors and steps... the design he has is plausible it just needs some materials, sensors, and geometry to better the design of his prototype
Did you think of using an inductive sensor? Also you can weigh outcome set of screws to quickly confirm it. Or extrude screws on a scale and give feedback to the machine.
Love the idea of an inductive sensor! I wonder how much his photoresistor is affected by ambient light noise. An inductive sensor wouldn't have that same problem.
Throughout the video i though to myself why he didn’t choose a inductive sensor it would be a much better solution for accuracy but only when you run the screws down a small enough tube next to the sensor.
@@thomashepi2674 yh should only not work if they are stainless
You could probably add a scale for validation, use it to detect a weight change
A contact microphone could also work, but neither the scale or the microphone can really be used alone due to things like bouncing/colapsing screw piles
If you're packaging ferrous screws you could tie in a hall-effect sensor (and probably a magnet)
Hi Nikodem, you are trying to detect a small object that is travelling quite fast, in a variety of different orientations. I think accuracy would be improved if the counting was carried out at a low speed part of the operation.
My suggestion is to have a light source on one side of the disc, and light sensor on the other. You can count either empty holes or full holes. This would require some form of indexing, such as a second light sensor to identify when the detection should be done. Alternatively, use a stepper motor instead of the little gear motor.
I agree. Alternatively you can use a inductive sensor, and have it placed behind the hole ring on the disk.
Was about to say this. Detect screws in the wheel as they're being brought up. This is a much more controllable environment.
Make a door that the screw has to push open as it falls. On the hinge side, there is a tab that acts as an NC (normally closed) switch. When the screw falls through the chute it pushes the door open, pushing the tab in the opposite direction and breaking the circuit. The door then falls back closed via gravity, closing the circuit. Or, as others have suggested, do it by weight, but I like your design!
This could be a bit tricky due to the low weight and size of the screws
A small door (20x10mm?) made from a piece of aluminum sheet around 0.5mm thick should be light enough to open with minimal force, if hung plumb. You could even cut it out of a soda can. I’m confident it would work reliably.
Maybe you can add a second or a third sensor and have them all count the screws. Add code that compares the results from all 3 sensor to clean up the error rate.
you basically DIY'ed a "vibratory bowl". Used often in industry for things just like this. nice work and good idea!
this is the first video i watch from your channel. I found your progress of making the machine is great! because you show how you encounter the problems and solve it. I love this kind of content. please keep it up!
That is really interesting. Have you considered a small load cell, and counting by weight?
not exactly going to be super precise imo. there is some variance in weight per unit measured and those stack up over time. when you have really small objects and only a few it can be off too much. load cells also have a tolerance. we are talking about 0.37 grams here for washers and around 3-4 grams for the m5-20 bolts.
@@jasonbenjamin1464 But when you are providing these screws as part of a kit for your customers to assemble, you anyway want to provide them a few extra to be sure that a single one with a bad thread doesn't have the customer complain. So, using the weight and ensuring it is N x (nominal weight + safety margin) seems the good idea to me.
Fajnie widziec, ze po tylu latach dalej dzialasz.
About a year ago, I was looking for videos on screw counting, as I was making a device to put x amount of screws in a cup. Back then, there were like, 4 videos on diy screw counters. In the past 12 months, I've seen more screw counting videos than I could count. Awesome stuff, and I love to see what people come up with
Thats some really good use of engineering, and the fact that you made your own PCB is cool too. Sadly I do have to agree with the others on the use of a weighing scale. This is an overengineered solution. But I hope you learned a lot of things that you can apply to your other projects!!
This is pretty cool. I think you should keep iterating on it. Maybe take a look at some other designs to see what changes you can make for the mechanism to be more reliable.
Building tools is always nice. Thanks for sharing your journey with this project!
Interesting project. As an alternative for a similar future project, for the screw detector, maybe try using a piezoelectric transducer and allow the screws to "bounce" off of it as part of the exit path. This will generate a small pulse that can easily be amplified. Also, for very accurate counting, have you considered just using a parts counting scale? I use one all the time for accurately counting even very light pieces such as 75mm x 1mm zip ties.
