Process Characterization (Model) for Control

P in the above equation of  control is usually represented by a set of parameters such as (Gain G, DeadTime Td, Response Time, (T1, T2..)). Collectively, all these parameters determine how process P responds to a unit change in  manipulated variable OP.  Depending on P, tuning of PID controller C be made to ensure E is minimized. Variations in one or more of the process parameters, re-tuning of C is warranted to keep E to its minimum value. Failure to do this would cause E to fluctuate widely. One of the most egregious problem is the difficulty of "knowing" values of these process parameters accurately and timely so as to be able to tune PID accordingly. 

​

At PIDAnalytix, we have devised the method of AI-PID tuning that reliably and robustly tune and re-tune PID no matter how little is "known" about P and when changes happen. One of the most affecting process parameter is gain G. Its variation affect E considerably. Basically, G defines change in P when OP is manipulated in open loop. However, when G changes later causing the loop to oscillate it becomes extremely difficult to know its value correctly. It is not always possible or permissible to re-identify value of G in closed loop operation. Consequently, majority of PID loops oscillate to a varying degree and remain so until it becomes unbearable. 

​

AI-PID tuning as demonstrated later show how re-tuning can be done in real time in closed loop. PIDAnalytix have devised adaptive intelligent (AI) method of re-tuning PID loops as and when needed. The art of perfecting PID tuning is what makes AI-PID so unqiue and powerful.

​

Valve Stiction is a "hidden" process parameter which is even more insidious than process gain G that greatly increases E.

​

To demonstrate, the capability and agility of AI-PID, we have included two sets of demonstration. In first demonstration, process Gain is varied randomly 256 times. A comparison is made of typical PID tuning with AI-PID tuning.

 

In second demonstration, deadband arising from valve stiction is randomly varied 256 times. Again, a comparison is made of typical PID tuning with AI-PID tuning. In both these two cases, AI-PID performance remains stoically same, the variations of process parameters are fully and perfectly compensated.

​

AI-PID has been validated to perform robustly for other process parameters (Td, T1, T2) including P being open loop unstable. AI-PID is designed to be highly adaptive for random changes in one or more of the process parameters in real time. AI-PID tuning is indeed limitless in its capability to keep E to its minimum value.

 

There is no known demonstration in last 100 years of  PID tuning of such wide variations and perfect control.

​

Welcome to witness a new dawn of AI-PID tuning to marvel at what can be done with mighty PID. Incidentally, simplicity of PID as a controller is what underlines adaptability of AI-PID. AI-PID demonstrates enduring power of PID arising from its simplicity.