Process Puzzler: Sidestep a Scale-up Scare

March 4, 2009
Readers grapple with a reactor production change.

Management wants to boost production from our batch polymer process by changing the operating procedure. Now, a portion of the acrylic monomer goes in with the bulk of the solvent and other ingredients. Then, monomer is slowly added until the reactor level rises to about 55%. Instead, management wants to increase the maximum level to about 65% and raise the final concentration of the polymer about 2%. The reaction is very exothermic; a chilled-water baffled jacket provides cooling. At the top of the reactor a horizontal shell-and-tube condenser that uses cooling tower water captures evaporated solvent. Agitation consists of a single axial impeller and the tank has reduced baffles to account for viscosity above 5,000 cP. To meet the production goal, it’s been proposed that we add all of the monomer at once; bench-scale tests have shown that this could work but our current approach seems safer. Management likes the plan and wants to put into practice immediately. Our production engineer is a little nervous — should he be concerned? How should we approach this problem? Are any process changes necessary? What do you think?

I would recommend using a static mixer with an integral heat exchanger to blend the materials before they go into the reactor. The static mixer/heat exchanger could aid in reducing the temperature of the exothermic reaction before fluid is introduced into the reactor. Controlling the temperature before it is introduced into the reactor is a safer solution [than adding the monomer all at once] by allowing the reactor and the mass inside to act as a heat sink/quench tank. Two mass flow meters, one for the monomer and the one for the solvents, would monitor flow. If the temperature in the mixer/exchanger increases above safe thresholds, you can continue to add the solvents while reducing the monomer level, thereby maximizing the monomer addition and eliminating risk. You may want to have multiple feed lines [for the solvent] with isolation into the reactor for faster addition, better control, and in case pluggage occurs or cleaning is required.

Fast temperature sensing at the exit of the exchanger and in the reactor, with either adaptive or artificial-intelligence temperature control, will be necessary to fine-tune and predict the reactor temperature.

Using this system eliminates the possibility of thermal runaway in the reactor. I agree with your production engineer, as once the reaction starts to run away it is too late. Piping is much easier to clean [and contain] than a reactor if an incident were to occur.
Joel Heidbreder, principal engineer
Monsanto Co., St. Louis

Your production engineer’s concern is perfectly right as far as safety is concerned. Adequate cooling has to be ensured as the rate of heat generation will be much greater [with the process change]. Try a continuous plug-flow-type reactor as it can handle higher concentrations and is safer.
Nandan Bhandari, managing director
Maple Biotech Pvt., Ltd., Pune, India

Yes, the engineer is right, they should be careful. If this is a free radical solution or emulsion reaction, as I assume, then it is a chain reaction that could lead to an explosion. This is especially true if all monomer is added at once; the temperature will rise out of control.

Instead, to increase the conversion, increase the initiator concentration, but not [enough] to affect the molecular weight, and lower the solvent concentration in the formula to increase polymer concentration. Use bubbling nitrogen inert gas and change tower water to chiller water in the condenser to avoid monomer evaporation further and lower inner pressure.
Emilio Malaguti, technical manager,
Chemtron, Hialeah, Fla.

Yikes! Sounds like “Synthron” all over again. Yes, there are significant process hazards if the reactor temperature is able to reach the “point of no return,” accelerating the exothermic reaction exponentially. Play the Chemical Safety Board video for management and see how they feel afterwards.
Chris Rentsch, process engineer
Dow AgroSciences, Midland, Mich.

With increased batch quantity, the agitation and cooling will have to increase. Normally, in a [batch] reactor, control of the rate of addition is very important to achieve the quality and yield. So to meet these requirements you have to make changes in the process procedures. You may want to look for an alternative catalyst. Finally if nothing works, adding another reactor would be better.
B.S. Gadodia, consultant
Indore, India

They should consider the following steps: 1) estimate or find by experiment the heat of reaction; 2) calculate the adiabatic temperature rise if all the reaction occurs instantly without cooling; 3) if the heat in step 2 is not enough to boil the reactor contents then all is okay to proceed with caution; 4) if the heat in step 2 can boil the reactor contents then proceed with great caution; 5) test the process using adiabatic reaction calorimetry to discover if the process’s behavior changes with the change in conditions (it is not uncommon for polymerizations to accelerate with temperature or to have competing reactions dominate at higher temperatures); 6) if maintaining a lower temperature becomes essential use the “efficiency method” rather than the usual log-mean, as it can handle transient conditions much more effectively; and 7) if in doubt, contact a chemical process engineer who understands the reasons for the steps above.
Ed Fish, associate consultant
Haztech Consultants, Winsford, U. K.