I needed such screw counting machine when making the ERC Timsav Foamboard Cutter for the community. But I don't have time to develop anymore. I just make a jig that makes counting easier and also easier dropping the screws into the bag.
THanks for sharing. Good filming btw! Replace the sensor with a weight detector that interacts with the rotation and you'll be good I think.
Inspiring project, seems like a lot of fun! 😄
Thanks Benjamin!
Amazing 6 categories of engineering in all! Very cool.
We had a scale that you placed like 10 on and you could see how many you placed in a box, worked wonders for m4 bolts m20 bolts, screws, deck screwes
From a production view you should add some extra screws which help with the error. 5 extra screws are way cheaper than a person counting it. IKEA, LEGO and other brands always add some extra
A tool I've seen in a pharmacy and a factory is a tray with a grid of pockets that fit one pill/part.
You shake around the pills in the tray until all the pockets are full, then tilt the tray to dump the excess out of a spout in the corner. Then you verify visually that all the pockets are filled and flip the tray into a bin. It's more manual but easier to manufacture.
As many other comments already captured: I'm very surprised, that you didn't use the method of weigh counting, since it's super simple and reliable.
Cool project! Did you consider counting based on weight? I'm curious if that would be more reliable.
lol. Definitely.
I like this because I wouldn't have thought to do it that way.
I guess for small quantities it would be a pain to manually weigh screws. So, I think I would add a scale to the project. If the weight is right, dump the screws into a train of plastic bins, else dump on a conveyor belt to return them back to the the tumbler. Since it is often right and fairly automated it won't matter if it takes a little longer than it could if it worked perfectly. When there are enough full bins in the bin train, use a custom funnel to transfer them from bin to bag.
Habe you thought about using an inductive sensor (like some 3D-printers use to level the printing bed)? This would allow to only change the big wheel if you want to switch to counting also smaller parts like washers or nuts. The quality of the sensor signal is also better, too.
Instead of checking manually several times whether the machine is wrong or not, it would be enough to use a scale. weigh 20 bolts and compare the weight after the machine has worked a cycle. and this system could be built into the machine itself. so that a green light would simply light up when the weight matches the previously set weight and a red one when it doesn't.
We have built this kind of machines before. A few tips. Lower the speed and reduce number of pockets to half, so it has time to stop preventing getting an extra screw before it stops. Make sure to have a "debounce" time on you sensor so you do not count a screw 2 times as it falls past. The sensor you are looking for in a inductive sensor or read fork sensor. But inductive is preferred. Make sure to get a sensor with enough read distance. This will solve it. Never have to count screws again.
Właśnie tego potrzebowałem do pakowania moich śrubek w zestawy! Bardzo przyjemnie ogląda się twoje filmiki! Pozdro!
Fantastic project! I would try to put a scale (100g or 500g load cell) under the box and use the weight to verify the correct amount of parts. The first part triggering the optical sensor will fall into the plastic box and will give you the first weight value, which can be used by the programm to write the weight per part variable.
This was a great video. You should also check out hall effect sensors/optical and scales as a possible hybrid solution. Also make sure you are using interrupts or very(very) low cycle times(or of course you could go wirh a latching circuit and reset on read). Best regarads.
Some things to consider:
1. Tumbling the screws in this manner but in a bigger quantities when they have a protective layer will wash it away and the customers won't be happy
2. Count them using better designed gravity based aligning section. Search for automatic screw feed or counting groups spinning machines
3. Weight based counting? Just air clean the screws before entering the scale
Great job. Excellent example of how to work out solutions.
Awesome job!
Nice project. Congratulations.
Love the chatGPT thing for the seven segment .... powerfull tool, if you understand that is there to save time, but will never replace skills that you need to come with this brilliant ideas.
Cool project and effort. Just as other comments stated there are small table top scales specially made for counting and display count and weight simultaneously.
Sometimes the best design is no design, rather find out it is already done.
Nice work. I did takeaway some ideas here. Consider a counting scale for production tasks. Thanks from Colorado.
