[curiouscat]

I have a reaction well-studied and optimized at kilo-scale (~1 kg batches). When I moved it to the Pilot Plant (~100 kg batch) there's a drastic fall in reaction rate. Batch times shot up to ~3 hours from what used to be 45 minutes in the 1kg-lab-reactor. Yields are approximately unaffected.

Geometrically the units are fairly similar; is such a reduction in efficiency typical? The only factor I can think of is rpm; was about 500 rpm in lab and is now about 150 rpm in the Pilot Plant.

What's a good scale-up criterion used by Development Chemists: constant rpm (sounds improbable), constant tip velocity (= r x ω ), or some other?

Out of curiosity, what's the max rpm stirrer anyone has seen installed on a Pilot Unit (~100 - 500 kg). All sorts of fancy stirrer gizmos are advertised for lab units (some going as high as 20,000 rpm!) But can any of this wizardry be pulled off in the larger units?

[eazye1334]

From my experience, a loss of efficiency such as what you are seeing is certainly possible. I had a similar situation where a very well-understood reaction in the lab took ~2 hours and the pilot scale took 4 or 5.

The fact that your yield is unaffected signals to me that it is a scale-up efficiency loss. Depending on your safe operating window, I would try tweaking some of your parameters and see if you can't speed it up some, say increase heat and stirring (if possible). Usually we scaled up using tip speed, but even that didn't always work for us (we were making chemical toners). Most times the lab scale worked as a representation and the process needed to be adjusted to make a viable pilot scale run, and then adjusted again to move up to production rates.

Our pilot scale was ~20 gallon reactors, so figure on ~75 kg. We were up to well over 300 rpm on our mixer, so I would definitely say you could go faster, but that all relates to your specific reaction as well.

[curiouscat]

From my experience, a loss of efficiency such as what you are seeing is certainly possible. I had a similar situation where a very well-understood reaction in the lab took ~2 hours and the pilot scale took 4 or 5.

Glad to have your example.

The fact that your yield is unaffected signals to me that it is a scale-up efficiency loss. Depending on your safe operating window, I would try tweaking some of your parameters and see if you can't speed it up some, say increase heat and stirring (if possible). Usually we scaled up using tip speed, but even that didn't always work for us (we were making chemical toners). Most times the lab scale worked as a representation and the process needed to be adjusted to make a viable pilot scale run, and then adjusted again to move up to production rates.

We need cooling; it is an exothermic reaction. The lab protocol asked to maintain temp. ~ 0 C. And it seems our pilot-unit has enough cooling capacity to maintain the same Temp. range. My worry is how to model the efficiency loss even if I may not be able to prevent it (worst case). i.e. From lab to pilot was a 1:100 Scale-Up factor; from Pilot to Production I anticipate another 1:50 factor (~5000 kg batch size). I want to try and predict the batch times as closely as I can.

Any ideas?

Our pilot scale was ~20 gallon reactors, so figure on ~75 kg. We were up to well over 300 rpm on our mixer, so I would definitely say you could go faster, but that all relates to your specific reaction as well.

Thanks! I'll ask our vendors how high they can go on rpm. Especially on the Production Reactor (~5000 kg) 300 rpm may be pushing the limit. (I'm glad to be corrected! Do any of you operate  1000 kg or larger reactors with stirrers faster than that?)

The other thing that is an option is to add more impellers on that same shaft.

[eazye1334]

We need cooling; it is an exothermic reaction. The lab protocol asked to maintain temp. ~ 0 C. And it seems our pilot-unit has enough cooling capacity to maintain the same Temp. range. My worry is how to model the efficiency loss even if I may not be able to prevent it (worst case). i.e. From lab to pilot was a 1:100 Scale-Up factor; from Pilot to Production I anticipate another 1:50 factor (~5000 kg batch size). I want to try and predict the batch times as closely as I can.

Any ideas?

Unfortunately, no. Our design team was never really able to work out a model that fit terribly well so 1) I don't really know how they did it, and 2) I probably wouldn't want to use theirs anyway! I don't have much experience scaling exothermic reactions, so I can't tell you their effectiveness when scaling (if it's really appreciably different than endothermic anyway).

