PID Tuning and Importance
- Thread starter m_abd_el_khalik
- Start date
Hello; dandrade;
I think you are looking for a "Feedforward" control algorithm, where you try to anticipate (known and measurable) disturbances on your process. Many PLCs offer this type of variation on their standard PID control.
Hope this helps,
Daniel Chartier
I think you are looking for a "Feedforward" control algorithm, where you try to anticipate (known and measurable) disturbances on your process. Many PLCs offer this type of variation on their standard PID control.
Hope this helps,
Daniel Chartier
dandrade,
its called "feedforward" or "disturbance".
The S7 software PID block SFB 41/FB 41 "CONT_C" has a "DISV" input for this purpose.
There must be some hardware PID controllers with this feature, but I dont use them so I cannot point you to any.
its called "feedforward" or "disturbance".
The S7 software PID block SFB 41/FB 41 "CONT_C" has a "DISV" input for this purpose.
There must be some hardware PID controllers with this feature, but I dont use them so I cannot point you to any.
Peter Nachtwey
Member
Control Algorithms with two inputs
Our company makes a motion controller that often takes 3 inputs. One position feed back the pressure feedback from each end of a hydraulic cylinder. This way we can seamlessly switch between position and pressure/force PIDs or do position/force limiting.
In animation the force across a piston are differentiated and fed back to the position PID this reduces or eliminates oscillations. After all, you don't want your kids to see characters that look like they have Parkinson's disease to you? Our competitors make a big deal about that. My response is that I can precisely calculate the spline or cam profile and limit the accelerations so the desired motion trajectory is smooth with out resorting to differentiated force feedback. This can save $$$. This is how the boat in "Master and Commander" was moved during the storm action. There are some cases where one may not have complete control over the motion profile. An example would be using a joystick as a reference position. In this case one can either filter the joystick input so the acceleration does not exceed the systems acceleration capability or limit the acceleration using a differentiated force feedback.
I recently help on an application where a log is moved through saws using hydraulic motors to grip each end. This required a trick where the lead dog is geared to the pushing dog and yet the force applied to the log by the top/lead dog has its force reduced. This way the bottom/pushing dog has enough excess force capability to push the log through the saws. We do this by putting pressure sensors on the ports of the hydraulic motor. This allows us to calculate the applied torque. Other common applications include presses, injection molding machines and drilling.
In motor applications, such as grinding, we can grind to as certain depth but still limit the torque on the motor.
There is another thread about a winder. This application can use position feed back from the roll motors AND the dancer between the rolls.
There are many applications that require position AND force feed back.
Feed forwards DO NOT require another feedback.
dandrade said:
Our company makes a motion controller that often takes 3 inputs. One position feed back the pressure feedback from each end of a hydraulic cylinder. This way we can seamlessly switch between position and pressure/force PIDs or do position/force limiting.
In animation the force across a piston are differentiated and fed back to the position PID this reduces or eliminates oscillations. After all, you don't want your kids to see characters that look like they have Parkinson's disease to you? Our competitors make a big deal about that. My response is that I can precisely calculate the spline or cam profile and limit the accelerations so the desired motion trajectory is smooth with out resorting to differentiated force feedback. This can save $$$. This is how the boat in "Master and Commander" was moved during the storm action. There are some cases where one may not have complete control over the motion profile. An example would be using a joystick as a reference position. In this case one can either filter the joystick input so the acceleration does not exceed the systems acceleration capability or limit the acceleration using a differentiated force feedback.
I recently help on an application where a log is moved through saws using hydraulic motors to grip each end. This required a trick where the lead dog is geared to the pushing dog and yet the force applied to the log by the top/lead dog has its force reduced. This way the bottom/pushing dog has enough excess force capability to push the log through the saws. We do this by putting pressure sensors on the ports of the hydraulic motor. This allows us to calculate the applied torque. Other common applications include presses, injection molding machines and drilling.
In motor applications, such as grinding, we can grind to as certain depth but still limit the torque on the motor.
There is another thread about a winder. This application can use position feed back from the roll motors AND the dancer between the rolls.
There are many applications that require position AND force feed back.
Feed forwards DO NOT require another feedback.
????????? Is limit concept, more feedback->multi-variables process.
Obliged the answer,understand good, trink do not exist to level fisic sometime algoritim. I have document, that lists 12 variations of PID and another of methods models.
They are tables that relates: Kind of variable, intensity of disturbe, relations between you varied and application.
Make, it produce (we, members) this turned out a table with the variantions of PID and its caracteristic, easy of decide which PID utilize Yours talk? (Mainly, not required math compresion-Hi rsdoran)
Obs: This board is of PLC,,, automation would must himself another (I went trink)
Tom Jenkins
Lifetime Supporting Member
dandrade, feedback and feedforward are two fundamentally different concepts.
Feedback measures a process variable, like tank level, feeds it back into the control algorithm (calculation or logic) and determines the error. It then manipulates a variable, like flow from the tank, to decrease the error.
PID is just one of many kinds of feedback loop.
