PID , PIDD , ART

PID is the "proportional integral derivative" that we all know and love.

For some slow processes (like temperature), it helps a great deal to use the acceleration as well as the speed, PIDD would certainly fit with "proportional integral double-derivative", but I'm just guessing. A quick web search yields damn little results, so I don't think it's a terrifically common term.

"Active reasoning technology" sounds like a way to automatically generate excuses. I would assume it's some controller manufacturer's name for a heuristic-based controller that uses an expert system of some type -- unless someone tells me I'm wrong, it may help to give some brand names that it's attached to.

I don't know of any terminology for the PIDD either. Perhaps PIDD'. In anycase the PID is good for tuning plants with two poles and PIDD is good for tuning plants with 3 poles as in the repeated poles thread.

Hydraulic systems can be simply modeled as a mass between two springs which is a pair of complex poles. There is one more pole for integrating the velocity into a position which makes three poles. I am assuming the valves is fast and can be ignored. A PIDD is very good for controlling hydraulic systems.

It all depends on the number of poles in the plant, not the response of the process unless one of the poles is so fast relative to the others and the controller that it can be ignored.

Peter Nachtwey

Hi,

Thanks for these first answers.

1 point to clarify, PIDD is also called PID square or PID2. I well understand that PID use the speed (D) and PIDD use the acceleration(DD). But I don't see exatly the which one is better and the difference for temperature regulation process.

Ebl

First, I have never seen or heard of a need for a second derivative feedback on a temperature control system. It would only be needed if one could identify three poles in the plant. Since most temperature systems are identified as having one or two poles one can get by with a PI or PID respectively. How many gains a controller needs is dependent on the number of poles you are trying to place. Remember, the integrator comes with its own pole so that can be counted. See the repeated poles thread below. One can see that tuning a system with three real poles can be done but there is only one solution that will provide a critically damped response. Adding the second derivative allows making the response faster. See the .pdf in the repeated poles thread below.

If anybody knows of a plant with three real poles I would like to hear of it. Meanwhile if they do exist then most of them can't be tuned easily with just the PID in a PLC. One could use a statistical method that just reduces the tracking error. I don't get involved with temperature control systems too often to say how this is done in practives. It there anybody out there that has?

Peter Nachtwey

Thanks for these answers. More information.

This is a copy of gammaflux provider. Gammalux build hot runner controllers for injection molding machine :

Sense - 20 times per second, Gammaflux controllers precisely measure the thermocouple;

Control - the proprietary self-optimizing Gammaflux PID2 control algorithm adjusts if the actual temperature deviates 0.1=BA F (0.05=BA C) from setpoint. The second derivative (PID2) monitors the actual temperature rate of change. As a result, the TTC module regulates the output to the heater in advance of achieving setpoint to limit or eliminate over or undershoot.

They speak well about PID2 and the second derivative.

Some systems use two thermocouples. They can tell by the temperature difference what the rate of change is much more quickly. Without the second thermocouple one must differentiate the the single temperature twice. The best resolution provide a minimum detectable second derivative of

0=2E1/(0.05)^ =3D 40.0 degree/min^2. Hum.

Since the injection machine cycles can be very fast I would think they would have a way of measuring the temperature disturbance in a cycle and have faster updates than every 3 seconds.

Actually, it first derivative is the rate of change.

There is nothing special about this. This should happen whenever there is a derivative term and no over shoot is desired.

I am sure it works well if their system identification can really find

3 poles. Implementing the second derivative is easy. Identifying a 3 pole system to to make good use of it is not.

I don't normally get involved with the temperature controls in an injection molding machine, just the hydraulic injection part. It would seem to me the best temperature controls would know the state of the machine and adjust a feed forward or bias according to previous history. I wouldn't be surprised if this is done already.

Peter Nachtwey

tp://

Sorry for my lack, but what is a pole?

Here is a mathematical explanation of what a poles is. Engineers use transfer functions to describe how a system will respond. These transfer functions have a polynomials in the numerator and denominator. These polynomials are function of s ( Laplace Transform ) or z ( z transform ). Values of z or s that make the transfer function go to infinity are poles and values of z or s that make the transfer function that go to 0 are zeros.

For the most part I find the above definition useless for providing an intuitive feel for what a pole manifests itself in reality. I prefer to think of poles as representations how energy flows to and from devices in a system. For example a mass and spring example has two energy storage devices. One is the mass which stores kinetic energy and the spring store potential energy. A DC motor stores the rotational energy and inductance of the armature store some energy. Thermal systems must heat material so there is a pole that reflect how the material absorbs and releases heat.

I see control systems as controlling energy. One must add or remove energy to control something. Some of this added energy does not go to the desired form or is lost and the controller must compensate for that.. For instance, when adding energy to a mass and spring system, one want to control the velocity of the mass ( kinetic energy ) but some of the energy goes into compressing the spring ( potential energy ). Therefor the mass will tend to fall behind the intended position because the spring absorbs some of added energy. The controller needs to compensate for the diversions of energy and energy losses ( friction ) to the mass will follow the desired path.

Does anybody have a good gut feel description of a zero?

Peter Nachtwey

If your heater has to warm up a column of material before your sensor starts getting warm, or if you are heating material that flows through a pipe before it gets to your sensor, there will be an infinite number of poles, and sensing the second derivative of the temperature would help if the measurement doesn't get subsumed by the noise.

As you point out elsewhere, however, one could do better by putting in two (or perhaps more) sensors, with one close to the point you want to control and one close to the heater. Jerry Avins has talked about adjusting the dynamics of a furnace controller simply by moving the temperature sensor closer to the heat source instead of increasing the derivative gain on a controller and dealing with noise problems.

I suspect that this particular company has arbitrarily decided to only use one sensor, and that their engineering folks have done their best within that constraint, and that their marketing folks are now trying to dress up a pig in a frilly dress.

But I'm getting cynical in my old age.

le texte des messages pr=E9c=E9dents -

It's one of main provider of hot runner controller. I just want to know what could be really PIDD in temperature process control. It's the main point of this provider and I would like to know if it's true. I know , like all provider of HRC, that they use 1 thermocouple by controled zone. I haven't any special information about PIDD but I know what it's PID. But I'm surprise that nobody know PIDD and that I don't find any information on the web. So if PIDD isn't a result of true calculation or reflexion , I'm surprise that a big compagny like gammaflux use it for sales.

texte des messages prcdents -

If you _had_ to do your control with just one sensor then using a PIDD controller could well be better than using just a PID, as long as you took care of noise issues. Sensing the acceleration of the temperature increase could help you control things quickly without overshoot. I'd still question why they couldn't use another temperature sensor -- there may be a very good reason, or it could just be inertia.

Never assume that just because a company is big that it does things right. Big companies survive by doing things the way they always have been done, whether that way is correct or not. They run on inertia until they fail, then smaller companies come in and start competing successfully with them. At that point either the small company gets big and starts running on inertia itself, or it gets bought by a big company and has inertia forced on it.