The engineer is right to be a little nervous. There are multiple changes happening here: addition rate, addition amounts, changing agitation pattern, and changing heat transfer surface. Small scale results often work well because the ratio of heat-transfer surface area to reactor volume is high but the ratio is much lower at larger scales. It shouldn’t hurt to make the changes incrementally, one at a time, over the course of weeks or a month. The results will show feasibility or demonstrate the system limits without making the six o’clock news.

Increasing the level and concentration means there is more reacting mass and likely a faster reaction rate. The level rise will increase the reactor heat-transfer area, but not in proportion to the additional duty. Further, the partial reactor baffling may mean that the increased level is in a part of the tank that is poorly mixed and has a low heat transfer coefficient.

On the other hand, it may be that the bulk of the heat transfer needs of the system are met by condensing evaporated solvent in the overhead condenser. In that case, someone needs to verify that the condenser and the vapor/condensate piping can support the additional load.

You can perform a number of calculations to check if the system can handle the extra load, but it’s easier, quicker and more accurate to slowly scale-up to the new conditions. I would start by increasing the rate of the initial addition; if all goes well, then either increase batch size or concentration or both in small increments. The traditional slow addition of the starting monomer charge may be needed to get the reaction initiated without “sandbagging” (no initial reaction followed by a sharp, maybe uncontrollable, spike in temperature). You want to demonstrate that this is not a problem before you have a lot of extra monomer in the vessel.
Alexander E. Smith Jr., engineer
Parsons Corp., Boston

Even though bench-scale tests show that adding all the monomers at once may be OK, it will be unsafe to directly implement it in the actual plant reactor. This will be mainly due to inadequate mixing and hence very high localized reaction rates in certain parts of the reactor. This situation may pose problems with reaction temperature control and hence may lead to a possible runaway reaction (due to exothermic nature of the reaction).

The best way to tackle this problem is to increase the monomer addition rate to maximum possible value (with present reactor and cooling system design) in small steps. Each time increase the monomer addition rate by a small step based on actual reaction controllability and product quality. If successful, increase the monomer addition rate further with a careful eye on the reaction controllability and product quality.
C. C. S. Reddy, lead process design engineer
Singapore Refining Company Pte. Ltd., Singapore

The process manager has a right to be concerned. As many others will mention, the surface area to volume ratio will change as reactor size increases. For a process that is cooled by a jacketed shell (surface cooling), the relative cooling capacity will decrease as a reactor gets larger. For this reason the results of a bench-scale test may not be an accurate predictor in a larger reactor. The other complication in this scenario is the fact that it’s an exothermic polymerization. As the reaction proceeds the increase in viscosity will hinder cooling at the same time the evolved heat will increase the reaction rate. This can lead to a runaway reaction. It would be best to proceed to the final goal of all monomer in at once in small steps with careful monitoring. One potential way to monitor this is to measure the heat removal from the jacket and the top-mounted condenser separately. The top-mounted condenser would be a faster indicator of heat evolution because it cools the boiled solvent, which is an indicator of the bulk temperature. One potential approach would be to see if it would also be possible to temporarily decrease the polymer concentration to help mitigate the effects of viscosity on cooling.
Rob Vermeulen, engineer
Metron Technology, Boulder, Colo.

Changing from semi-batch, by slowly adding monomer, to batch, by charging all the monomer at once, can be dangerous. In semi-batch mode, a limited amount of energy in the form of unreacted monomer is available at any point in time. In batch mode, the entire energy release of the reaction is available at once. Should the reaction proceed adiabatically, as it would in the case of loss of cooling or reflux, the final temperature and therefore pressure due to the vapor pressure of the final mixture could be quite high. The elevated temperatures will also greatly increase reaction rate. Worse, at this elevated temperature, exothermic or gas-producing decomposition reactions may initiate, exacerbating an already bad situation. Before instituting batch-mode production, you must: 1) quantify the heat of reaction for the desired reaction; 2) determine the final temperature and pressure (due to vapor pressure) if the reaction should occur adiabatically; and 3) test the final product for thermal stability from room temperature up to the maximum achievable adiabatic temperature rise. If the final temperature or pressure of the adiabatic reaction exceeds the vessel’s design capacity or if thermal stability testing shows a decomposition reaction, adequate emergency relief must be designed for the system (as the existing emergency relief system my not be adequate). This will likely require testing in an accelerating rate calorimeter or vent sizing package testing to obtain maximum rates of temperature and pressure rise, as well as consideration of two-phase relief flow.
John C. Wincek, process safety manager
Croda, Inc., Mill Hall, Pa.

