PID controllers (classical controllers) and advanced process controllers can be summarized as follows:
1. The PID controller does not use a process model while advanced controllers usually needs a process model and advanced controllers calculate the control output through the simulation(or prediction) on the basis of the process model. So, it is clear that the PID controller is structurally simpler and the implementation is much easier.
2. Actually, PID controllers are good enough in controlling most processes in industry with acceptable accuracy. Also, PID controllers show good robustness to uncertainties and variations of the process because the control output is a linear equation with respect to the control error and the direction of the output is intuitively right. So, most controllers in industry are in the form of PID controllers due to 1 and 2.
3. The performance of advanced controllers depends on the accuracy of the process model. Usually, a plant test (process excitation) is required to obtain the process model. The plant operator hates this. He would not cooperate in most cases except the case that the PID controller cannot provide acceptable control performances for the process.
4. Advanced process controllers are recommended for unusual cases such as MIMO processes with strong interaction, highly nonlinear processes, batch processes requiring extremely high control performance etc. In these cases, the plant engineer has no choice to consider advanced process controllers even though the cost is expensive and the whole procedure for the implementation is sometimes painful.
5. In the case of mass production, a small improvement of the control performance can make a huge amount of money. In this case, also, advanced process controllers can be a good choice.
Hello;
The new methods in many cases, its expensive more than the system!
I agree with Mellah. The new methods are rarely needed and overly expensive.
PID should be able to control most systems, or simple variations on it. In fact, one uses D very seldom, so most systems are fine with PI control.
What works for me is the use of a micro controller, with analag filtered and buffered inputs and some "tricks" e.g. oversampling and decimation, simple filters of moving averages, ignoring rogue readings etc. Use a fast processor with high ADC sampling rate - typical 12-15 bit with 1 - 2 Msps.
We can say that the PI Control is Hardware Oriented, actually based on Output and Input and is exact as it depends mainly on Real Time Variables and Deterministic Mathematics.
On the other hand the Intelligent methods like Fuzzy Logic ,Genetic Algorithm,Neural Networks etc. are all Iterative and depend on Initial Selections like Framing of Fuzzy Rules,Selection of Population or Number of Layers,Weights given etc and are iterative in Nature and the experience of the Designer matters a lot.
Deterministic Techniques are always Superior compared to Techniques that is based on Search Algorithm or Iterative.
Most of the Real Time Control Systems using PI Controls work satisfactorily.
I fee,l PI control is not outdated or dead.
P.S.
Most of the industrial problems are solvable with PID controllers and the new methods are research practices.
In my knowledge. Non-linear control is especially appropriate in non-linear problems, which have otherwise to be linearised to be used with linear controllers liked PI. This linearisation can provide acceptable results or not according to the approximation errors around the linearisation point, the frequency of linearisation and indeed the target level of precision. A non-linear controller especially designed for the nature of the problem, the application, could bring a better predictability and stability to the solution (i write a paper on this in the application of motion control). New approaches seem in general takes better advantage of the knowledge of the nature of the problem to bring better fitted solutions. Nevertheless, PI can often be used in conjunction to new approaches for example in order to provide a fast-closed control loop at the output of a non-linear controller that further reduces control errors. Linear controllers such PI, in general, have also the advantage to be easy to use and to adjust, and to be handled by commercial, industrial devices. They can thus be more accessible and appropriate for non-experts.
There are many types of controller, but there is no way to say that one controller is better than the other in all cases. To compare two controllers, you will have to identify first your criteria.
If the criteria, for example, is the ability to satisfy many robustness conditions, then PID should work better than PI because PID has three degrees of freedom (parameters that could be tunes), i.e. Kp, Ki, and Kd, while PI only have two, i.e. Kp and Ki.
If the criteria is circuit complexity, then PI becomes better than PID because PI has lesser components as compared to PID.
If the criteria is step response, there are times that PI is better than PID, depending on plant to be controlled and the tuning method that is used. The PI and PID tuning handbook by O'Dwyer has all the formulas and you could try some of them in determining which is better.
In conclusion, we cannot say which is better.
