PID Controller: Order, Stability & Labview

[Darren93]

I'm a physics student so don't do much in the way of electrical engineering, pardon my ignorance. However I'm looking at using a PID controller with a resistance thermometer sensor and heating element plant, with a reference point of some resistance on the thermometer. That is heat up a device to a particular temperature. However when considering generally what's going on here, I get a bit confused. That is considering the stability of the system, the PID controller seems to be 2nd order, surely there is no unstable point for a 2nd order transfer function? I know I can simply use the Ziegler–Nichols method to come up with tuning parameters, what I don't get is how there is a point at which the system starts to become unstable?

Separately does anyone know of any digital tuning scripts that would work with labview, well free ones anyway? (Ignore this really, it's a long shot to a separate issue).

 

[Bystander]

how there is a point at which the system starts to become unstable?

You are controlling a real system with thermal conductivities, and diffusivities between the error sensor and the response heating.

 

[jim hardy]

That is considering the stability of the system, the PID controller seems to be 2nd order, surely there is no unstable point for a 2nd order transfer function?

Please excuse if i missed the question here...

Remember stability is for the closed loop
If transfer function of process is G and controller is H
you have to make G/(1+GH) stable.

 

[donpacino]

I'm a physics student so don't do much in the way of electrical engineering, pardon my ignorance. However I'm looking at using a PID controller with a resistance thermometer sensor and heating element plant, with a reference point of some resistance on the thermometer. That is heat up a device to a particular temperature. However when considering generally what's going on here, I get a bit confused. That is considering the stability of the system, the PID controller seems to be 2nd order, surely there is no unstable point for a 2nd order transfer function? I know I can simply use the Ziegler–Nichols method to come up with tuning parameters, what I don't get is how there is a point at which the system starts to become unstable?

Separately does anyone know of any digital tuning scripts that would work with labview, well free ones anyway? (Ignore this really, it's a long shot to a separate issue).

To add to what others have said, do some research on phase and gain margin
http://en.wikipedia.org/wiki/Phase_margin

 

[stevenb]

the PID controller seems to be 2nd order, surely there is no unstable point for a 2nd order transfer function? I know I can simply use the Ziegler–Nichols method to come up with tuning parameters, what I don't get is how there is a point at which the system starts to become unstable?

What is your model for the plant (thermal system) you are trying to control? The closed loop response is a combination of the controller transfer function and the plant transfer function. So, you can't predict stability or optimim PID gains without consideration of your plant. If you don't know your plant model, then you must tune experimentally (e.g. Your Z-N method). Keep in mind that these tuning methods only get you in the ball-park of a good solution, and you should manually tune it from there.

Typically thermal systems require only a PI controller and the D part is usually not needed.

 

[CalcNerd]

Most PID controllers have an auto tune setting. You can configure an initial guess of parameters (best to be educated, but will probably zero in with a poor guess too). Set your control points and the controller will monitor rise time and error and self correct to a point of very small error as the controller oscillates around the controlled parameter.

I suspect you are over indulging in the control literature and theory vs just setting up your PID controller. Unless you are operating near a critical set point ie where you can't afford a moderate overshoot (which will happen to an automated PID controller with a poor guess for the PID settings) can overshoot the control point by adding too much energy (what ever the control medium) to the system and overshooting the control point badly. And that is an excellent reason to understand the math and controls. But many process can allow the auto tune feature to work. After the PID controller kicks in and does its job, you can usually go in and read the auto tune PID settings for a future configuration or help you understand the process. Often your "best guess" may be a poor guess, because you overlooked some aspect ie heat losses greater than expected or temperature of ingredient's not high or low enough, etc.

If you are attempting perfection on the first pass, well ignore all of my advice, because the above will result in one control swing that will result in a large process overshoot compared to a calculated "good" set of PID values. However, its also been my experience that I have always found something in the process that negated all my efforts for a first pass calculation on PB, PI and PID controllers. As Stevenb stated, you probably don't need the derivative feature of your controller, and most PID controllers can be configured as simpler PI and PB controls. However, a properly tuned PID controller is better.

That is why I am a bit more lackadaisical about analyzing the system to obtain what will probably be approx. values for the numbers anyway. (unless you want to prove something to yourself).