PID phillosophy

Hey Guys and Gals

Just want to spitball some things in regards to closed loop control.

If i were to ask you all the different between indirect feedback and direct feedback what would you tell me?

I want to chat a little bit about the different types of feedback specifically related to a some kid of process control.

Could be flow could be rpm could be temperature. Doesn't really matter

What i am further interested in discussing is the matching of units so to speak and do we really have to do that.

Lets look at a speed control for a second.

a motor accepts a 4-20 speed signal, the setpoint is rpm and the tachometer reports rpm back to the controller.


If someone asks me if that is direct vs indirect feedback i would say direct.

So the output of the controller is a signal directly proportional to RPM. It is ramping up and down partly based on the motor load.

The input is a setpoint in rpm obviously.

the feedback is also reporting back in rpm.

This is what i mean by consistent units.





Now the other case, Say we have an indirect control system.

Lets say for example a controller is moving a linear actuator. For the sake of the argument this linear actuator has the ability to place load on or off the shaft of the motor from the above example. (stupid i know but bare with me)

The controller is moving the actuator in and out and can increase or decrease the load which is also of course effecting the rpm

The feedback is still the tachometer.

If it were to be direct control i would say since the controller is controlling the stroke of actuator, the feedback should be an instrument that knows the position of the actuator stroke.


However, its not, its the tachometer. So we are reading RPM, and controlling an actuator.

How do these two scenarios effect our controller.

Questions....

Are both scenarios possible?

Considerations for each?

Why do this you ask?

Say you have a customer that does not want to buy an instrument to read the actuator feedback and wants to use the tach instead because they already have it.



I know this is a bizarre example but i want to stimulate some thought on direct vs indirect control.

Just for thoguht, here is another indirect exmaple

output is a 4-20 controlling a fan

The fan is blowing over something hot.

The controller is charged with controlling temperature by speeding up and down the fan

See attached file

Rscott:
When there are external variables that affect your process you can either use feedforward or cascade controller.

I have personally never heard of the distinction between direct and indirect feedback. But I live in a pretty closed world.

Your definition seems to focus on the unit or form of the control variable versus the process variable. If they are the same unit or type it is direct feedback, if they are different it is indirect. Given that criterion I would say most if not all of the control loops I am involved with are indirect.

The company I work for does a lot of web tension control. In our normal case the setpoint would be some amount of force in the web, the process variable would be the same thing but the control variable would be either torque or speed. So your definition would refer to this as indirect (I think).

However, what if we abstract the whole thing and we convert everything to percent. At that point I have a percent setpoint, a percent feedback and a percent correction command. Now the exact same process is direct feedback. In a very real sense this is what you had back in the old analog controller days.

If I am oversimplifying your distinction please take another swing at defining what direct versus indirect feedback really is.

Keith

Direct feedback as it was described, seems like some crude analog construction without even analog signal conversion. It's probably more expensive to source the components than to convert to digital.

Actually, to be honest, your post is perfect

You hit the name on the head

In your example, essentially almost everything is "indirect" by my first definition.

The key point is the process variable definition.

Say we are controlling temperature.

Yes, the output may control a jet burner or some kind of furnace or whatever.

That that output still has an effect on the process variable.

So then the question becomes,
What is the relationship between the signal and the process variable.

The 20 million dollar question is???

Do we really care what that relationship is!



Your example of percentages is a good one.


We dont really care that the output might be controlling a fan speed, the point is , that output of fan speed has some direct relationship to the temperature.


So, then the new question becomes, can we build a controller that is ineffective. Well the answer is yes , of course

Can you have a machine to where even with 100% duty cycle output, you can never reach setpoint. Sure, i can do it in my house by turning my thermostat to 150 degrees f.


So then the question becomes, how can the design engineer insure that the system is capable of the requirements

Of course, that different system to system. Bigger furnace smaller furnace.


So in short, it doesn't matter if the output is in fact the same "unit" as what we are trying to control. AS long as it has some effect on the process variable itself

Is that safe to say?

I feel like this should be a journal entry.

kamenges said:

I have personally never heard of the distinction between direct and indirect feedback. But I live in a pretty closed world.

Your definition seems to focus on the unit or form of the control variable versus the process variable. If they are the same unit or type it is direct feedback, if they are different it is indirect. Given that criterion I would say most if not all of the control loops I am involved with are indirect.

The company I work for does a lot of web tension control. In our normal case the setpoint would be some amount of force in the web, the process variable would be the same thing but the control variable would be either torque or speed. So your definition would refer to this as indirect (I think).

However, what if we abstract the whole thing and we convert everything to percent. At that point I have a percent setpoint, a percent feedback and a percent correction command. Now the exact same process is direct feedback. In a very real sense this is what you had back in the old analog controller days.

If I am oversimplifying your distinction please take another swing at defining what direct versus indirect feedback really is.

Keith

Click to expand...

When you think about it many closed loop systems are indirect. In power generation you have several modes where you are simply controlling the throttle of the engine, but you are controlling on KW export or Kvar, etc. Distillation towers in refineries have all kinds of indirect controls. Flows in and out of the tower cause temperature changes which affect the amount of light ends produced or the liquid level in the tower, etc.

The other 20 million dollar question,

Does the setpoint and the feedback always have to be the same unit?

I would say yes.....

