8 replies to this topic

[curious_cat]

We have a semi-batch reactor where three Liquids A, B & C are co-dosed. Better the control over proportions better is the batch yield. Ideally we'd want not more than 1% variation from design flow. All are low viscosity fluids with SG between 0.8 to 1.1. 

 

I was ruminating over three options (all attached as sketches) and would love opinions about pitfalls etc.:

 

Scheme 1:

 

Three dosing pumps (motor driven / diphragm type / simplex) with downstream flow rate sensors for totalization & validation. 

 

Issues: Not sure if without feedback a pump can consistently deliver a 1% accurate flowrate. Especially as dosing tank level falls. I've contacted vendors and they claim they can, but I'm still a bit skeptical.

 

Scheme 2: 

 

Dosing pump feedback controlled via the flow sensor. Probably by control of stroke length. VFD control of strokes / min is another option. Strictly speaking, the ratios being more critical & the actual flows  can be allowed to vary a bit, I was thinking a ratio control strategy might be a good fit?

 

Scheme 3:

 

Use CFG pumps & control a throttle valve via a flow sensor. 

 

This seems simplest in terms of cost, complexity & maintainace too but will it work? Especially since these are not large flow rates. 

 

Other details are on the diagrams, but if I left anything out please ask.

 

Notes:

 

(1) Ideally we don't need the dosing tanks at all. Filling them is a hassle. We had them in the default plan since a level gauge on them provides a passive backup to the dosing rate control.

 

(2) One of the fluids is H2O2 so has a bit of a gassing problem. That does make accurate dosing and measurement a challange. 

 

(3) Coriolis meters would be great as a strictly mass measurement sensor. But they are expensive. Good quality thermal / magnetic or ultrasonic sensors are cheaper and promise 0.5% accuracy so might be a reasonable compromise. 

 

(4) Other combinations might also work, e.g. a CFG pump with a bypass line controlled etc. Not sure how to choose.

 

(5) Is there utility to having a single driver drive all three pumps to compensate for any driver speed variations. So that the ratios are unaffected?

Attached Files

•  Scheme1_Page_1.jpg   58.36KB   6 downloads
•  Scheme2_Page_1.jpg   65.07KB   2 downloads
•  Scheme3_Page_1.jpg   61.37KB   2 downloads

Edited by curious_cat, 12 October 2013 - 04:45 AM.

[fseipel]

Milton Roy and many other vendors offer metering pumps with +-1% accuracy.  I would consider pulsation dampening in design if using metering pumps -- make sure meter can handle it or include a bladder.  You might want to consider level transmitters in the feed tanks -- you can then have DCS calculate rate of change in tank level vs flowmeter indications & alarm if there is significant disagreement between the two rate indications -- this also allows you to confirm there is enough stock before initiating any of the charge & won't be impacted by the Oxygen offgassing and also won't have cumulative error.  The metering pumps will not be impacted significantly by suction head changes as feed tank levels drop, and slippage should be minimal since you're discharging to an atmospheric pressure reactor.  Their Achilles heal is the check valves and diaphragms.  If you ALWAYS dose three chemicals a triplex head could be used & thus requiring only one motor -- motor speed (slippage) is going to be minimal, especially at low discharge pressures, so only advantage there is cost/wiring savings vs 3 motors, won't really improve accuracy.  Stroke length on each head could be varied via PID loop.  It may be cheaper to just vary stroke volume with 4-20 mA signal rather than use VFD to change speed.  I'd include discharge pressure gauges on pumps, isolation valves to allow pump servicing, and possibly drain/recirculation valves to aid in priming pumps and cleanouts if you switch over reagents.  May want to include 4th pump with hose jumpers to allow you to run in case of a pump failure in middle of run.  Liquids that offgas such as Hydrogen peroxide may tend to vapor lock a positive displacement pump, but as long as you vent it prior to start-up, it should run fine throughout batch running continuously.  You can also opt to use more dilute H2O2.  In situations prone to vapor lock I prefer centrifugal pumps with recirculation lines though I think metering pump is fine here.  If you have a steel superstructure and can elevate tanks you could also feed by gravity or pressure liquid over from tanks & go with control valve scheme (no pumps req'd).  Others may recommend against it, but in a situation where little is changing (pump discharge pressure, reactor backpressure nearly constant) simply using a vee cut ball valve with positioner is quite economical with the centrifugal pump option vs a conventional control valve -- also the seats are cheap.  If you don't truly need to co-meter, you could alternately feed for 5 min from tank A, 5 min from tank B, 5 min from tank C using actuated valves to switch between them, but that won't be as accurate.  If operators check level at 15 or 30 min intervals, may need need flowmeters -- can just make chart of target level vs time and they can adjust pump stroke volume to meet target levels.

Edited by fseipel, 12 October 2013 - 06:30 PM.

[breizh]

Regarding metering pumps  you may also consider Pulsafeeder brand , this is based on my experience with latex .

 

Breizh

[curious_cat]

@ fseipel: 

 

Thanks for those great tips. 

 

1. Any comments about what sort of sensing element you tend towards using? 

 

2. A solution I was especially attracted to is this sort of integrated dose+meter by Grundfos or Bronkhorst. 

 

I'd love to hear other vendors or models that offer an integrated dosing + measurement / control unit. 

