Weight based small parts sorter

I am looking to build a parts sorter based on part weight. The range is 3 to 12 grams. I need 0.1 gram resolution. The load cell will be interfaced with an Arduino for the actual sorting. I need advice on components, code, etc.

We unfortunately don’t offer a load cell/force sensor that can measure a 3-12 grams range with 0.1 grams accuracy.
The smallest load cell we offer would be the 100g Micro Load Cell that you can use with the HX711 Load Cell Amplifier or the Load Cell / Wheatstone Amplifier Shield.
Another alternative would be to use a capacitive force sensor as the Capacitive Force Sensor 8 mm 1 N (0.2 lbs) with the Capacitive Force Sensor Electronics DAQ that can be directly interfaced with an Arduino Board as the Arduino UNO. But again, this capacitive sensor can measure loads up to 0.2lbs (about 90 grams).

Thanks for your answer. I assume you mean the capacitive sensor would not have the required accuracy in the 3-12 gram range.

In the page 2 of the datasheet of the Capacitive Force Sensor 8 mm 1 N (0.2 lbs), it is indicated that the force resolution is 0.2% of full scale (FS), therefore, the theoretical resolution would be 0.2% of 0.2 lbs which is 0.0004 lbs or 0.18 grams for a range of 0-90 grams.

I think that should work to differentiate between parts with a 2 gram weight difference. That’s the minimum weight difference with all the parts. Most are 4 grams or more different.

Thank you so much for digging into that for me.

It looks like all I need besides the sensor is the electronics DAQ package and my Arduino and I have a good starting kit. Obviously I’ll have to develop the mechanical structure and the devices to perform the actual sorts, but this should be a complete starting point, do you agree?

What did you plan to use as a sorting actuator ? Note that choosing the sorting actuator will strongly depend on the parts dimension and size.
If you are planning to use a linear actuator, you can take a look at our Linear Actuators category.
Since the parts you want to sort are not too heavy (12 grams maximum), you might be interested in the PQ12 Actuonix actuators.
Here is also our Servo Motors and Steppers category for your reference.
Also, another idea on the side, you said that you are looking to build a parts sorter based on weight, does the parts with different weight have different colors and shapes ?
If so, you might be interested in using the Pixy. Here is a brief video presenting the Pixy to have an idea. It can easily be interfaced with an Arduino board. If the parts have different colors and shapes, it can be easily detected and sorted using the Pixy camera.
It can remember up to 7 color signatures so that you can have up to 7 different parts with unique colors.

What I am proposing is a machine to sort brass cartridge cases of various sizes collected from shooting ranges. My initial idea was to sort by weight, however your suggestion of the Pixy opens more possibilities. The cases are all the same approximate color, but they all have unique shapes. It’s possible that a weight + shape detection system might give the highest accuracy.

What I have envisioned is to moderately clean the brass in a tumble wash to remove mud, dust, grass, etc. Then by a method yet to be devised, feed the cases one at a time to the sorting machine. This will present the case vertically, although whether base up or down is in question. Presumably Pixy could detect an object in either orientation.

At this point the case is standing on a long strip of Delrin or similar slick plastic, about an inch wide and four feet long. The strip is slightly inclined up away from the case. The case is supported inside a horseshoe or vee-shape piece of Delrin about two inches tall. By moving this part, we drive the case along the strip. Gravity will keep it inside the horseshoe or vee.

The case will be pushed onto the weight sensor and concurrently be examined by Pixy so that by weight and shape the particular caliber of the case will be determined (or not).

Then the pusher block will push the case up the strip (slanted upward only enough to ensure the case stays in place). Along both sides of the strip will be 2" diameter PVC pipes which lead down to containers for the sorted brass. Underneath the strip will be microswitches which correspond with each PVC tube pair (one on each side). Each switch can be numbered and one side of the strip is A and the other B.

