PID Autotune algorithm

My first job had something to do with PLCs but then I switched my trade and went on to become a "Design engineer". Recently, I've designed a PID type temperature controller. Now I want to add auto-tuning feature to it. The problem is I cannot find any relevant documents to come up with an auto-tuning algorithm. I remember from my days in this forum that there happen to be quite a few experts on PID here. So please help me by giving me some good website links or refrence of books on the subject.

Regards.

Search for the Ziegler Nichols tuning method. It's a good start.

I have found nothing good on the web or in a book about auto-tuning.

I have three control books. None mention auto-tuning. There are plenty of articles on the web. None are detailed enough to implement auto-tuning unless you can 'read between the lines'. That is the trick.

The reason auto-tuning is not mentioned in the books is because auto-tuning is really a series of steps or procedures.

1. System identification. This is the hard part. There are many simple methods that don't work well. The methods that do work well may require a lot of math. Check out least squares system identification. If you want real auto-tuning check out recursive least squares system identification. Even a small 8 bit microntroller should be able to do this on a temperature system. You must be able to compute the gain, poles and zeros of a system. When you think you have it right your model or transfer function should be able to mimic the actual system very accurately given the same control signal.

2. You need to pick a control algorithm that can control the model in the desired way. This means you must be able to chose between PD, PI, PID, I-PD, or state feedback. Sometimes one needs more gains than provided by a PID. In your case you should look at I-PD or state feedback.

3. You need to pick a desired response. In motion control, overshoot is not desired so the resonse should be underdamped or critically damped. In temperature control a slight overshoot may be tolerable. Many systems don't let you choose the response the customer may not get a result that works on his system. For instance, in motion control a critically damped response should normally be the best result but I have found that there are things that can't be modeled properly so one must back off the theoretically best result for the more conservative overdamped result.

4. Finally you need to place the poles to provide the desired response. This is the easy part as it is covered well in just about any control book.

This will take you lots of time and effort if you want a truely good auto-tuning algorithm. My attitude is that others have figured this out. I can to and do it better. Ziegler Nichols is not acceptable good. The competition is doing much better.

Good luck.

I do not have much talk.Basic the all, includes to evaluate the tax of error for the current status, modifying the parameters individually or in group.

Peter, could complement! Which the observations comments, criteria to guide in the decision of the resources (or features) of a code of autotune?

Why Auto-tune ? don't try something auto that you can't do in manual .
Autotune for what ? temperature , flow , pressure , big soak , little soak ?
as a PLC engineer firstly , these questions would be in the front of my thinking before redesigning somthing that is already proven .

Again , something worries me - you were a PLC engineer , became a design engineer , and have now re-written a PID algorithm , didn't you find these technical references when you were doing the above .
You will answer you own question when you can tune your own PID loop manually from first principles - You are clearly a clever guy , having designed your own loop , the autotune should already be part of your thinking .