You make no mention of molecular weight desired in the polymer. Normally, with the thermoplastics I’ve worked with, the higher the temperature, the lower the molecular weight. And I don’t see anything about mixing or blending. Most polymers achieve a narrow molecular range when a somewhat uniform temperature is achieved in the reactor. If one has nominally high production rates, maybe a continuous reactor is the best choice.
Tom Murphy, CEO
Puritrol, Inc., Centerville, Mass.

Your production manager should be concerned for the following reasons:
1. Changing the way you add your ingredients is extremely likely to change your final polymer. The average molecular weight and the distribution of molecular weights will change. Both of these will affect the quality of your product.

2. In your current reactor configuration, I presume that if you lost cooling in either the jacket or overhead condenser you would be able to stop adding monomer — which you won’t be able to do if you add all your monomer at once as in the new configuration.

3. Assuming that the vessel is closed, which seems likely, the pressure in the vessel will go up by a minimum of 29% just with the decrease in headspace from 45% to 35%. With the same cooling as you have now, the pressure will go up significantly if not catastrophically, as you will be giving off more heat in a shorter period of time. This, in turn, will drive the reaction, causing it to accelerate until something gives. Even if the vessel is vented much of the cooling is probably via the overhead condenser. The higher temperature anticipated will reduce the effectiveness of the condenser and more solvent will be lost.

4. Besides being a potential environmental concern, evaporation of the solvent caused by the heating will further increase the concentration of polymer. The viscosity will increase and reduce heat transfer from the jacket.

5. At some point the pressure relief system will relieve. The system will depressure, quickly causing intense foaming. Even with an open vent, which loses solvent, the higher liquid level and viscosity will tend to cause more foam. Foaming could reach and foul the condenser, reducing still further the ability to take heat out of the system. If either of these scenarios occurs, it could seal off your pressure relief system as the polymer comes out of solution due to solvent vaporization.

6. If the system does not blow up, be prepared for a giant molecule of polymer to form, which may require vessel replacement or jack hammering out polymer. In either event downtime would be significant.

You should start by exploring the thermodynamics. Find out exactly how large the exotherm is and then define the reaction rate at various temperatures. This is the only safe way to increase production; a logical progression vs. pushing all the yield parameters all at once. You will probably find that you can safely increase production to some extent. Beyond that you will need to make improvements to the process to increase your production.
Barry Bershad, consultant
Mansfield, Mass.

As a process engineer, I would be very concerned with scaling-up a solvent process from a bench scale to production scale with no in-between checks to verify the scale-up. Safety should be a primary concern.

Follow a gradual plan for scale-up. Scale-up first from the bench scale incrementally to 5 gallons and measure the temperature of the jacket-cooled batch to verify that there are no safety concerns or flashing. If the process stays within a safe temperature range you should move to the next step, which is a 20-gallon or 30-gallon batch. After verifying that the process is safe at that scale, next scale up to 50 gallons, then 100 gallons, then 250 gallons, then 500 gallons, and then 1,000 gallons. There are also other concerns that may arise with rapid addition of one raw material.

Things to look out for are two phases/phase separation/phase changes, viscosity changes, different agitation requirements, changes in consistency, etc. For agitation concerns, consult a known reliable vendor. If there are changes in the amount and or rate that the solvent evaporates, verify that the condenser above the reactor is sized accordingly. Also, measure the solvent concentration in the air on scale-up to verify the equipment has adequate venting. Battery operated portable meters are available.

