PID tuning in FlexLogix System

I am putting together a small rack to control water pressure in a plant H20 system. I am very comfortable with Logix 5000, but wonder if anyone could suggest some good reading or procedures for tuning my loop. I have never actually tuned one from scratch.

My system is using Flex Logix and RS5000. I am using the PID in Ladder style not function block.

Some quick and simple rules...

Use caution with these, as the system will dictate the response. But, when tuning PID from scratch and you either don't have the time to do the mechanical calcs on it (or there are too many unknowns to complete the polynomial) you can try this:

First, start with only Proporational gain. Keep in mind that most loops only require P and I. Get the gain set to where you can see some response that is appropriate for the differentail action you need. Then feed in the I coef. I is usually 1/10 what P is, just as a general rule. Do not add any D unless you have a "spring" effect that is caused directly by the mechanical system, not your own P/I calc. Too many engineers tend to go for the D to fix the poor tuning; but, instead end up with a less than responisve and sometimes over-responsive system.

See how that works, for a first shot. It usually takes me about 1/2 an hour to get something runnable using this super simple approach. (Time depending on the mechanical system - of course).

PS: Set limits in your code during this, to make sure you can't cause any damage!!!

You can practice here and not get wet!

Try these sites for PID loop tuning practice:
Number one
Number two

Patience is the key to loop tuning. You can not make the system run better than it is designed to, only more consistant than if you were doing the control by hand.

Good Luck!

?????? What is all this garbage? Is is urban legends or PID mythology?

I is usually 1/10 what P is, just as a general rule.

Click to expand...

Where is that written? You better have your stuff together before answering this. Don't tell us this comes from experience.

Do not add any D unless you have a "spring" effect that is caused directly by the mechanical system, not your own P/I calc.

Click to expand...

Where is that written? Again, have better have your facts, physics and control theory up to date.

Too many engineers tend to go for the D to fix the poor tuning

Click to expand...

Are you implying that using a differentiator is just a crutch for poor tuning?

; but, instead end up with a less than responisve and sometimes over-responsive system.

Click to expand...

Which is it? More responsive or less responsive? I prefer just right. I think just right can be achieved.

I found this in Randy's second link"

The first thing one should know about derivative is it should never be implemented in a flow controller.

Click to expand...

Never is a big word and rlong69 is not controlling flow. I have found the differentiator to be key to responding quickly to disturbances or following pressure changes quickly. I have BIG problems with those that have tuned 10 systems and call themselves experts when all the systems they have tuned are the same.

Rlong69 did you know that the gains will change depending on the set point? Why is that? Does anybody know?

I guess we are no longer interested in a simple approach?

Personally, my preference is for the engineer to perform the calculation based on the mechanical system. Afterall, the control system is really trying to "control" the mechanical process - whatever it may be (flow, pressure, cell growth, pH, temp, etc.).

However, I've found very few engineers, even ME's, who can appropriately complete the calculation for the method employed by the controller that they are using. Additionally, few even remember how to complete the series transformation in order to determine the actual mechancial system's response based on the nature of the system (e.g. viscosity, pipe size,

So, for our friend who is trying to just get a good start on tuning a loop, using the AB/Rockwell system (which approaches the Z-N method; but, doesn't follow when it comes to derivative or damnpening especially), I maintain - start with P, and as nessasary, employ some I. My suggestion with regards to D still holds as well. MOST loops that I've re-tuned have been critically dampened with D, apparently as a solution to a poorly tuned loop. Is it a crutch? Only when mis-applied. Is it nessasary sometimes? Absolutely!!!

When everyone is ready to "hang" it over the edge and compare size, I suggest we move over to flight controls. Forget PLC, DCS, loop controllers, and let's move into the real world.

So, our friend from WA is correct. You should know your system! But, if you can't calculate the coef, then you need to start somewhere....especially for your first shot!

Also, the gains entered are an approximation. Often, in truely well tuned loops 3 sets of coef can be used and you can perform a linear interpolation between. Again, we can get into the differential calculus that applies here; but, we were just trying to be helpful weren't we?

PS: After 12 years in controls, and tuning on tens of loops using everything from PLCs (Mitsi, AB, GE, Modicon, Koyo, Siemens...etc), DCS (Fisher and Landis) and C based systems, I might have a bit of a clue. And, I HAVE worked on flight controls....