Should I use a PID Loop?

[Cody Anderson]

I am not sure of the exact amount of time, but the peanuts do fill up. So there is some time the flow can stop for but it usually runs as a constant flow. I believe this time is for emergencies if the peanuts were to stop flowing. Thinking like 20 to 30 minutes max. But the system will mostly be running all shift, just stopping every so often as the photo eye sees it.

 

[drbitboy]

Is the actual target an instantaneous value (say, 10-20s) or is it over the current "run" or "batch" or other longer period?


E.g. for the latter, "25% small" means that, over the current shift, we want to use three times as many bags of large as of small (assuming all bags are about the same weight).

If it's the latter, you could keep track of the total of each put in from some base time, and simple increment or decrement the vibrator a bit with changes in the overall percentages, which percentages should be more accurate, and less noisy, because of the longer time base.

 

[Cody Anderson]

It would be the latter, so you would want 25 percent smalls in the whole shift and 75 percent bigs, or whatever percentage they choose.

 

[drbitboy]

Would it then be acceptable to empty three bags of large peanuts, while the small peanut vibrator is turned off, then one bag of small peanuts while the large peanut vibrator is turned off, and repeat that sequence over the entire shift and say the system delivered ~25% small peanuts?

 

[Cody Anderson]

I do not think that would work or be acceptable. I don't know the process of blending peanuts very well, but from what they wanted, they want a constant flow of the blend not really one than the other. I wish I could do it that way, it would be super easy lol!!

 

[drbitboy]

Okay, but in previous post you said that it did not have to be instantaneous (continuous) 25%, so the question is, over what time frame to you need it to be a measurable 25%?


E.g. when a bag goes empty, it is 100% one or the other, and that seems to be acceptable because it's over a short time frame (a minute or two?).

 

[parky]

Just remembered a project I did some years ago, this started just as one loss in weight for feeding uncooked rice at a given weight (tonnes/hr), however, it was expanded to do the same for one of the other ingredients, this was a continuous feed system the rice bags sat on two hangars with load cells & fed into a sieve & then blown into a continuous rotary cooker, only one bag would be active (one in use other for instantaneous changeover), this was controlled by a PID loop (cannot go into the mechanics of how the feed was controlled as may still be a secret), this was an existing system not done by me, however, I modified it to incorporate load cells for a loss in weight of one of the other ingredients, this too had a PID it worked extremely well, the system would stop for short periods (not the cooker just the feeds).
Over time, we trended the feed systems along with the buying dept came up with the typical accuracy of the blend, this turned out to be within about 1.5%.
The only problems I think you may have is if there is vibration from the system that could affect the load cells may cause fluctuations in loss in weight but can be smoothed out & the speed of recovery of a PID on a start.

 

[Manglemender]

Coming late to this conversation but I would model the system. Build your self a lookup table for the vibrating feeders of setpoint vs output. eg, 1% = xxxkg/H, 10% setpoint = xxxKG/H, 20% = xxxKG/H and so on. Do this for both Bigs and Smalls. 11 points is really the minumum you should considder and you can then do a linear interpolation between the points to determine the predicted setpoint for any given dose rate. Now using your lookup tables, you can predict the required setpoints to deliver the required quantity and ratio. You can then add a correction to the setpoints based on feedback from the loss in weight system if needed.

Nick

 

[parky]

Manglemender is right, if the feeders do not give a linear feed from x% to 100% (not included 0-x as this will almost certainly be well out) then use a lookup table to adjust the values, I did not have this problem, however, I did a similar thing on some silos that were a very odd shape, in that case I used an array of 100 this was ideal as the silo levels were 0-100%.