Polling of SCADA and DCS

what is SCADA?..What is DCS?...Polling?

SCADA? supervisory control and data acquisition..What is DCS?distributed control system...Polling? Whacking someone in the head with a large stick.

Oh well, two out of three ain't bad.

Maybe things have changed since I was involved with DCS systems, but such blanket statements were usually suspect to me. You have to check the specifications of specific SCADA and DCS systems--you may find that some SCADA systems are more efficient than some DCS systems, or vice versa.

Also, you have to define what you mean by efficiency--is it scanning efficiency, in terms of points scanned per second, or storage efficiency (how many hours of data stored), and if data is stored on an exception basis, what is the minimum change that defines an exception, and is it (and, in fact, are all of these things) adjustable?

Hi rhkrammer

"You have to check the specifications of specific SCADA and DCS systems--you may find that some SCADA systems are more efficient than some DCS systems, or vice versa."

- what will be good points to check? pls mention in short.

- what will be the normal points scanned period for SCADA and DCS?

- what will be the storage efficiency (is it measured in terms of amount of data stored in a give period of time?)

- what is a small example of 'minimum change that defines exception'

(- what is the meaning of data stored on exception basis (is it related with exception handling in java?)

Regards,

sks

General comment: I think you need some qualified help--I'll try to answer some of the questions below, but some of them would require a lot of explanation. It might help to read the wikipedia articles on SCADA and distributed control systems.

My experience was on complicated systems with lots of control loops and data points, including a computer to do what I'll call supervisory control, and requiring a great deal of customization to make it work to fit the needs of the client. (Things like control of a reheat furnace and multi-stand cold rolling mill for a steel mill, or of a continous caster for a steel mill, where things like haeting profile, mill speed and bite, and cooling water flow had to be controlled per various general recipes depending on the type of steel being cast or rolled. My answers here are, no doubt, colored by that experience. Other systems can be much simpler, and "off the shelf" solutions may be adequate.)

You could probably get better advice by telling us more about what your process is.

sauravbabu wrote the points below preceded by ">":

> (- what is the meaning of data stored on exception basis

Both DCS and SCADA systems are digital--all analog inputs are converted to digital to use in the system (including for storage). Also, they don't collect continuous samples of those analog (or digital) inputs, but collect them on some schedule, from which comes the concept of scanning / polling.

Given some signal at some scan rate, let's just say once per second for convenience, the system could store the results of every scan--in other words, one sample every second. This would use a certain amount of storage if it was saved for whatever purpose.

An alternate approach is to compress the data in a variety of ways. One way of compressing the data is to only store a new sample if the data changes. For a (1 bit) digital signal, this is straightforward, the sample is either a 1 or a 0. Whichever it is, in storage do something like storing the value and how long it has been in effect (or the time when it changes)--thus, if the value is 1 now (say 1:00 PM GMT), and at 2:00 it changes to 0, you store exactly that, as opposed to storing 3600 samples of 1 between 1:00 and 2:00 (and, of course, more samples after 2:00.

For analog signals, they may be converted to digital signals with 8, 12, or more bits of precision. Quite often that is more precision than required (or reliably available from the sensor) for your purposes. Considering it in terms of pressure (and I'm going to make up these numbers rather than make them accurate), suppose you are measuring pressure in a pressure cooker.

Your main interest is probably keeping track how long the pressure is above some threshold (maybe 15 psig) (and maybe controlling the pressure to stay slightly above that threshold--i.e., a setpoint). 11 bits of precision is about 1 part in 2000, i.e., with a 10 bit converter (and a 0 to 20 psi range on the input sensor), you could measure the pressure to a resolution of 0.01 psi.

If you want to store a sample of the pressure every second, you need a certain amount of storage. (Now the other thing I won't go into atm is that you probably don't even need a sample every second for control of a pressure cooker, perhaps a sample every 15 seconds is more than adequate.)

Anyway, if you can determine that you only need the pressure to a resolution of 0.1 psi (instead of 0.01 psi), you can reduce the storage requirement significantly. For example, as long as the pressure can be rounded off to tenths of a psi and not change, you treat the pressure as unchanged, and don't store changes for the hundredths of a psi. (E.g., as long as the pressure is between 14.86 psi and 14.95 psi, you treat the value as being 14.9 psi. And, like the digital example above, you don't store a value for this every second, but instead just store the time that the value became 14.9 psi (after rounding), and later, the time when it moved to some other value (like 14.8 or 15.0), thus reducing the storage required.

I've oversimplified or mistated some of this (or maybe just didn't mention all the relevant points). A control engineer should decide what the sampling rate should be for a given control loop to give proper control, considering all relevant aspects. (One minor point: Note that the next value (after 14.9) might not be 14.8 or 15.0 if the process variable (pressure in the cooker) can change quickly enough.)

So, that is an example of data stored on an exception basis. I'd want to have a means to choose how much of a deviation is an exception--in the example above, do I want to ignore hundredths of a psi, tenths of a psi, some multiple of either?

> (is it related with exception handling in java?)

Not really, except to the extent that the word exception is used in both cases.

> - what will be good points to check? pls mention in short.

You've mentioned some with your questions here. But a control engineer would ask (or investigate) many more. Is your system going to do any direct digital control (like the example of the pressure cooker--will it turn on and off burners or the flow of steam to control the pressure in the cooker, or will that be done by some dedicated controls, and the DCS / SCADA system will only supervise that control)--if it will do direct digital control, someone must investigate the scan rate required for each loop, and then, assuming more than one loop will be controlled (possibly for other process functions), does the DCS / SCADA system have enough processing / scanning capacity to handle all the loops at the required rate.

> - what will be the normal points scanned period for SCADA and DCS?

Well, you or your control engineer have to figure out what scan rate is required, then you have to investigate SCADA and DCS systems to see which ones can satisfy that requirement.

On some projects I've worked on, we've controlled things that required very fast scan rates (in the millisecond range)--this was usually when a (special) motor was involved in the process. We've often used dedicated / specialized controllers for things like that, or PLCs where the control was a little less critical, but we needed to maintain sub-second responses for operator controls.

> - what will be the storage efficiency (is it measured in terms of amount of data stored in a give period of time?)

That sounds good, but you need to understand what a vendor may be telling you--one vendor may specify his storage capability (maybe I like that word better than efficiency) based on a certain strategy (or parameters) of data compression, another may use different strategies or parameters. How do you compare them? (It takes some digging or past experience with that vendor's equipment.)

> - what is a small example of 'minimum change that defines exception'

Go back to the pressure cooker example. In that case, I chose tenths of a psi as being that minimum change that defines an exception. I might have wanted to choose 2 tenths, or 5 hundredths of a psi--does the system give me a means to make that choice?

FYI;

polling.

A network access method where a master controller manages the communication

process by interrogating each slave controller under it to determine

whether the slave has any information to send.

i had never thought of SCADA and DCS as mutually exclusive systems. SCADA is used for the day-to-day running of an entire process, with control of on/off, speed, cycle time, etc of the process. Data for these events are updated by the second (or so, depending on the requirements). With DCS, control is to a finer degree, and of a part of the process. So the SCADA could be controlling the DCS, but only polling that information which is of relevance to the overall running of the entire system. The additional data stored by(for?) the DCS would be used for problem solving/avoidance (maintenance) and polled separately.