Hi all. I need to tune a PID controller with time constants (period of the scillation) on the order of hours. Any suggestion for procedures or algorithms? Literature? Rune
Tune PID controller?
Started by ●May 11, 2011
Reply by ●May 11, 20112011-05-11
Rune Allnor wrote:> Hi all. > > I need to tune a PID controller with time constants > (period of the scillation) on the order of hours.Do you have a model of the controlled plant or the PID has to be tuned empirically?> Any suggestion for procedures or algorithms? > Literature?PID is a basic thing, you will get zillion of references on google. * The first that comes to mind are Ziegler and Nichols methods. * Tim Wescott has an article "PID without PhD" on his web site. No pun intended :-) Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●May 11, 20112011-05-11
On Wed, 11 May 2011 09:21:58 -0500, Vladimir Vassilevsky wrote:> Rune Allnor wrote: >> Hi all. >> >> I need to tune a PID controller with time constants (period of the >> scillation) on the order of hours. > > Do you have a model of the controlled plant or the PID has to be tuned > empirically? > >> Any suggestion for procedures or algorithms? Literature? > > PID is a basic thing, you will get zillion of references on google. > > * The first that comes to mind are Ziegler and Nichols methods.The Ziegler-Nichols method requires knowing only the gain at which a proportional only controller begins to cause oscillation and the period of the oscillation. When I last tuned PID controllers ~15 years ago I found that Z-N produced exactly the same tuning as the auto-tuning controllers I used. Original paper (which uses somewhat archaic terminology) at: <http://www.driedger.ca/Z-N/Z-N.html> Regards, Glen
Reply by ●May 11, 20112011-05-11
Glen Walpert wrote:> On Wed, 11 May 2011 09:21:58 -0500, Vladimir Vassilevsky wrote: > > >>Rune Allnor wrote: >> >>>Any suggestion for procedures or algorithms? Literature?>>PID is a basic thing, you will get zillion of references on google.>>* The first that comes to mind are Ziegler and Nichols methods. > > > The Ziegler-Nichols method requires knowing only the gain at which a > proportional only controller begins to cause oscillation and the period > of the oscillation.There are several methods of Ziegler and Nichols, some based on the step response, the others on the Bode plot, self oscillation or something else. There are also numerous improved methods based on the Ziegler-Nichols classics.> When I last tuned PID controllers ~15 years ago I > found that Z-N produced exactly the same tuning as the auto-tuning > controllers I used.For the simple plants, there isn't really much of a difference.> Original paper (which uses somewhat archaic > terminology) at: > > <http://www.driedger.ca/Z-N/Z-N.html>Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●May 11, 20112011-05-11
On 05/11/2011 07:21 AM, Vladimir Vassilevsky wrote:> > > Rune Allnor wrote: >> Hi all. >> >> I need to tune a PID controller with time constants >> (period of the scillation) on the order of hours. > > Do you have a model of the controlled plant or the PID has to be tuned > empirically? > >> Any suggestion for procedures or algorithms? >> Literature? > > PID is a basic thing, you will get zillion of references on google. > > * The first that comes to mind are Ziegler and Nichols methods. > * Tim Wescott has an article "PID without PhD" on his web site. No pun > intended :-)If there's a reason _not_ to use the method that I present in this case, it's because with time constants in the hours, tuning will take _forever_. My favorite tuning method for most systems is to collect a frequency response, then use Bode plots to tune for stability and bandwidth. With settling times in the hours, this is going to be even more absurdly time consuming than the "millwright's method" that I present in "PID Without a PhD". Ziegler and Nichols has been superseded by the Astrom-Hagglund method. A-H is very like Z-N, so Vladimir may have been lumping it in with Z-N when you mentioned it, however. If you have any prior information on the plant behavior -- particularly if you have a model that you're pretty confident with -- getting the step response of the plant, using it to refine your model, then tuning from the model may be the best way to go. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●May 11, 20112011-05-11
On 05/11/2011 08:33 AM, Vladimir Vassilevsky wrote:> > > Glen Walpert wrote: > >> On Wed, 11 May 2011 09:21:58 -0500, Vladimir Vassilevsky wrote: >> >> >>> Rune Allnor wrote: >>> >>>> Any suggestion for procedures or algorithms? Literature? > >>> PID is a basic thing, you will get zillion of references on google. > >>> * The first that comes to mind are Ziegler and Nichols methods. >> >> >> The Ziegler-Nichols method requires knowing only the gain at which a >> proportional only controller begins to cause oscillation and the >> period of the oscillation. > > There are several methods of Ziegler and Nichols, some based on the step > response, the others on the Bode plot, self oscillation or something > else. There are also numerous improved methods based on the > Ziegler-Nichols classics.If I were going to use the Z-N method or one of its descendants for this problem I think I'd use the one that starts with an open-loop step response -- unless I had days to spend characterizing the plant. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●May 11, 20112011-05-11
