Distributed or concentrated?

Hi

When comming to design a factory control is it wise
to concentrate the whole control & data storage in few ruggged
& powerfull computers instead of 30-40 controllers from several different manufacturers, for easy maintenance.

Good to see the new career is working out for you, Pete.

I personally like to see the entire factory controlled by one or two powerful computers with software custom written by one person, in open-source platforms. That way you're just one OS patch or a car wreck away from shutting down the entire facility for hours, weeks, or months.

If you use a DCS then you can get away with one or two controllers as theyre alot more reliable & redundancy is built in everywhere.
So is the increased cost.

If its PLCs and windows-based SCADA I would split it up into sensible chunks.



"Good to see the new career is working out for you, Pete."

Agreed, hope everything is going well. A bit of a change I know!

I prefer to keep all seperate machines controlled seperately. Some related machines may share controllers, but otherwise, to each there own.

This allows for additional machines to be added to the shop easily without interrupting your existing machines. Also, as Ken said, if your own controller failed, your entire shop goes down. Where as if you had seperate controllers, and one fails, you still have the rest of your production.

Plus, who wants all that communications running around the entire shop. It just clutters up with the existing lighting, air, hydraulic, phone, network, PA system, etc. There's already too much stuff running through the rafters.

Can I have your old job?

Depends how you'd measure up.

Not everyones cut out for that sorta thing.

There's always a catch.

I just bought a 3/4 ton diesel truck, whenever someone asks why I need such a big truck I tell them I have a small ....

Peter North said:

Hi

When comming to design a factory control is it wise
to concentrate the whole control & data storage in few ruggged
& powerfull computers instead of 30-40 controllers from several different manufacturers, for easy maintenance.

Click to expand...

I do not know about several manufacturers but if they are stand alone machines let them have their own controller, even if there are 30-40 they could still be controlled (if necessary) by a few PC's running SCADA....Personally, do not want all my eggs in one basket.

What I would consider, IF there are multiple stand alone machines, is have them work stand alone with an HMI for specific control. It is alot easier for the guy on the floor to do much of the control, with SCADA PC's available to monitor, provide recipes, and log data.

THERE IS NO SUCH THING AS EASY MAINTENANCE, if there was you would never need BUBBA's.

Its wise to do what makes sense.

I normally prefer to have one controller per machine.
Our machines are quite large and complicated, so we still need some quite large PLCs, even though it is one per machine.

Our mid 70s vintage machines have 10 PLCs each, this is just because the PLCs used did not have the individual power to control the entire machine by themselves, although the later S5s were up to the job.

I was being a little flippant, because of your username and the leading nature of your post.

In my experience, the most efficient and easiest to maintain facilities are the ones that standardize on one or two control system suppliers and choose a level of decentralization that is appropriate to their machine systems. Although a history of different control systems as the market changes and decision-makers change is common, intentional diversity is not.

If you have independent systems that can be shut down and maintained while others run, or systems that must run regardless of faults in other systems, you should have independent controls for those systems.

I know of no successful complex facilities that centralize their machine controls in just a few PCs.

Ok, now I will be serious.

My vote is for one controller per machine where possible.

With a name of Peter North I doubt the OP is actually someone in charge of designing controls for a factory but who knows these days. My guess is a student.

This is a classic debate. Should I have 1 controller that controls 10 things or 10 controllers that each control 1.

If you have one controller and it goes down all 10 things stop. If you have 10 controllers you are 10 times as likely that something will go wrong (not really but you get my point).

I'll give you an example of a machine that I have dealt with that I thought should have had one or two PLCs.

The machine is a pipe mill, about 600' long with several different sections. This machine had 10 controllers on it and was designed in 1999. Several OEM's were involved with building it which explains the mess. Here is what everything did.

- Main OEM decided to split machine up into three sections. Accumulator, mill, and cutoff (3 SLC 5/04,s DH+)

- Entry OEM had their own 5/04 but where apparently too stupid to figure out a simple linear analog brake pressure controller for the mandrel brake and bought a DL-something and little HMI to do that.

- Welder had another 5/04

- 3 seam annealers each ahd their own 5/03

- Drive guy decided he needed an Automax controller for the drives (Reliance Flex Pak 3000).

- HMI guy decided he liked wonderware so main HMI was Intouch, the rest were panelviews.

- Motion control was necessary on the cutoff so an Indramat controller and a SOT (HMI) was installed there.

- Hydraulic control was necessary also on the cutoff so 2 rexroth HNC's were installed.

What a nightmare!. All of the PLC duties could have been handled easily by 1 5/04.

At the time the drive stuff was reasonable. Hey why not install the same system that you have been installing for the last decade even though it is about to be obsolete. Since the Reliance DCS won't talk DH+ lets install a third party protocol converter so it can talk to the PLC's, brilliant!

The HNC's had a bad habit of frying the 24V valve outputs if they got got shorted (approx 3K a pop), not to mention other problems, So I pushed for a Delta RMC. They were smart enough to go with a RMC after I left but too dumb to let one controller control both axis. So they bought two RMC's and networked them both with profibus (plant manager figured it was redundancy)

My point is that a machine (system) should be made as simple as possible and still do the job. If you have a million I/O points buy a fancy DCS, most of us don't.