Thanks! I'll ask our vendors how high they can go on rpm. Especially on the Production Reactor (~5000 kg) 300 rpm may be pushing the limit. (I'm glad to be corrected! Do any of you operate  1000 kg or larger reactors with stirrers faster than that?)

The other thing that is an option is to add more impellers on that same shaft.

That's actually exactly what we did for production: larger baffles and more impellers. The production tanks were very large (5000 gallons, so ~19,000 kg) and the toner was not much more viscous than water, so increasing rpms on the mixer caused too much aeration and lots of foaming. Speeds were actually slower in production than in pilot, they just added extra impellers and expanded the baffles which seemed to do the job nicely. Definitely something I would check out as well.

[curiouscat]

Thanks! Very glad to have your advice!

Larger baffles is something I hadn't thought of. We have standard D/10 baffles. 4 total.

To throw ideas at people; another thought I have is to add a pumped circulation loop to the reactor with an inline static mixer and a heat exchanger.

Would definitely help with cooling; but not sure how much it'd help with the mixing.

[eazye1334]

To throw ideas at people; another thought I have is to add a pumped circulation loop to the reactor with an inline static mixer and a heat exchanger.

Would definitely help with cooling; but not sure how much it'd help with the mixing.

That actually might help. I haven't done something like that for an application like yours, but I have used inline high shear mixers for some incredibly viscous fluids and it definitely helped keep things flowing. Not exactly the same scenario, but I could see this helping. Like you said, it would at least help with cooling which certainly isn't going to slow you down at all. Divide and conquer!

[discodermolide]

May I ask what solvent you are using and is 0°C really required?

Get the lab to do one reaction at the solvent reflux temperature.

[curiouscat]

May I ask what solvent you are using and is 0°C really required?
Get the lab to do one reaction at the solvent reflux temperature.

Of course, you may. Glad you asked. It's essentially a solvent-less reaction; both reactants are liquid. H2SO4 (~65%) and  a higher alkene.

I'm assuming you are thinking about using the boiling solvent as an effective way to get rid of the heat of reaction, correct? If so, I had indeed considered it; but gave up since I couldn't think of an effective solvent with a 0 C BP. Any ideas? Maybe I misunderstand you?

The protocol I had from a previous experimental program claims that we lose selectivity at temps. above 0 C. I haven't investigated that line in much detail ; so far I was assuming they are right. But I could try more high-Temp. expts. if you think it's worth a shot.

[discodermolide]

Yes I was thinking along those lines of using the solvent reflux to remove the heat. But as you are doing it neat and the selectivity gets worse at higher temperature keep with what you've got.
Looks like you are going to require extra special stirring for this one.

[curiouscat]

Yes I was thinking along those lines of using the solvent reflux to remove the heat. But as you are doing it neat and the selectivity gets worse at higher temperature keep with what you've got.
Looks like you are going to require extra special stirring for this one.

One idea I had was, of applying a vacuum to reduce the BP of the system down to 0 C ; and then I could use the reflux-to-get-rid-heat idea. I haven't done any tests or calculations yet; but what do you think? The alkene itself boils at 100+ C at room temp. so might be quite a high vacuum we need though.

Another idea was to add a low-boiling solvent just to get rid heat (and it would probably aid stirring too by reducing viscosity). Solvent separation shouldn't be too bad since the reaction mix is pretty high boiling.

But I'm still having trouble coming up with a suitable solvent / solvents.

[discodermolide]

I don't know the scale you are doing this reaction but I would guess it may be very hard to get the vacuum down far enough so that the alkene boils at 0°C.
If you can do it then dichloromethane may be a candidate. However the acid is not going to be miscible with DCM.

[curiouscat]

I don't know the scale you are doing this reaction but I would guess it may be very hard to get the vacuum down far enough so that the alkene boils at 0°C.

Thanks for your tips discodermolide! Right now I'm trying out small batches but I'll have to scale-up to production scale (~5 tons/day or so ) eventually..

[discodermolide]

5 tons/day, small scale stuff then

[curiouscat]

5 tons/day, small scale stuff then

If we made 50 we'd exceed the world demand....