Super cool design! Have you considered a small load cell? Measuring the weight increase each time a screw lands at the bottom of the chute would be quite reliable I think.
You could steal one from some cheap digital scales designed for sub 1kg measurements.
Nikodem: "Secret" usable sensor is called *SCALE* (skala) It's very accurate, works for a wide range of screws and even includes final check: _TOTAL_ _...but nice vid as usual_
As suggested weith at the end will be very precise because of each screw does weigh a lot. pottiing in a bag directly is the second. Third I would suggest putting them a second time in the machine, but the weight will do it for sure. Last I would suggest having the output shrank to a minimum so that there is a minimum space for the screw to get out and min space for those sensors. If you have too much space, you might miss some. Those are my suggestions. I play also with esp and esp32. Great project !!! I am also almost sure you could connect to a scale to decode.
The solution is adapting the output chute to orient the screw straight using some clever geometry, the chute needs to be a slightly larger diameter than the head of the screw and get a laser optical sensor for a larger signal change when sensing, and the chute near the sensor needs to be a matte black material reduce reflections that could cause inaccuracies. And adding a very short retrigger delay to the sensor causing a counting event to then further the count accuaracy
you would love brass threaded inserts that can be melted in easily with soldering iron. Very clean and you can use smaller holes as opposed to nut that has a big outer diameter compared to inner thread diameter.
Weight is definitely the way to go for something like this but your screw capture/conveyer system is really cool
Although there are other ways to count screws, this it's an interesting project. Congratulations!
Add a flap to the output chute that when it moves will trigger an opto-interrupter logically AND that output with the screw detector to maybe get a more reliable count?
I've worked in factories where you sit all day sorting between pass and fail components coming off of a vendors line, and every so often you have to count up the pass components. Most of these are done by using a counting scale. Tarr the scale with an empty bin, drop enough parts into that bin to give you an acceptable accuracy by weight, say 10 or 20, tell the scale how many that weight is, and dump the parts to b e counted, nd the scale tells you how many parts are in the bin. I believe they used a load sensor, as it was a digital scale, but if you are counting out a consistent number of parts, you could use balance beams to do that. In either case a cup gets set on a scale, pargs get dropped into it (you may also be counting them as part of that drop, then when the expected number show up in the cup, that cup moves to the bagging station. If you are counting using two methods, you can very likely move on to doing QA checks ever 5 or 10 or 1000 bags depending on your confidence goes up.
One other possible solution would have been a capacitive sensor. That allows for metallic and non-metallic parts. You will have to account for the dead band, the sensor cannot be too close to the object being detected, and there may be some improvements that you could make in the firmware. If you are not familiar with them, look it up. There are tons of information all over the web.
Overall, it's a great initiative and congrats on rolling up your sleeves and actually doing something. Perfection is never being truly satisfied with the results and searching for solutions for improvements, buy also knowing when to stop. Keep up the good work, challenge yourself, keep records of your progress and create some kind of version control.
How about using an induction sensor to detect screws? Those are generally extremely reliable and their sensitivity can be adjusted. Also add a cheap weight sensor with i2c to communicate with the RPi and double check if the weight makes sense.
Your photoresistor is totally like the same ones used in smoke alarms, neat as!
As always, it's interesting to watch you tackle an issue you are facing in real life. It seems like a good design. I'm sure you'll get the bugs ironed out in the next version.
I think you can improved the results by simply .. using IR sensor ( break beam) looks like you are using photo (DIY) break bream,
Also make the IR beam smaller, make the funnel when the screw travels smaller to insure the sensor gives the best result.
Nice results so far !!
There are inexpensive digital weighing scales that has Clint function. You can program it quickly and count screws.
We have a similar problem as you have. We're basically assembly hardware bags, but each bag contains several different bolts / hex nuts / hex key etc. To be honest we didn't find a solution to the problem that would actually speed up the manual process. When packing bags manually it takes us around 10 sec per bag. Having that in mind and a machine that would do the counting, we would spend more time transfering counted screws into the bag, than we would actually spend packing them manually.
