PID Controller for Delayed Process Control...

[sameenkhan]

Hi Guys,

I have experienced a problem on one site during tuning of PID controller which was installed on clarifier in utility plant... The system response was delayed by an hour...meaning change in controlled output was causing a change in output after delay of an hour...

First of all... Being a Control engineer.. how would I know that a process with delayed response is linear or not... 2nd ... what procedure should be adapted to tune the controller...
3rd how to tune controller in case of multiphase fluids??

Best Regards,

Sameen

[tosif]

Sameen, as a standard practice a process design engineer provides control limitations of that particular process to process control engineer. This is the responsibility of process control engineer to design such a logic, so that process remains in provided limits. Tunning of control scheme is one of his job descriptions. Softwares are available to analyse controller response including time lag along PID tuning parameters. Mostly used softwares are HYSYS Dynamics and DYNSIM. The response of such a controller is also dependent on the control system (TDC-3000, Bailey or others) due to difference in scan frequencies and PID configurations in those systems.
Now lets peep into delayed response. First of all be informed that there is no link between delayed process response and behavior of the control output. As far I know the process of clarifier, normally flocculant dosing is the output response of the disturbance in suspended solids concentration. This can either be installed in feed forward or feed backward configuration. Can you elaborate further more the process description, so that I can answer you specifically.

[sameenkhan]

Hi Tosif,

I agree with most of the explanation. Except that there is no link between delayed process response and behavior of the control output.

Controlled output influence the process and that influence can be detected by change in PV instantly or after a delay, all is subject to process dynamics..

For large dead time (1 hour or variable time > 45 min )... what techniques should be used to tune a feedback control loop. Using HYSYS & DYNASIM we can get controller tuning parameters, but how would you determine tuning parameter without these softwares??...

Furthermore, I am talking about how to tune the loop physically in the DCS... I am sure you won't wait for process to react in an hour... Then how would you tune such control loops quickly??

Process Engineers used to tune such loops without any softwares but how???

Regards,
Sameen

[tosif]

sameen, lets start with "how to tune any PID controller offline". process engineer has to study the dynamics of process and gather info to further analyze the control, being 1st or higher order systems. then these dynamic systems will be solved by process engineer, either using differential equations or any simulator. there process engineer ends up with values of tuning parameter such as time constant, gain, damp and lag. that reflects proportional, integral and derivative parameters of PID controller.
when we talk about "online tuning of PID", the major limitation comes with it is real time tuning with real time results. generally, process engineer comes up with good starting parameter, due to the knowledge of process and particular experience. but they also still have to wait for the actual response, also incase of delayed responses.
no doubt that most of the natural and normal processes behave as 1st order dynamics, but when two or more 1st order systems overlap, specifically in series, they moved to 2nd or higher order system. the delayed response hence reflects in 2nd and higher order systems also. for example as in your case, if we assume that we have an RO (reverse osmosis) installed at downstream of that coagulation system. a feedback system is installed to control the coagulation dosage and pressure delta across RO membranes. here both systems are linear 1st order. but being in series they will become 2nd order systems, where delayed response still exists.

hope you get my point.