Process Control vs. Discrete Manufacturing

[Jieve]

I've recently gotten involved in some small process control projects, and having had most of my controls background be involved in discrete PLC control, I'm hoping someone here can shed some light on a few things. Two major differences I see are the way the program cycle works (tasks every 50, 250, 1000ms) for example, as opposed to the continuous cycle in a typical PLC, and the additional diagrammatic programming languages as opposed to the standard IEC languages. So my first "question" is, what is the reasoning behind the task based structure and how do they work exactly? I'm assuming if I select a 250ms cycle for my main program, that it reads inputs, runs the code, writes outputs, then waits until the rest of the 250ms have elapsed and repeats the process. Is this correct? Is this determinism especially advantageous for process control tasks in a way that a traditional PLC program cycle is not? Second, what is the significance of the diagrammatic languages? I know Siemens has PCS7, AB PlantPax, ABB 800xA, etc. and they all have preconfigured blocks in graphical languages that can be used. Is it most common to use these languages in process control as opposed to the IEC languages like Ladder or FBD? Is there really a benefit to one over another? Thanks for any input!

 

[Saffa]

I'm in the water industry and the majority of our controls are just implemented in continuous task.

I personally use mostly FBD and ladder, with a small amount of structured text when appropriate (mostly for comms mapping).

I can't really see any need for timed execution like you detail but maybe it's specific to your industry. Most stuff in water is sloooow. I have RTUs out there that only have a fixed 500ms scan time... for everything. Perfectly fine for controlling a few pumps but pretty rubbish for anything else.

 

[mk42]

Not in the process industry now, but I did a brief stint there a decade ago. The reason I see for using the timed tasks vs the continuous program cycle is that process control is twofold.

1) The timed tasks are relatively deterministic, which can be useful for math in things like PIDs. Not only can it affect the analog math, it makes the control smoother.

2) Different parts of the chemical process have different action times. If you're monitoring a slow temperature, you really only need it to run at a slow rate, whereas some things (maybe a pump?) need faster control. COULD you do all that in a continuous process? Sure, it's done all the time. It's just less efficient.

Motion control has the same type of requirement where it wants a dterministic scan, but at the other end of the scale; it typically needs to be very fast. You might need some axes at a 4ms update, and then a few other critical axes at a 1ms update, and the basic machine control in a continuous cycle that runs as it can. Could you just run everything at 1ms? Sure, but you're wasting CPU to do it.

It's possible that the diagram languages you speak of come from looking similar to the documentation they are generated from. Ladder logic was designed to look like an electrical print. FBD was designed to look like a https://en.wikipedia.org/wiki/Functional_block_diagram, which is a great way to chart a process defined by mathematical relationships, instead of mostly electrical ones.

I think CFC in PCS7 is nothing more than a souped up version of FBD. The "normal" FBD in Siemens is limited in that you can only connect boolean pins together, presumably so that it maintains interoperability with LAD. The "full" FBD is called CFC, and it allows any pin to be interconnected, whether bool or int or real.

 

[TurpoUrpo]

In process control world the way to program is different and the way to control is usually also different.

You can easily often run even 1s tasks in process world without any problems anywhere. Then run faster if really needed.

If you think about typical cycle of process control, say some stuff is transferred from container to reactor via pipes with pump.

The amount of stuff is often high, you only "open the route" with valves, start control of the pump or control valve and run the stuff to the reactor and then close.

In many ways process control is much simpler than machine automation. Usually you get pretty decent stuff already with simple sequence and pid's.

CFC type programming is very well suited to the process world.