Back to the example of a fan blowing over a hot element to keep it at a certain temp

You cant make the set point the speed of the fan and then claim the process variable is temperature.

Who agrees?

I have to crack open my college Process Control text book where the very first chapter have an example control loops and it doesn't use that the direct or indirect terminology either. I think you are mixing up some concepts

In your example:

"You cant make the set point the speed of the fan and then claim the process variable is temperature. "

Eh, yes but that's an incomplete PID where it will never get to PV to the setpoint. You will need to close the loop by adding a 2nd PID loops.

Outer Loop, PV=Temp, SP=Temp, CV=SP of inner loop (Speed)

Inner Loop, PV= Fan speed, SP= Fan Speed as set by Outer Loop, CV= Speed out

If you are going to use a single loop hen

PV= Temp, SP=Temp, CV=Fan Speed Out

rscott9399 said:

You cant make the set point the speed of the fan and then claim the process variable is temperature.

Click to expand...

Why not? Temperature IS the process variable, or the measured value. Measuring the speed of the fan motor will not tell you how hot the element is, and that is what you are trying to control in this example. You receive the feedback from the thermocouple, and then the fan's RPMs are adjusted according to the error signal that is produced.

I've never seen a concern for direct or indirect feedback. I have heard of direct and indirect acting controllers. These, in a nutshell, determine how the system determines the error signal. I think direct is the process variable - the setpoint = error signal. Indirect is setpoint - process variable = error signal. In your example, a direct acting controller, the temperature minus the setpoint would give you the error signal, the error signal would be used to control the speed of the fan to get the temperature back to the setpoint.

Aaarrrggghhhh!!!!?
Let snuff out that direct and indirect feedback stuff right now.
It is usually best that the feedback units are the same as the setpoint units but that doesn't need to be the case. In motion control the feed back is usually in position or angle units but some times we just want to control speed. There is nothing wrong or strange about that.
In the case of the fan controlling the temperature. This is an example of a cascaded loop. The fan in in the inner loop and controls the RPM with RPM feedback. The outer temperature loop has an output of RPM that is used as the set point for the inner fan control loop. There is nothing strange here. It should be straight forward.

You really should think of systems as being integrating ( type 1 ) or non-integrating ( type 0 )
Temperature and speed control are examples of non-integrating systems. Position and level control are examples of integrating systems. Feed back devices used for integrating systems can be used for non-integrating systems. For example, it easy to differentiate positions to get velocities.

There is no PID philosophy. Everything can be calculated.

I'm not sure the fan has to be on its own PID loop. I will say these processes can be tricky as they can possibly reverse themselves. If you are moving a fluid through/across something to cool it, the flow of the cooling media can get high enough as to not actually cool it enough.

Lets say your fan is running at 50% and the material is at the set-point. If you increase the fan speed to 60% the materials temp reduces. Works great. The fan speed is raised to 70% and the materials temperature doesn't change. HMMM? You raise the speed to 80% and the piece gets hotter. The process has reversed itself. If the PID was controlling it. It would continue to raise the fan speed and the temp would run away.

LoL, I was waiting for Peter to show up and straighten us out.

SD_Scott said:

I'm not sure the fan has to be on its own PID loop.

Click to expand...

You are right. It doesn't. It is just an example of how the inner and outer loops can have different units.

I will say these processes can be tricky as they can possibly reverse themselves. If you are moving a fluid through/across something to cool it, the flow of the cooling media can get high enough as to not actually cool it enough.

Click to expand...

Yes again, but I don't think your example as being "reversed"

Lets say your fan is running at 50% and the material is at the set-point. If you increase the fan speed to 60% the materials temp reduces. Works great. The fan speed is raised to 70% and the materials temperature doesn't change. HMMM? You raise the speed to 80% and the piece gets hotter. The process has reversed itself. If the PID was controlling it. It would continue to raise the fan speed and the temp would run away.

Click to expand...

I find this hard to believe. I can believe the cooling is non-linear.
Can you provide an explanation as to why there is less cooling at higher speeds? Is it a case of different heat removal capacities with non-linear vs linear flow?

Peter Nachtwey said:

You are right. It doesn't. It is just an example of how the inner and outer loops can have different units.


Yes again, but I don't think your example as being "reversed"


I find this hard to believe. I can believe the cooling is non-linear.
Can you provide an explanation as to why there is less cooling at higher speeds? Is it a case of different heat removal capacities with non-linear vs linear flow?

Click to expand...

I'm not sure it would happen in this situation with a fan blowing air on a a hot work piece. In some heat exchangers the coolant flow can be high enough so it doesn't have time to transfer enough heat. Yes it is cooling but the other part of the heat exchanger is being heated so it runs away. You can measure the outlet temperature of the coolant and see that once the flow is high enough it starts decreasing in temperature because its not in the exchanger long enough to carry the heat away. Once again it depends on the process and it would most like be known by the process engineers.

The process does indeed reverse because in normal operation a rising PID output results in a cooler hot side process and a hotter coolant output. Once a certain flow is achieved an increase in output results in a hotter hot side process and a cooler coolant output temperature.

Also if you use a VFD you can get the motor load and the speed through the drive if that is supported by the drive. If the fan is belt driven you can detect a broken shaft or belt. You can also add an airflow sensor.

Oh well, the crazy world of PID.