what would be the best scan time for the refinery closed loops in DCS. As of now the chosen scan time for Pressure & flow is 500ms and Temp. and Level 1sec. Since considering the criticality of refinery, is advisable to go with 250ms & 500ms vice versa?

 

The scan time of a PID loop has absolutely nothing to do with the "criticality" of your process. The required PID scan time is related ONLY to the response time of the process. Typically, for a refinery - and many manufacturing processes using Flow, Level, Temperature and even most Pressure controls - 500ms or 1sec will be just fine.

Rob
www[.]lymac.co.nz

 

Rob wrote:
> The scan time of a PID loop has absolutely nothing to do with the "criticality" of your process. The required PID scan time is related ONLY
> to the response time of the process. Typically, for a refinery - and many manufacturing processes using Flow, Level, Temperature and even most
> Pressure controls - 500ms or 1sec will be just fine.

It is not true. Scan interval does matter. It defines uncertainty of the derivative part. Common practice is to choose the scan interval close to 1/10 the cutoff frequency of the object under control.

Best,
Vladimir E. Zyubin

 

Yes Vladimir. But the question as asked was about whether you should scan faster because you're in a dangerous oil refinery.

 

> Yes Vladimir. But the question as asked was about whether you should scan faster
> because you're in a dangerous oil refinery.

I think if the question is about safety then the solution can be solved by two processes -- one for the PID loop to control temperature (and 1 sec scan time looks a bit strange because it depends on the cutoff frequency) and the second process for safety control (to catch the critical events, where the scan time for the process have to obey the principle the short the better... 0,1 or 0,01 sec).

though I do not aware of programming traditions for the refinery domain... so the above is just a remark.

Best regards, Vladimir

 

sir,

My question is dealt with the execution period, i,e., the update time for the field valve to the controller.

 

Sorry for the confusing response you've received so far....

Vladimir & I agree that in general "faster is better" and in an ideal world you should scan any PID loop at 5 to 10 times faster than the response time. But that isn't always practical.

PID algorithm's take up a lot of CPU power, so a compromise usually gets made. In most cases you are not going to use Derivative and the response time of a flow loop will be around 3 to 6 seconds - so a 1 second scan time - while not ideal in theory - is "good enough" in practice for most situations, and 500ms is plenty. Yes, I know this is a thumbsuck guess about your plant.

Depending on the DCS you are using you may be limited by the "base scan rate" ie. the fastest that the CPU can scan loops. Typically with older style systems you do not have the flexibility to choose a range of scan times for different loops and must instead scan at a multiple of the base rate. Shortening the base scan rate can dramatically increase the CPU load which can only then be resolved by buying more expensive controller hardware. (Though why some customers will spend millions on plant but refuse to spend 50K on an extra CPU is beyond me). I'm guessing here that this is why you're asking about using scan times of only 500ms or 1 second.

Ideally you will be using a control system that has much more flexibility and will let you put some faster PID loops in a faster task if you find that you really need fast control on something. Or you will have an unlimited budget for your control system.

Rob