As much as this method isn't the best one for counting it's a fun way to solve the problem, I kept thinking that attaching a contact microphone to an angled surface and bouncing the screws of it would have been very reliable.
You will change the world young man! Genius!
instead of a photoresistor and an LED, what about a hall effect sensor with a bias magnet?
they can be used to count the rotation of gears, so i suppose they are quite reliable to detect fast moving ferrous metals.
as others said you could also add a load cell under the container that holds the counted nuts/bolts as a sanity check.
or maybe a small piezoelectric speaker under the container that receives the screws to sense the their hits when they fall. this might get way to unpredictable with lots of screws moving around though
If you count to many you schould add a small cooldown after detecting an object. This eliminates counting a screw more than once. You schould also think about weighting the output to verify.
What a super cool project. Do all screws have the same weight with a tiny error rate? If yes, you can add a weight checker to the screw counter :)
Measure weight with a scale. Do it in chunks so that the error rate is minimized. 1 screw weighs x grams, 10 x =, done, verify. Another way is to watch for the weight vs time and sync with the optical sensor. Adding a Schmitt trigger might help too. Optical sensors are tricky with unusual shapes. One last possible attempt would be to install your optical sensor just before the drop, while the screw is in the "lift" and more constrained in shape/orientation.
I like your attempt and commend the efforts!
You should use a hall effect sensor, if you get a good one you can sample at a very high frequency and have it work 100% of the times
postage scale works just as good.
used it in printing to weigh stacks of paper to make reams. also for screws in an assembly job.
Cool project. I've done something similar only I measured the weight of the screws (or whatever) until it matched a known amount. I just pored them in :)
For fast and accurate counting, count by 3. 3 is a number you can take on sight without really having to 'count'. Do this 3 times for a total of 9, this you can again do without having to 'count'. Now take 1 and put this apart on a separate pile (each 1 will count for 10 later). Repeat till you run out of parts. Note the last quantity lower than 10. Now repeat this for the parts you put on a separate pile (the 10's), take 3 x 3 parts and then put again 1 aside on a separate pile (each 1 will count for 100) Keep repeating for the 1000's, the 10.000 till done.
This system works well once you get the hang for it, does not require any equipment and requires no real concentration since you never count anything bigger than 3 x 3. No problem if your phone rings or somebody comes and bothers you.
Maybe some sort of little bucket where one screw falls in and rest. So the machine can detect it easily. And if it detected one screw it opens the bucket so the screw can fall out
Countin screws is usually done by measuring wight instead of actually counting them. But your machine has a potencial as a screw feeder. If it outputs all screws in head up direction than it is more useful machine.
The world needs more bright minds like you
Nice project u can improve the sensor by reducing the rectangular hole so every screw cut the beam , use infrared led and infrared photodiode to get ride of ambiant light error.
Have you considered automating the bagging part of this project as well? Interesting concept
almost all digital scales have the count mode you usually count some items together for example 100 items and it calculates the weight of the single one with very good accuracy, its the perfect case for bolts and nuts, If the method is accurate enough for (bulk) diamond sellers and jewelers to do that, should be good enough for screws and bolts. you could do this using a $10 cheap eBay digital scale, but I'm guessing it wasn't good enough for making a video and putting it on youtube!
Great material. 💪
May I ask you which software are you using for editing your videos and add text effects? It looks awesome 😊
it's Davinci Resolve
get a small loadcell and weigh the output to verify the results. Bottom line is you can do it all via the loadcell but having two technologies that check each other does eliminate the manual check. Use an HX711 module which has a 24bit A/D which is perfect for counting parts. You should use a 500g or 1kg loadcell depending on the maximum number of parts that you're counting vs the minimum weight. As an indicator, I'm using a 5Kg loadcell with an accuracy of 1g (0.5g if I push it) so the 500g and 1kg will give really good accuracy. One of the reasons you could still use the light source is that you are counting before you put the component in the container. If you use the weight as the primary method of counting then you could go over and then you'd need to find a way to remove the part. There are of course ways to make it work accurately but this would solve your problem. Good luck
You could print a basic plane piece with exactly 10 screw-shaped holes (or whatever you need) and fill it up in a box full of screws
Have you thought about weighing as a validation technique or even as an input to the controller?
you need to make a tinny slot ahead of your led, it makes a narrow light beam, and reduces reflecting from metal screws
You could add weighting and conveyor belt, so it would be great automated device
Did you consider making a vibratory feeder bowl?
