Webinar on SOPDT tuning

[agarb]

OkiePc, you should look at the the Scilab autotuning program here
http://www.deltamotion.com/peter/Scilab/AutoTune/
and look at the format of the .txt files. If your data is in that format then the Scilab auto tuning program should work for temperature systems.

Peter - I am definately interested next time if I can work it in my schedule.

Just downloaded scilab + your files to play with.

 

[Peter Nachtwey]

What I am learning

I am finding out that using the Smith Predictor makes a HUGE difference in the ability to control temperature systems. The IMC method on the controlguru site basically reduces the controller gain as the dead time increases. This works but only to a point. I talked to Pandiani and he said they have long dead times there of up to 4 minutes so I am now simulating controlling a SOPDT plant with a 4 minute dead time but using only FOPDT gains with a Smith Predictor. Even though the controller gains aren't theoretically right and the tuning is aggressive the results are good when using the Smith Predictor. Those that do temperature control of plants with long dead times should study and implement the Smith Predictor.

 

[ndzied1]

Peter,

Do you think the Smith Predictor would help in a system to control flow (based on a flow meter) where the flow meter has some lag time for averaging?

Thanks

 

[Peter Nachtwey]

Yes, if the flow device is your feedback for a closed loop

Peter,

Do you think the Smith Predictor would help in a system to control flow (based on a flow meter) where the flow meter has some lag time for averaging?

Thanks

The Smith Predictor isn't too hard to implement from a programming point of view but one must have a pretty good idea of what the dead time is along with the plant gain and time constant.

Implementing a Smith Predictor for flow would be a little tricky because the dead time would vary if the flow changed. This would affect how far back in the past one would need to look for computing the correction.

 

[ndzied1]

I'll do some testing

I can get some pretty accurate readings on dead time by making open loop step changes to the spool position and watching the flow meter response. They all seem to have some averaging routines that slow the response down.

This one looks to be about 750 msec but I have been off on other things for a while and will get back to it next week.

Thanks

 

[Peter Nachtwey]

We can model the sensor feed back response as a time constant

A SOPDT temperature system has two poles. One is the heating of the thermal mass and the other is the response of the sensor. So your flow meter has a filter. If it is a low pass filter we can estimate that too.

 

[keshik]

I know I'm dragging up an old thread, but did round 2 of this ever occur?

 

[Pandiani]

Not yet, but will soon, I hope.

 

[Peter Nachtwey]

There hasn't been a second webinar yet.

This was good timing. I was going to find this thread and make the announcement. I plan do the next webinar Saturday morning a 9am PDT.

I am also going to narrow the scope down so the webinar isn't too long.
This time I am going to show how to do ZN tuning along with other systems that are similar. I am not a fan of ZN tuning but I have found that it works pretty well as long as the dead time isn't too long relative to the other time constants. There were also some fine details I had to work out. It takes a long time to do ZN tuning properly.

I plan to do future webinars about other narrow related topics like:
Using the Bias.
Internal Model Control ( IMC ) ala Control Guru
Sliding Mode Control ( SMC )
Smith Predictor. The Smith Predictor webinar would be an advanced one.
Different forms of PID. PID vs PI-D vs I-PD.

If anyone has ideas then let me know. I don't want to get into PLC specific things. The PLC manufacturers should support their own 'peculiar' features.

I will get using goto meeting. I will post the meeting id on this thread about 15 minutes before starting. Make sure your audio equipment works. I will be sharing my screen

 

[keshik]

Good timing indeed. Is there anyway that you can record the session? I'm not sure if that time will work for me.

Thanks!

 

[TurpoUrpo]

Omg I hate all this "shorts" for times. Why can't everybody just use UTC+-Y.

PDT seems to be UTC-7.

Its nice time to make that webinar, im in as its just 19.00 in here (if i did not mix anything).

 

[Peter Nachtwey]

I am getting ready. Finding Ku is time consuming

 

[Peter Nachtwey]

1. Please join my meeting.
https://www1.gotomeeting.com/join/985475272

2. Use your microphone and speakers (VoIP) - a headset is recommended. Or, call in using your telephone.

Dial +1 (314) 627-1505
Access Code: 985-475-272
Audio PIN: Shown after joining the meeting

Meeting ID: 985-475-272

I can only support 25 people but I doubt that many will attend
I will start at the 9 am PST.
THat is about 20 minutes from now.

GoToMeeting®
Online Meetings Made Easy™