As some people have commented, there are several tunning rules for PID tunning. Some of them are based on identifying your system supposing a specific model (a first order, first order plus time delay, second order model...). I advice you to take a look to the bibliography. You could take a look to the literature I referenced in one of my works, here : http://www.ifac-papersonline.net/cgi-bin/links/page.cgi?g=Detailed%2F42690.html;d=1;browse=c Hope it helps, -Mario>On 05/11/2011 08:33 AM, Vladimir Vassilevsky wrote: >> >> >> Glen Walpert wrote: >> >>> On Wed, 11 May 2011 09:21:58 -0500, Vladimir Vassilevsky wrote: >>> >>> >>>> Rune Allnor wrote: >>>> >>>>> Any suggestion for procedures or algorithms? Literature? >> >>>> PID is a basic thing, you will get zillion of references on google. >> >>>> * The first that comes to mind are Ziegler and Nichols methods. >>> >>> >>> The Ziegler-Nichols method requires knowing only the gain at which a >>> proportional only controller begins to cause oscillation and the >>> period of the oscillation. >> >> There are several methods of Ziegler and Nichols, some based on thestep>> response, the others on the Bode plot, self oscillation or something >> else. There are also numerous improved methods based on the >> Ziegler-Nichols classics. > >If I were going to use the Z-N method or one of its descendants for this >problem I think I'd use the one that starts with an open-loop step >response -- unless I had days to spend characterizing the plant. > >-- > >Tim Wescott >Wescott Design Services >http://www.wescottdesign.com > >Do you need to implement control loops in software? >"Applied Control Theory for Embedded Systems" was written for you. >See details at http://www.wescottdesign.com/actfes/actfes.html >
Reply by ●May 11, 20112011-05-11
On May 12, 3:05�am, Glen Walpert <g...@null.void> wrote:> On Wed, 11 May 2011 09:21:58 -0500, Vladimir Vassilevsky wrote: > > Rune Allnor wrote: > >> Hi all. > > >> I need to tune a PID controller with time constants (period of the > >> scillation) on the order of hours. > > > Do you have a model of the controlled plant or the PID has to be tuned > > empirically? > > >> Any suggestion for procedures or algorithms? Literature? > > > PID is a basic thing, you will get zillion of references on google. > > > * The first that comes to mind are Ziegler and Nichols methods. > > The Ziegler-Nichols method requires knowing only the gain at which a > proportional only controller begins to cause oscillation and the period > of the oscillation. �When I last tuned PID controllers ~15 years ago I > found that Z-N produced exactly the same tuning as the auto-tuning > controllers I used. �Original paper (which uses somewhat archaic > terminology) at: > > <http://www.driedger.ca/Z-N/Z-N.html> > > Regards, > GlenWhat if you cannot form an oscillation due to safety concerns?
Reply by ●May 11, 20112011-05-11
You dont need to experiment an oscillation. There are tuning rules based only by the step response of the system. -Mario>What if you cannot form an oscillation due to safety concerns? >
Reply by ●May 11, 20112011-05-11
On 05/11/2011 12:08 PM, HardySpicer wrote:> On May 12, 3:05 am, Glen Walpert<g...@null.void> wrote: >> On Wed, 11 May 2011 09:21:58 -0500, Vladimir Vassilevsky wrote: >>> Rune Allnor wrote: >>>> Hi all. >> >>>> I need to tune a PID controller with time constants (period of the >>>> scillation) on the order of hours. >> >>> Do you have a model of the controlled plant or the PID has to be tuned >>> empirically? >> >>>> Any suggestion for procedures or algorithms? Literature? >> >>> PID is a basic thing, you will get zillion of references on google. >> >>> * The first that comes to mind are Ziegler and Nichols methods. >> >> The Ziegler-Nichols method requires knowing only the gain at which a >> proportional only controller begins to cause oscillation and the period >> of the oscillation. When I last tuned PID controllers ~15 years ago I >> found that Z-N produced exactly the same tuning as the auto-tuning >> controllers I used. Original paper (which uses somewhat archaic >> terminology) at: >> >> <http://www.driedger.ca/Z-N/Z-N.html> >> >> Regards, >> Glen > > What if you cannot form an oscillation due to safety concerns?There's a number of different ways to extract the relevant parameters -- establishing an oscillation is just one. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
