Pressure application

I've done several CIP systems on a plc in large spiral freezers etc. If its piped correctly there is no reason why you cant do it with a modulating return valve or VSD.

i'm sure you are aware of the non linear effects of VSD on centrifugal pumps.

Personally I would prefer a variable/modulating valve.

sQuick..

I think I'd rather the valve also. The pump will have to work a little harder if you use the valve, but that's life. There may be a problem getting the right sized valve to handle the entire control range with the pressures required.

I'd seriously consider using a positive displacement pump and a dump valve for this application. Use a dump valve to lose the extra pressure. Each zone can have their own dump valve so that it only needs a shutoff valve to run. That makes things quite easy.

Michael

Michael

The closest I've come to your problem is with itty bitty air pumps and valves, small enough to mount on a PC board. So I don't know enough to share your trepidation, but I've seen other smart people trepidating about this sort of thing on this group, so I trust that it is, indeed, a hard nut to crack.

It seems that your problem is both a technical one and a political one

-- technical because you are going down a path that you fear may not succeed, and political because you are being set on that path by someone you are working with.

I can think of some things that I might do to resolve this sort of issue.

• Line up a bunch of guys with experience who say it's a bad idea (or who tell you how to do it right). This is what you seem to be doing here, I hope it's all you need.

• If the guy you're working with is generally reasonable but gets stuck on things every once in a while I'd take the 'dumb yokel' approach: "gee Mr. Chemical Engineer, I'm just a dumb PLC programmer. Every time I or any one of my friends tries to do this it fails. Could you please detail how you make it work?" If he's done it before with success grill him (nicely, naively) on how that job was like this job, and why the differences won't make a difference. If he gets over your head revert back to "I'm just a _dumb_ PLC programmer" and make him simplify. Hopefully he'll start to explain, stop in the middle and say "oh sh--" and hurriedly order a valve for you.

• An alternative to the "dumb PLC programmer" is the "prima donna PLC programmer": "I'm sorry sir -- I don't believe that I can make this work, and I refuse to guarantee my results. I'll take your money for spending time on it, but you won't get any refunds if it fails." In my experience this is a battle you want to choose carefully -- you can really get people's back up doing this, and make them stick to their guns even after it's obvious that it's going to be a screwup.

• Another thing I'd consider doing would be to prototype the hard part

-- the pump with a head on it -- and (a) see if I could control it and if not (b) show the setup to him and get _him_ to try to make it work. This approach has the difficulty that you have to get working pieces: either you need to get the pump in question going without a bunch of effort, or you need to get a small pump and then argue about how representative it is.

Are you an employee or a contractor? If your'e an employee, sorry for your luck. If you are a contractor, don't dive into a job that you know to be a sure-fire cluster-f*ck. While SuperChemist aims the shots the target will be your backside.

I've had those jobs...where at the end I would have had more dollars in the bank if I had wrote the customer a check for $5000 and told him "screw you, I don't want the job!".

Depending on the flow and the gain of the valve, operating on the pump relief valve is not a desirable condition. Heat, valve chatter, pressure oscillations are typical results. Most of all, pump flow can drop off a cliff once the valve starts to open. Having said that, the temperature rise across a centrifugal pump can sky rocket at low flows. Also, controlling flow with a centrifugal pump can be a challenge since the pressure rise across the pump is a function of pump speed, system flow demand and the system line loss. It can get complicated---typically, a metering valve (or orifice) along with a pressure regulator to keep the Delta P across the metering valve constant is used to control (or vary) flow using a centrifugal pump. MLD

Your concerns are valid. There's a particular problem controlling flow with centrifugal pump speed where the static head is high, in that the gain between flow and speed can be very high, leading to (possibly) uncontrollable instabiity. One possible solution is to add some hydraulic resistance into the system, there may already be enough via the nozzles themselves. If not, then just putting some manual throttling valves in the lines may reduce the gain enough to make control using pump speed acceptable.

Prior analysis may or may not give you a clear answer on whether things are going to be OK. It would be worth planning for some manually adjustable throttling in the lines, eg. allow for easy isertion of valves, ensure there's some excess discharge head available from the pump. The system will certainly become more controllable as the resistance increases. The point at which things become 'OK' is probably something to find out live.

I have not experience with the CIP projects but I do with hydraulic motion, pressure and force control. I too have seen difficulties in controling a pump to control flow or pressue. due to the slow response compared to quickly changing loads.. It is far better to have an accumulator on the output of the pump and a flow control valve after that. The accumulator will slow down the pressure changes which will make the control of the pump less critical and much easier. The valve will have a relatively constant pressure source and is much faster responding than the pump motor. I know it cost more money but it can be thought of as insurance.

After you get into the project a bit you should find the RPM the pump needs to run at for each valve combination. You should use this knowledge to provide a feed forward or bias for your PID. This way you don't need to rely on the integrator winding up or down to match the new flow requirements for the new valve combination. If done right you should be able to run the pump with the bias and proportional gain.

Use the past to predict the future. Make your system learn. Hopefully the pump will find a constant RPM to run at for each combination. Update your bias data using this RPM. You may wish to filter the changes a bit. This way you system will almost self tune and adapt as the system ages.

application?

I had to do just that in the previous case I mentioned. After much headache it was the only way to reliably control the process. I like to use standard control methodology whenever possible and hate to complicate a process, but sometimes the initial designer leaves you no other choice :)

Years ago I installed a nutating disc flowmeter that had an electronic head that received pulses from a reed switch. The unit controlled a pneumatic ball valve downstream. The program in the unit had a parameter that compensated for valve wear, air pressure variations, or anything else that might cause the batch to overrun set point. I thought it useful and have implemented that particular type of logic in a PLC for a couple of different applications. Useful indeed.

Either a VFD pump or a pump/valve system can be capable of meeting your needs. The problem that you will face is the problem if non-linearity.

For example, "the zones differ by as much as 110 feet", means roughly

45 psi of difference in head. Since your nozzle pressure requirements are 110 psi, you can expect at least 40% difference in process gain from this alone. Factor in longer 135 different pipe runs, with different bends, and you can expect a highly non-linear system response.

The non-linearity will cause the pressure/flow controls to appear slow and sluggish for some nozzles, and unstable for others. The trick is to find the "right" tuning for this system. If the system is highly non-linear, you may also need to apply a characterizer to the control loop.

Ideally, you could find the tuning by doing a "bump and tune" for each nozzle. But I am going to assume that this is unlikely, due to time constraints. You should try to find "robust" controller tuning, that will keep the system stable under all conditions.

This article from Control magazine may be helpful to you:

You can find some more articles on tuning, using "Relative Response Time", at

Good Luck!

-George

BobG wrote: