Peter Nachtwey
Member
Good questions.
I start with a couple of openloop moves. To a process person that would translate in to controlling the system in manual. I do this just to see how the system responds. I can get a rough idea of what the system gain is. For a motion system I would see what speed the system moves. If it goes 3 inches per second then the system gain is 3 inches per second per volt. The first two spreadsheets with the sinusoidal wave forms where motion system where the system gain was 2 inches per second per volt.
When giving the openloop/manual command, I also look at how quickly the system responds. Look at the real poles spread sheet from last week. Look at the curves carefully. You can easily tell a single poles system from a multiple pole system. Systems with imaginary poles will oscillate or ring. These are all clues that I use.
Next I increase the proportional gain until I get control of the system. This means that I want to be able to keep the actual or PV from drifting around or oscillating. However, there is a point where increasing the proportional gain does no good and increasing the gain just makes the system oscillate. I like critically damped systems.
Next I adjust the feed forwardS. Yes, there can be two or more feed forwards! We have talked about feed forwards yet, but FEED FORWARD ARE THE MOST IMPORTANT GAINS! Ideally 100% of the control effort should come from the FEED FORWARDS. In reality the feedforwards should be able to do 95% of the work and the PID is only required to provide the other 5%.
After I get the feed forwards adjusted the system should look pretty good. I then set the feedforwards temporarily to zero and then tune the PID for a critically damped response so the overshoot in minimized. Normally, people tune their systems so that it is slightly underdamped to get a faster response but I don't because the the feed forwad should do most of the work and it isn't necessary to tune the PID that hot. Tuning the system to be critically damped is usually just a matter getting the ratio between the integerator gain to the proportional gain right. The higher the ratio of the integrator gain to the proportional gain, the quicker the response and the more likely the system is to overshoot. The differentiator may or may not be required. Usually it is but there are two cases I can think of where it isn't. One is the simple velocity system of last week. The other is a system with imaginary poles ( one that rings in response to a open loop step change in the control output ).
Yes, that is the point. I assume you increased time constant A's value from .1 second to .3 second as I suggested. This slows down the response of the system. See the Real Poles Spread Sheet I posted last week. See how a one, two, three and four pole system respond to a step input. Increasing the time onstant is similar to not having enough force to move the mass. In a hydraulic system this occurs because the system pressure is too low or the cylinder is too small in diameter.
Actually, a cylinder is more like the spring part of a mass on a spring.
Yes. You are on the infield playing the game.
When you start to tune is there any particular place to start?
I start with a couple of openloop moves. To a process person that would translate in to controlling the system in manual. I do this just to see how the system responds. I can get a rough idea of what the system gain is. For a motion system I would see what speed the system moves. If it goes 3 inches per second then the system gain is 3 inches per second per volt. The first two spreadsheets with the sinusoidal wave forms where motion system where the system gain was 2 inches per second per volt.
When giving the openloop/manual command, I also look at how quickly the system responds. Look at the real poles spread sheet from last week. Look at the curves carefully. You can easily tell a single poles system from a multiple pole system. Systems with imaginary poles will oscillate or ring. These are all clues that I use.
Next I increase the proportional gain until I get control of the system. This means that I want to be able to keep the actual or PV from drifting around or oscillating. However, there is a point where increasing the proportional gain does no good and increasing the gain just makes the system oscillate. I like critically damped systems.
Next I adjust the feed forwardS. Yes, there can be two or more feed forwards! We have talked about feed forwards yet, but FEED FORWARD ARE THE MOST IMPORTANT GAINS! Ideally 100% of the control effort should come from the FEED FORWARDS. In reality the feedforwards should be able to do 95% of the work and the PID is only required to provide the other 5%.
After I get the feed forwards adjusted the system should look pretty good. I then set the feedforwards temporarily to zero and then tune the PID for a critically damped response so the overshoot in minimized. Normally, people tune their systems so that it is slightly underdamped to get a faster response but I don't because the the feed forwad should do most of the work and it isn't necessary to tune the PID that hot. Tuning the system to be critically damped is usually just a matter getting the ratio between the integerator gain to the proportional gain right. The higher the ratio of the integrator gain to the proportional gain, the quicker the response and the more likely the system is to overshoot. The differentiator may or may not be required. Usually it is but there are two cases I can think of where it isn't. One is the simple velocity system of last week. The other is a system with imaginary poles ( one that rings in response to a open loop step change in the control output ).
Yes, or the valve is too small.When you reduce the system gain from 1 to .5 What does that
mean? If it were a hydrolic system would this mean the pump
was to small?
And when you go to .3 on the time the system
seems to slow.
Yes, that is the point. I assume you increased time constant A's value from .1 second to .3 second as I suggested. This slows down the response of the system. See the Real Poles Spread Sheet I posted last week. See how a one, two, three and four pole system respond to a step input. Increasing the time onstant is similar to not having enough force to move the mass. In a hydraulic system this occurs because the system pressure is too low or the cylinder is too small in diameter.
Actually, a cylinder is more like the spring part of a mass on a spring.
Is this any where in the ball park??
Yes. You are on the infield playing the game.