 

http://us.grundfos.c...dosing-dme.html

e.g. DME 940-4

 

One problem I've faced in the past is the sensor vendor blaming pulsations and the pump vendor blaming a bad sensor or control loop. 

 

 

3. Pulsation dampning we already have in the plan. Sorry I should have mentioned that. So also, drains, bleed valves, isolation, strainers, pressure relief etc. I neglected to draw those in. 

 

4. Triplex pump is a great idea, do you know any models that allow three independent dosing's with a common driver? 

Edited by curious_cat, 13 October 2013 - 01:29 AM.

[fseipel]

I prefer coriolis mass flowmeters but they're expensive -- Micro Motion, E&H Promass, etc.  I concur w/Breizh, Pulsatron and other Pulsafeeders are reliable, especially if you keep spares kits (Kopkits).  Pulsatrons are good for +-2%, though with flowmeter, that should obviously be improved, to whatever flowmeter accuracy is -- I note this since you stated you needed +-1%.  I've never tried Grundfos dosing pumps, but the operator interface looks much better than the Pulsafeeders and I like the click-stop mounting -- these look much more sophisticated than any I've ever used from LMI or Pulsafeeder -- the integral pressure monitoring and auto flowadapt are great concepts as is use of stepper motor vs solenoid.  Pulsafeeder offers a flow monitoring option but I think it's just a flow switch.  The user interface on those new generation Grundfos appears to be an incredible step up from the klunky LMI&Pulsatron interface -- large display, multiple colors for alarms, wheel to select menu option -- I like the idea of programming pump without needing manual -- like an Endress&Hauser pH transmitter.  I can easily see the panel & they are +-1%, though again I don't think that's critical on the pump if you have a flow controller.  I suppose you could also just use 1 mass flowmeter for calibration (allow any pump to feed through it as a bypass) & then use extremely inexpensive meters such as turbine or paddle -- if calibrated at operating point against mass flowmeter at beginning of run, it won't matter if their zero and span aren't dead on; control loop will still work great.  If chemicals aren't reactive with one another, consider teeing them into a common line into reactor to save plumbing costs. Out of curiosity, are Grundfos metering pumps more expensive than Pulsafeeder or less & is lead time/availability as good as Pulsafeeder?  I've used Grundfos well water and multistage centrifugal pumps before.  LEWA manufactures triplex pumps with individually adjustable stroke length on each head & other vendors do as well.  If you're always running at the same rate calibrating inexpensive paddle or vortex meters may make sense IF there is little pulsation (effective dampener).  A 1/4" vortex meter in PVC is around $1000 from Cole Parmer (ASAHI meter); if you need x-proof, I/S or solvent compatibility this won't work.  Accuracy=1%.  Kobold's DVZ is ~$300 but they are brass/plastic.  I've also gotten by with Brooks variable area/rotameters with transmitters even without dampener -- just by filtering the input from the flowmeter.  Thermal mass flowmeters are inexpensive but need recalibrated if you change fluids.  They are especially well suited to gases.  Magmeters tend to be a good choice for conductive aqueous fluids (cheap, accurate).  Nutating disc and mini oval gearmeters from Tuthill Sotera et al are reasonably inexpensive.

Edited by fseipel, 13 October 2013 - 08:25 PM.

[curious_cat]

Thanks again @fseopel:

 

I don't have a quote from Grundfos yet so not sure. How much are similar Pulsafeeder  units?

 

Lead time isn't bad. 20 days they said, these are modular, mostly off the shelf units they said. 

 

There's also this very similar unit  by Bonkhorst called Coriflow. 

 

http://www.bronkhors...s/cori-flow.pdf

 

It looks pretty good too. Has integrated pump / Corolois MFC & a control loop.

 

 Most of their models are for smaller flow rates but I found one going up to 600 kg/hr. 

[fseipel]

I don't know enough about the head material to directly compare.  Pulsatrons don't go up to the flow rate you require but Pulsafeeder Omni's do -- around $2300 w/PP head for Omni DC6 -- of course this will vary with diaphragm materials, etc.  Once feature I always liked about Pulsafeeder was I could look up their prices online without need to quote.  Cori-flow controller looks nice.

[breizh]

Hi ,

I have seen  people feeding reactor by gravity flow , with load cells underneath the dosing tanks . The quantity of material was managed by PLC adjusting the opening of the feed valves by calculating the mass flow rate :(delta mass/delta time)  .  Probably not precise enough for your process , this was to produce resins.

 

Just for information.

 

Breizh

Edited by breizh, 14 October 2013 - 09:15 PM.

[curious_cat]

Hi ,

I have seen  people feeding reactor by gravity flow , with load cells underneath the dosing tanks . The quantity of material was managed by PLC adjusting the opening of the feed valves by calculating the mass flow rate :(delta mass/delta time)  .  Probably not precise enough for your process , this was to produce resins

 

Thanks. I've seen that too. But typically for smaller dosings. Often for solids. 

 

Our quantities will come to ~ 3000 kg each, over a 5 hour period. Not sure if a load cell is viable to fit to a 3-4 kL dosing tank and if so whether it will yield the +/1% accuracy we desire especially during early dosing stages.