Feedforward measures a factor that will disturb the process, like flow into a tank. It then manipulates a variable, like flow out of the tank, to maintain a process variable. The manipulation is based on a known relationship between the two. Flow out = flow in to maintain level for example.
You can use feedforward and feedback together as well. For example, you could "trim" the flow out based on a level measurement to allow for unknown or annaccounted for influences (a leak in the tank, for example.)
Feedback measures a process variable, like tank level, feeds it back into the control algorithm (calculation or logic) and determines the error. It then manipulates a variable, like flow from the tank, to decrease the error.
PID is just one of many kinds of feedback loop.
Feedforward measures a factor that will disturb the process, like flow into a tank. It then manipulates a variable, like flow out of the tank, to maintain a process variable. The manipulation is based on a known relationship between the two. Flow out = flow in to maintain level for example.
You can use feedforward and feedback together as well. For example, you could "trim" the flow out based on a level measurement to allow for unknown or annaccounted for influences (a leak in the tank, for example.)
Peter Nachtwey
Member
Disturance Rejection or Feed Forward?
Actually, that is disturbance rejection. Feed forwards and disturance rejection are related. The main difference is that feed fowards are dependant on the SP and not on a measured value. Both techniques can greatly reduce errors during a dynamic process.
Feed forwards use the SP and its derivatives to calculate an estimated output. There is no measuing involed. In motion control the SP is the target position and the derivatives are target velocity, target acceleration and target jerk. Each of these terms is multiplied by a feed forward gain which is derived from the plant model. The result is an estimated output that should be very close to required output to achieve the target velocity, acceleration and jerk. Now the PID is only required to adjust for changes in load and non-linearities. Feed forwards ARE THE KEY to minimizing error between the target and actual position. Disturbance rejection is also very helpful in many process applications where there is time to react to a distrubance.
Tom Jenkins said:
Actually, that is disturbance rejection. Feed forwards and disturance rejection are related. The main difference is that feed fowards are dependant on the SP and not on a measured value. Both techniques can greatly reduce errors during a dynamic process.
Feed forwards use the SP and its derivatives to calculate an estimated output. There is no measuing involed. In motion control the SP is the target position and the derivatives are target velocity, target acceleration and target jerk. Each of these terms is multiplied by a feed forward gain which is derived from the plant model. The result is an estimated output that should be very close to required output to achieve the target velocity, acceleration and jerk. Now the PID is only required to adjust for changes in load and non-linearities. Feed forwards ARE THE KEY to minimizing error between the target and actual position. Disturbance rejection is also very helpful in many process applications where there is time to react to a distrubance.
Peter,
You have got it all wrong. I did not use ANY PID instructions in this blender program. The PLC progream merely looked at the weight of dust in the blender, and calculated how much water should be added, then opened a valve and measured the water until the flowmeter showed the correct gallons. No, the model did not assume perfect mixing, but only 90% had to be mixed to get suitable results. It was a 4-minute, 500-lb batch operation, so that the first drop of water did not have to mix evenly, just so long as it mixed during the mix time of 4 minutes, same as you mix up your dough batter when making bread.
The real problems (not known until it was too late) were that the dust was coming in at varying moisture levels, varying temperatures, and varying densitys. In his calculations my boss had ASSUMed that all these things were constants and not a factor in the mixing. I was measuring the water added with a flowmeter, so that was controlled well enough, but the weight of dust was not accurate, because the entire blender platform was mounted on load cells, and the vibration of the platform caused the weight reading to fluctuate wildly. Even when I did manage to filter out the wild weights (by stopping the blender for 10 seconds while the PLC read the weight), the weight did not tell me how much water to add, because the density of the dust varied from minute to minute, so that the correct water/dust ratio was continually varying. I could not go back and install density sensors and thermocouples on the incoming dust, because we spent the entire budget in trying to find out the problems. By that time the customer was fed up and threw us out of the plant. As a result of this project, the manager who hired us was fired for the screw-up.
You have got it all wrong. I did not use ANY PID instructions in this blender program. The PLC progream merely looked at the weight of dust in the blender, and calculated how much water should be added, then opened a valve and measured the water until the flowmeter showed the correct gallons. No, the model did not assume perfect mixing, but only 90% had to be mixed to get suitable results. It was a 4-minute, 500-lb batch operation, so that the first drop of water did not have to mix evenly, just so long as it mixed during the mix time of 4 minutes, same as you mix up your dough batter when making bread.
The real problems (not known until it was too late) were that the dust was coming in at varying moisture levels, varying temperatures, and varying densitys. In his calculations my boss had ASSUMed that all these things were constants and not a factor in the mixing. I was measuring the water added with a flowmeter, so that was controlled well enough, but the weight of dust was not accurate, because the entire blender platform was mounted on load cells, and the vibration of the platform caused the weight reading to fluctuate wildly. Even when I did manage to filter out the wild weights (by stopping the blender for 10 seconds while the PLC read the weight), the weight did not tell me how much water to add, because the density of the dust varied from minute to minute, so that the correct water/dust ratio was continually varying. I could not go back and install density sensors and thermocouples on the incoming dust, because we spent the entire budget in trying to find out the problems. By that time the customer was fed up and threw us out of the plant. As a result of this project, the manager who hired us was fired for the screw-up.