Or perhaps just get a scale that's accurate enough to weigh them?
There’s an app called “count things” take a picture of a pile of screws spread out, and it will count and label each screw you can then add dots if it missed anything
Use a small digital scale to validate the results instead of counting manually. Weigh a batch of 100 screws, tell the machine to count 100 screws, weigh the result, rinse and repeat until you are confident it is counting accurately every time.
nice work
Glad you went with this approach. You would've gotten in trouble if you made a machine that screws counts.
you could make a "shelf" that the screw will hit when it falls out of the back, and use a piezo element and op-amp to detect the pulses you'll get when the screw makes contact. if you don't have a piezo element, you could probably get a similar result with an electret microphone. you might get more than one if the screw hits both ends, so you'd need to make the code smart enough to say all pulses within X millioseconds are counted as 1 screw
with your current design, you could make the part where your detector is go all the way down to just a little bigger than the screw size, so the screw will go across the sensor in the long direction, so you'll have more time that it's blocking the light from the sensor, or maybe you just need a second sensor to catch the 5% that fail
in industry there is a sumilar application which we call it as a biabrator feeder wher it can count reliably.
you can upgrade your device to that
good luck
good project
Would be very interesting to know where the error was coming from. Is it from the sensor part? Or does the mechanical part some times deposit multiple screws at the same time?
Check out Christopher Helmke's videos on a screw packing machine. He uses a magazine to sort the screws and then a sprocket to count the screws.
Maybe a simpler variation might be to sort the screws into a magazine, them manually dispense them based on their height in the magazine.
Seems like you having a grounding issue on the outlet your monitor is plugged into. I noticed its turning on and off. You may want to consider an AVR and make sure your ground wire on your LCD is working/connected.
great job, I also want to try making one. However, when it comes to counting standard parts, an accurate scale should be more convenient.
You could simply augment this by adding a little flap in the outlet tube where the light sensor is. Make the flap so that is open with even the slightest pressure so that a screw can pass through, But the flap will completely block the LED or the Sensor every time. This way the screw can't slip past. Then I would ensure that there is no external light making it's way to the sensor at any point.
Have you thought at using a switch with a long lever: When the screw falls on the lever it trigger the switch and add to your counter, then you shouldn't have any error?
If you know the percentage of error you know how many extra screws you have to set to always give the right minimum amount
I just watched a video by Christopher Helmke where he used sprockets to count screws, he has a very interesting modular production system that is worth a look at. Cheers
I would have put a trapdoor in the exit shut that would count the screws as the fell. (I know that mechanical solutions have a shorter life-span than optical, but for these small parts I would expect that it would be more accurate.)
The other option is to just weigh the screws, because they are made with such high precision.
Really good seeing that PCB done without using a third party
Try adding somekind of weight verifying system.. Basicly after the machine is done counting it should check how much do these screws weight and therefore check whether there is too much or too little. Ofcorse there will still be some margin of error but definitly not as drastic.
Thanks for open sourcing this. I love little projects like this.
Happy to do so! Thank you!
In this case, it is necessary not only to measure the weight, it will have a low accuracy. A small mechanism is needed that measures the weight of ten or twenty bolts in a small container, and this small container, after measuring the weight, is tipped over into a large container.
This will be much more accurate and hundreds and thousands of elements can be measured in this way.
A. that's impressive
B. anytime you can make something that is also helpful for business, double win! I know you'll keep working at it until it's accurate.
If the screws are ferrous (or maybe not?) what about a magnet across from a hall effect sensor? Would the screws falling by disrupt the field reliably enough to count with?