If we have programmed our machine to place a certain case at point 8B, then when the pusher trips microswitch 8, a 100mm pneumatic cylinder will retract from the half-way position to move the pusher and case off the strip to the B side and drop the case into the PVC tube. Had it been 8A, the cylinder would have extended from the mid-stroke position to push it off the A side. After pushing off the case, the cylinder returns to half-stroke (probably determined by a micro-switch) and the pusher block assembly (which includes the 100mm cylinder) is returned to home to fetch another case.

The linear movement of the pusher block assembly could be accomplished by a motor driven belt system.

By using the weight to narrow the possible choices, Pixy would have an easy time of identifying the case.

Obviously many details have to be worked out and the devil is amongst them, but the principal of the system seems straightforward enough. It can be adapted to sort only a few or many different cases. The length of the track will have to be determined, but by being linear instead of a circular system, we have more flexibility.

I’m not sure of the commercial potential for such a machine. The speed will probably limit it to the hobby market, that being the amateur as opposed to commercial reloader. The cost to build might be low enough to generate some sales in that market, but my primary object is to develop it for myself initially.

What I am proposing is a machine to sort brass cartridge cases of various sizes collected from shooting ranges. My initial idea was to sort by weight, however your suggestion of the Pixy opens more possibilities. The cases are all the same approximate color, but they all have unique shapes. It’s possible that a weight + shape detection system might give the highest accuracy.

What I have envisioned is to moderately clean the brass in a tumble wash to remove mud, dust, grass, etc. Then by a method yet to be devised, feed the cases one at a time to the sorting machine. This will present the case vertically, although whether base up or down is in question. Presumably Pixy could detect an object in either orientation.

At this point the case is standing on a long strip of Delrin or similar slick plastic, about an inch wide and four feet long. The strip is slightly inclined up away from the case. The case is supported inside a horseshoe or vee-shape piece of Delrin about two inches tall. By moving this part, we drive the case along the strip. Gravity will keep it inside the horseshoe or vee.

The case will be pushed onto the weight sensor and concurrently be examined by Pixy so that by weight and shape the particular caliber of the case will be determined (or not).

Then the pusher block will push the case up the strip (slanted upward only enough to ensure the case stays in place). Along both sides of the strip will be 2" diameter PVC pipes which lead down to containers for the sorted brass. Underneath the strip will be microswitches which correspond with each PVC tube pair (one on each side). Each switch can be numbered and one side of the strip is A and the other B.

If we have programmed our machine to place a certain case at point 8B, then when the pusher trips microswitch 8, a 100mm pneumatic cylinder will retract from the half-way position to move the pusher and case off the strip to the B side and drop the case into the PVC tube. Had it been 8A, the cylinder would have extended from the mid-stroke position to push it off the A side. After pushing off the case, the cylinder returns to half-stroke (probably determined by a micro-switch) and the pusher block assembly (which includes the 100mm cylinder) is returned to home to fetch another case.

The linear movement of the pusher block assembly could be accomplished by a motor driven belt system.

By using the weight to narrow the possible choices, Pixy would have an easy time of identifying the case.

Obviously many details have to be worked out and the devil is amongst them, but the principal of the system seems straightforward enough. It can be adapted to sort only a few or many different cases. The length of the track will have to be determined, but by being linear instead of a circular system, we have more flexibility.

I’m not sure of the commercial potential for such a machine. The speed will probably limit it to the hobby market, that being the amateur as opposed to commercial reloader. The cost to build might be low enough to generate some sales in that market, but my primary object is to develop it for myself initially.

Follow up: it occurs to me that some cases are made of aluminum or steel and those need to be rejected. Perhaps Pixy’s color sensing can detect the difference between brass, aluminum and dark steel. That would be great and those cases go to the reject bin.

Pixy detects objects using hue-based color filtering algorithm. The objects need to have distinct hue for the Pixy to be recognized. Brass, aluminum and dark steel objects might work but need to be tested.
You can take a look at the Pixy’s Wiki for more information on the objects that Pixy can detect.