Peter Nachtwey
Member
A couple weeks ago Jack was having problems ramping up pressure using a valve. He was having problems with the feedback not matching the SP. I pointed out two things that could be done to make the pressure follow the SP a little closer WITH OUT using a PID. I updated the files pointed to by the link. They now the files show what can be done to compensate for the lag in the response of the valve by using a openloop compensator. The openloop compensator modifies the SP ramp and provides a little lead which compensates for the valve's lag. It is kind of like holding the carrot out in front of the donkey to get it to follow the path you which to travel.
So why bother?
Some people that frequent this site have pointed out that PIDs are over used. I agree, but not that the PIDs are used too often. It is that the PID is used to carry the whole burden of control when PIDs should really be something that just corrects for disturbances and unmodeled features in the system. The bias or feed forward should carry 90% or more of the burden of control. People use/abuse PIDs as if it can solve all the control problems and it can't by itself. Evenn now there are threads where people want to use PIDs but they don't see the big picture and they will wonder why the PID doesn't work.
Jack's valve problem seemed to be a good example to proved a point which is a lot can be done without the PID.
See Jack SLC 500 ramp problem
Notice that I ramped the control signal more smoothly than Jack. I don't see why he chose to update the control signal every 200ms instead of every 10ms. Ramping the control signal smoothly is required for a lead compensator or the output will spike and saturate.
Parts of the files show equations in the s and z domain. I don't expect anyone to follow all the details. However, the equations are very simple and it is really the result that matters.
Pressure Response of Jack's Valve.
I expect only 5 or so people to understand this. I just what to open some eyes.
So why bother?
Some people that frequent this site have pointed out that PIDs are over used. I agree, but not that the PIDs are used too often. It is that the PID is used to carry the whole burden of control when PIDs should really be something that just corrects for disturbances and unmodeled features in the system. The bias or feed forward should carry 90% or more of the burden of control. People use/abuse PIDs as if it can solve all the control problems and it can't by itself. Evenn now there are threads where people want to use PIDs but they don't see the big picture and they will wonder why the PID doesn't work.
Jack's valve problem seemed to be a good example to proved a point which is a lot can be done without the PID.
See Jack SLC 500 ramp problem
Notice that I ramped the control signal more smoothly than Jack. I don't see why he chose to update the control signal every 200ms instead of every 10ms. Ramping the control signal smoothly is required for a lead compensator or the output will spike and saturate.
Parts of the files show equations in the s and z domain. I don't expect anyone to follow all the details. However, the equations are very simple and it is really the result that matters.
Pressure Response of Jack's Valve.
I expect only 5 or so people to understand this. I just what to open some eyes.