Emm
By and large, PID and its variants along with simpler schemes available in ladder logic are the most commonly applied approaches by far. This is so for a number of reasons, unfortunately, not all good. First and foremost, frequency domain techniques provide what I call a big hammer - they'll work for almost every physically realizable system (it's simple to construct an abstract system for which PID will not work for instance, by Brockett's theorem, no nonsmooth system can be controlled by a smooth feedback but, in reality, very few physically realized systems are truly nonsmooth though that is changing with with more and more cyber componentry being introduced into systems across the board).
Just because something will work, doesn't mean it should be used. Unfortunately, that is not the philosophy typically applied in practice and, in a lot of cases, the PID-based control systems are a mess with loop after loop added in an ad hoc fashion and kludges galore (gain scheduling anyone?). The practical effect is that the design, implementation, and tuning of these systems is often a "black art." In general, the more moving parts one has in the system being controlled, the less apt PID-type control systems are.
Frequency domain methods are SISO. For MIMO applications, the problem must be cast as a noninteracting control problem and often the design of the physical systems is performed in manner that provides this (for at least nominal operating conditions), for example aircraft in the trimmed condition. These systems do not work well away from the nominal system trajectory, but, in many cases, the use of PID is "baked into" the system design for a number of, typically legacy, reasons (industry "best" practices, NOBODY wants to go through V&V with the FAA for a new control methodology).
While many (even most) applications are well-suited for application of PID control, there are a substantial number of important applications that aren't. I agree with Mellah that in some cases, the new approaches are more expensive than the system but not in all, and I think that this is changing as base systems are becoming more complex and are having to meet more stringent performance criteria and low power sensing, communication and processing is becoming ubiquitous. Adherence to PID-based control approaches is, IMHO, "penny wise, pound foolish." There is no intrinsic reason why many modern control techniques are more expensive (some more exotic methods perhaps, but not basic pole placement approaches) - it requires no new hardware in most cases, it requires no additional design effort, simulation, or hardware in the loop testing. The additional cost is associated with the designers learning (or refamiliarizing themselves with) some new techniques.
That being said, I don't think that you'll see a lot of control engineers embracing modern control methods any time soon. I also think that, in part, because of this obeisance to a legacy mindset, the stagnation of the field has a fair chance of consigning control engineers to irrelevance in many areas as more and more of the control functions are absorbed into computer science and engineering.
PID is still the most popular and applicable control scheme due to its simplicity and low cost. However, PID is taking advantage of the intelligent methods in tuning. We find now fuzzy PID, etc...
PID controllers (classical controllers) and advanced process controllers can be summarized as follows:
1. The PID controller does not use a process model while advanced controllers usually needs a process model and advanced controllers calculate the control output through the simulation(or prediction) on the basis of the process model. So, it is clear that the PID controller is structurally simpler and the implementation is much easier.
2. Actually, PID controllers are good enough in controlling most processes in industry with acceptable accuracy. Also, PID controllers show good robustness to uncertainties and variations of the process because the control output is a linear equation with respect to the control error and the direction of the output is intuitively right. So, most controllers in industry are in the form of PID controllers due to 1 and 2.
3. The performance of advanced controllers depends on the accuracy of the process model. Usually, a plant test (process excitation) is required to obtain the process model. The plant operator hates this. He would not cooperate in most cases except the case that the PID controller cannot provide acceptable control performances for the process.
4. Advanced process controllers are recommended for unusual cases such as MIMO processes with strong interaction, highly nonlinear processes, batch processes requiring extremely high control performance etc. In these cases, the plant engineer has no choice to consider advanced process controllers even though the cost is expensive and the whole procedure for the implementation is sometimes painful.
5. In the case of mass production, a small improvement of the control performance can make a huge amount of money. In this case, also, advanced process controllers can be a good choice.
Recent trend is so called Intelligent controls are being compared with PI controller performance and show that damping of the system has improved. Obviously PI controller damping is poor. If good damping is required we have to go for PID. Accordingly these Intelligent controls results are to be compared with PID performance, but not with PI. It is not good to compare oranges and apples.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering