Flow control using dosing pump and Coriolis flow meter

[mylespetro]

Hey everyone,

I was on a customer's site on friday where they have a large bioreactor that they bought secondhand from a previous user. System is controlled by a SLC 5/04 and one of the feeds into the reactor is glucose. I'm not sure what sort of pump was used for the feed, but the customer has installed dosing pumps for most of the ingredient feeds. This has worked out pretty well for the most part, but the glucose is giving them problems lately.

The original control scheme was pump -> Coriolis flow meter -> pneumatic control valve -> bioreactor. Just to reiterate, I'm not sure what kind of feed pump was used in the original install. The new control scheme is the same, except they're sending a 4-20mA to the dosing pump (4-20mA = 0-30L/h) and have the control valve set to a certain percentage to maintain a little bit of back pressure on the line, keeping the Coriolis meter full. However, this has caused the loop to be unable to be controlled even in manual. If we set the dosing pump to 50%, the flow measurement (0-600g/m, don't love the mismatch in engineering units, but we've been unable to open the HMI file to edit it) can be anywhere from 150-300 g/m, and fluctuating +/- 50g/m. I think this is a combination of having the control valve closed off to maintain back pressure and the periodic nature of the pulse on the dosing pump going through the Coriolis meter.

I feel like ideally either the feed would be more of a constant flow and the control valve would regulate it based on the flow measured by the Coriolis meter, or you would just have an on/off valve (for isolation from CIP/SIP backflow into the line) and send the desired rate straight to the dosing pump and assume it's pumping the correct/desired amount, allowing it to pump freely into the tank. I just feel like a dosing pump on a Coriolis meter isn't an ideal setup, but I admittedly don't have a ton of experience in the biochemical processing world and was just curious what everyone else's thoughts are on this.

Thanks in advance!

 

[parky]

Dosing pumps tend to pulse so not a good with flow meters

 

[drbitboy]

what is the timescale of the fluctuation? is the timescale dependent on the dosing pump speed? is the dosing pump a positive displacement (PD) pump (that seems the most likely cause)?

what is the sampling interval of the Coriolis meter? would it be reasonable to do a moving average, or other filter, of many samples, i.e. over a timescale much longer than that of the fluctuation?

if it is a PD pump, and the pressure rise does not vary greatly, then it may be adequate to calibrate the flow rate against pump speed, and use the Coriolis meter only to detect when the pump is failing (e.g. hole in diaphragm).

 

[mylespetro]

Dosing pumps tend to pulse so not a good with flow meters

This was my thought, it should be either one or the other. My gut feeling is to revert to a regular feed pump and let the control valve do its thing.

what is the timescale of the fluctuation? is the timescale dependent on the dosing pump speed? is the dosing pump a positive displacement pump (that seems the most likely cause)?


what is the sampling interval of the Coriolis meter? would it be reasonable to do a moving average, or other filter, of many samples, i.e. over a timescale much longer than that of the fluctuation?

If I had to estimate, I would say it could fluctuate from 150 g/m to 100 then up to 200 within 10 seconds. Pretty sure it's a PD dosing pump. Unsure about the sampling interval on the Coriolis. Moving/rolling average could work, but I just feel the whole setup isn't conducive to good control, as a fixed output on the pump isn't providing a steady flow whatsoever.

Really just in the preliminary stages of addressing the issue now, so nothing is off the table as far as I'm concerned.

Thanks for the input!

 

[doctord]

the flow meters I have worked with before also need to be purged, by that I mean no air in the meter are you can get wrong readings

 

[drbitboy]

P.S. the control valve maintaining back pressure is probably reducing the measured variation, though slightly.

The process needs capacitance, whether via software,

• e.g. a moving average or other filter of meter readings,

or hardware

• e.g. a tank upstream of Coriolis meter with a controlled level that provides
  • either variable head to control the flow through a fixed control valve setting (cf. here - gotta love a paper when the first reference is BS&L ),
  • or fixed head with flow controlled via the valve position;
• The "tank| (capacitance) could be a single tall piece of pipe fed by the dosing pump and drained to the control valve;
• this all assumes the process operates near atmospheric pressure.

 

[drbitboy]

within 10 seconds.

Is that consistent with what you see the pump doing i.e. is there a connecting rod that could be driven by a crank at 6RPM?


Let's say it is, and setpoint is 150g/m, and specific gravity is 1. so 150g/m = 150ml/m = 2.5ml/s, so 25ml per 10s, so 50ml would be the volume of one stroke, assuming cylindrical piston and D = stroke ~ 4cm.

I am jsut guessing; does any of that seem reasonable with what you are looking at?