Tom Jenkins , I understood, always good know the limits about the concepts. But as feedforward associates to the PID? (Better guide to an illustration)
Peter Nachtwey It cleared that only associates for SP (cascate) and explained the operation, however did not I understand the example. (equilibre of the energies)
See this application(draw): Heat the pieces
Variations: Size, espace from the piece (disturbances)
With one variable, lives easy understand. Explain new. Feedforwards - ideal aplication and result! Correct?
My comprehension is better is said: event; cause; effect; correcting action; turned out
Peter Nachtwey It cleared that only associates for SP (cascate) and explained the operation, however did not I understand the example. (equilibre of the energies)
See this application(draw): Heat the pieces
Variations: Size, espace from the piece (disturbances)
With one variable, lives easy understand. Explain new. Feedforwards - ideal aplication and result! Correct?
My comprehension is better is said: event; cause; effect; correcting action; turned out
Tom Jenkins
Lifetime Supporting Member
Peter: I've never heard the term "disturbance rejection" for a control technique before, and I can't find it in my references. You definition makes sense, but I would have classified it as a variation of feedforward.
Dandrade: PID and feedforward are not related, and in some ways are the opposite of each other. They are different control techniques, although they may be used in conjunction with each other.
I would interpret your diagram as two loops.
The feedforward loop would sense the size and temperature of the incoming material. Based on a mathematical model using heat capacity, mass of the material, the feedforward loop would calculate how many watt-hours of energy were required to raise the temperature and using the expected time at the heater, calcualte the watt output required.
A second independent PID feedback loop would then control the heater output to provide the watt energy rate to the material.
OK, Peter, how close did I get?
Dandrade: PID and feedforward are not related, and in some ways are the opposite of each other. They are different control techniques, although they may be used in conjunction with each other.
I would interpret your diagram as two loops.
The feedforward loop would sense the size and temperature of the incoming material. Based on a mathematical model using heat capacity, mass of the material, the feedforward loop would calculate how many watt-hours of energy were required to raise the temperature and using the expected time at the heater, calcualte the watt output required.
A second independent PID feedback loop would then control the heater output to provide the watt energy rate to the material.
OK, Peter, how close did I get?
Peter Nachtwey
Member
Distrubance Rejection or Feed Forward, there is a fine difference
Tom, I would approach the problem like you stated. This is more like disturbance rejection because the PID must measure the speed of the conveyor and how much material is on the conveyor to determine how much heat is required to change the material from the inital to final state. A faster moving conveyor and a high amount of material/converyor would increase the demand so why not add this estimated demand to the bias term? This looks like veneer drying application although there must be many similar applications.
Feed forwards don't require an outside measurement. The are based on the SP and its derivatives. Other that that they are the same and often get confused. I know it is a fine point and splitting hairs.
Our motion controller uses feed forwards because the target velocity and acceleration are calculated not measured. However, sometimes when an axis is geared to another axis we use the measure speeds and accelerations of the master to aid the slave axis in following the master instead of relying on just the PID. I still use the same feed forward gains for both. Sometimes the difference is small.
Tom, did you google distrubance rejection? I got about 75000 hits.
I post the link below because it applies to another recent thread about winders and distrubance rejection. How does one go about explaining the link below? I think Keith did a good job of explaining the general idea, but the details are complicated.
Disturbance Rejection for Winders
Tom, I would approach the problem like you stated. This is more like disturbance rejection because the PID must measure the speed of the conveyor and how much material is on the conveyor to determine how much heat is required to change the material from the inital to final state. A faster moving conveyor and a high amount of material/converyor would increase the demand so why not add this estimated demand to the bias term? This looks like veneer drying application although there must be many similar applications.
Feed forwards don't require an outside measurement. The are based on the SP and its derivatives. Other that that they are the same and often get confused. I know it is a fine point and splitting hairs.
Our motion controller uses feed forwards because the target velocity and acceleration are calculated not measured. However, sometimes when an axis is geared to another axis we use the measure speeds and accelerations of the master to aid the slave axis in following the master instead of relying on just the PID. I still use the same feed forward gains for both. Sometimes the difference is small.
Tom, did you google distrubance rejection? I got about 75000 hits.
I post the link below because it applies to another recent thread about winders and distrubance rejection. How does one go about explaining the link below? I think Keith did a good job of explaining the general idea, but the details are complicated.
Disturbance Rejection for Winders
Similar Topics
Hello, I am attempting to tune a PID loop on a process. The process involves a valve with electronic actuator that has quite a high deadband...
so i have 4 25gpm wells feeding a 1000gal tank (T-1), with an additional 15gpm from a decant tank for 3hrs every 12hrs.
P1 and P2 both controlled...
Hi everyone, yet another PID problem. I'm hoping I understand enough of the process I'm controlling that my request for help is reasonable. If a...
A few months ago, I started to look into PID controllers and the tuning of first order processes. This has, partly thanks to you, resulted in a...
I haven't had to tune a PID loop in a very long time.
It's actually a PI loop for a pulse width modulation s.v.
What was the name of that tuning...