 

[JTCat]

I've used Coriolis flow meters to calibrate dosing pumps. I assisted a massive inline blending operating using Emerson Coriolis meters that used gear pumps with a bypass line to measure dosing. The gear pumps always ran, a three way valve opened and closed when the batch called for an ingredient. The meter therefore was never subject to line pressure. Pressure can interfere with the vibrators and pick-ups

 

[parky]

Coriolis meters use either a single or double vibration tube, pulsing of product will have an adverse effect on it's measuring capability, as will air bubbles.
If you want to use a flow meter for controlling dosing then the only reliable way is not to use that type of pump.

 

[widelto]

I learnt the hard way too many years ago, piston or pulsing pumps do not get well with flowmeters.

 

[mylespetro]

Hey everyone, thanks for the replies. Office closed early yesterday due to snow so I didn’t get to reply sooner, but I’m onsite now gathering some info. I had a feeling that the pulse wasn’t going to work well with the flow meter, so now we have to decide what to do with the system as it is. As I said previously, I think either some sort of feed pump with a steady/constant flow, or change the control to open loop, letting the dosing pump do its thing and take its reading as gospel.

I’m taking some videos and pictures of the trends and equipment etc, and will try to remember to upload them when I get back to the office.

Thanks again everyone

 

[mylespetro]

Back at the office now, had a mildly successful attempt at controlling the loop in auto. With the PB set to 500% (P gain of 0.2) and an integral term of 1min/repeat, the loop was fairly stable at setpoint, fluctuating +/- a couple percent. Not as tight as I would like, but again, the PV is pretty noisy. The batch process they run starts with a setpoint of 31g/m, so I put that in to check the response, and it was slow to reach setpoint, but once it reached approximately 31, it kind of bounced around from 30-40. The customer said this is actually fine, because any excess glucose will be consumed by the reaction down the line. With the setpoint at 200, it was a bit tighter, bouncing around from around 197-206, sometimes going a little bit outside this range.

I think the small oscillation at the lower setpoint is due to the fact that the PV reading seems to oscillate around 12.5% output on the loop, and once it gets above that it starts becoming more stable, however, at a setpoint of 31, the output seems to hover around 13-15.5%, so I think it's just a bit noisier down that low because of the longer time between pulses.

While the customer was happy with this performance for the time being, I view it as a temporary solution and recommended the following options, from cheapest to most expensive:

1. Convert to open loop control, having an on/off solenoid to isolate the line when the process isn't running, and full open when glucose is required. Allow the dosing pump to pump freely into the vessel and trust that it's pumping the indicated amount.

2. Install a pulsation damper in the system as it is now to smooth out the pulses and keep the flow a bit more constant through the Coriolis meter.

3. Install a positive displacement feed pump and revert to the original control scheme, allowing the control valve in the line to regulate the flow into the tank.

I'm working on getting some pictures of the trends on my computer and uploaded, will post when I do so.

 

[drbitboy]

So the primary cause of the large oscillations was less-than-ideal PID tuning, and pump pulsation was a minor cause?

What is the timeframe of the oscillations now?

What is the loop update time?

.

 

[drbitboy]

If there is a bypass or recirc valve on the pump, cracking it a bit to allow some recirc might settle things a bit (and make the pump run faster for a given flow), although that might require re-tuning, and it is not a good long-term solution.

 

[mylespetro]

So the primary cause of the large oscillations was less-than-ideal PID tuning, and pump pulsation was a minor cause?

What is the timeframe of the oscillations now?

What is the loop update time?

.

I think the oscillations are a combination of the noise on the PV from the pulses from the dosing pump which in turn are causing a bit of oscillation on the output of the PID. I think this is why the P term is set so low, and we actually turned up the integral action quite a bit to ideally integrate the error out eventually while avoiding oscillation from a higher P term. This seemed to work fairly well, as when we were onsite on Friday, the P term was 0.2 (500% PB) and the integral action was 6min/repeat. With these tuning values, the loop was quite unstable in auto. As I said in my post from earlier today, we lowered the integral action to 1min/repeat, and the performance was much improved, if slow.

The oscillations aren't quite regular, but they're pretty slow. Didn't think to check the loop update time, and unfortunately our computer with an RSLogix500 license on it is still onsite there hooked up to the PLC, so I can't check the loop update time right at this moment.

If there is a bypass or recirc valve on the pump, cracking it a bit to allow some recirc might settle things a bit (and make the pump run faster for a given flow), although that might require re-tuning, and it is not a good long-term solution.

Not exactly positive if the pump has a recirc, it's a Grundfos DDA dosing pump, which I'm not overly familiar with. I'll look into it and check it out.

Thanks for the feedback, wish I could offer a bit more info, but I don't have the information on hand when I'm at the office.