Since you are also implementing a formula change by increasing the polymer concentration by 2%, you should also do a full quality-control analysis on each scale-up batch to verify that your targets are staying consistent. Other tips for improving efficiency and turnaround time are to pre-weigh and stage raw materials before addition, use a pump to discharge the product instead of discharging by gravity, also discharge into secondary containers like totes or a secondary tank if filling smaller packages. Depending on how the scale up goes you may need to upgrade your processing equipment.
Errol Williams, production engineer
AVEKA Inc., Woodbury, Minn.

Polymers are formulated for specific purposes. Changes in the process
could affect the final product. For instance, more or less monomer and
high heat and curing changes may make the polymer more or less appropriate for its intended use. Also, by forcing the reaction, will appropriate cross-linking occur for product performance and are studies needed?
Richard Ashley, associate director, technical affairs
Barr Pharmaceuticals, Pomona, N.Y.

The restraint of “slowly” adding the monomer implies that it is controlling the rate of reaction. The first step should be to increase the rate of addition in discrete steps to verify control of the reaction. Then the batch size may be increased if the head space has not been compromised by foaming or carryover as the vapor disengages. The objective may be attained in a safe manner and allay the nervous concern. The condenser will remove most of the heat of reaction as the heat transfer to the jacket will be negligible at high viscosities. Bench-scale equipment does not have the same concern for removing heat as a production-size reactor. Going directly to a bulk charge is not recommended. To go to a bulk polymerization requires a review of the heat removal capability of the condenser and the relief area of the reactor. Adjusting the free space in the reactor may be necessary if foaming is a concern or cannot be controlled by defoaming agents; if this is true, the faster reaction rate will aggravate the situation, possibly to an unacceptable level that may require smaller batch sizes rather than larger ones.
Jim Morris, senior process engineer
Flint Hills Resources, Odessa, Texas

Obviously the production increase should be delayed until it can be tested at a laboratory scale (1–2 gallons), then a bench scale (10–50 gallons) and finally a pre-production scale (100–200 gallons). But what will the results tell us? Perhaps, not enough.

I remember my first engineering assignment: to suggest the reason for an explosion caused during scale-up of a new solid rocket propellant. Tests at the laboratory scale provided few answers other than the obvious ones: mixing and temperature control decline with scale-up. For this problem it may be best to address the symptoms.

If the batch is thickened by chemical reaction and by concentration, agitation will decline and the bulk temperature will increase as surface area to volume decreases. Agitation cannot be easily improved for viscosities much above 100 cP. However, heat transfer can be increased by changing the media and temperature differential.

First, the solvent could be chilled prior to addition to the reactor. I suggest changing the cooling media for the jacket and the condenser to chilled brine; this may require a material change in the condenser and jacket from carbon steel. Use an operating minimum of about 40oF. Perhaps a maximum of 80oF would be acceptable. Changing the differential temperature might more than double the actual heat transfer.

Avoid a temperature so low that a phase change occurs as this will be a barrier to heat transfer.

An alternative coolant could be 30% propylene glycol (PPG) in water. This could avoid the cost of replacing the heat exchangers because of corrosion. PPG is a safer choice than ethylene glycol and less of an environmental risk.

Because it is too expensive to move the reactor volume through a pump, it may be possible to capture the solvent as it evaporates, chill it, and inject it into the reactor. Inerting the reactor may also help. By careful experimentation it may be found that additional solvent is best to control the temperature.

Now, we turn to product quality. Because it may not be desirable to have a product with the solvent, the temperature could be used to drive off the solvent at the end of the batch. This could work best under vacuum. It may be necessary to add an additional step of drying, perhaps vacuum drying with a drum.

All of these changes require a radical change in the process steps but might be managed with a modest investment of time, capital and research.
Dirk Willard, consultant
Highland, Ind.

We react two organic chemicals in a stirred tank. The reaction is exothermic and highly sensitive to temperature. We control the temperature by adjusting the feed flows, particularly reactant A, which makes up 75% of the flow. Reactant B is ratioed off reactant A.  (See Figure 1: Our new control engineer thinks we need to program some lag in the control valve for A. His first idea was to install an electric valve positioned on control valve A. The product quality has declined. Is he right about the lag? What other improvements should we consider?

Send us your comments, suggestions or solutions for this question by April 13, 2009. We’ll include as many of them as possible in the May 2009 issue and all on Send visuals — a sketch is fine. E-mail us at [email protected] or mail to Process Puzzler, Chemical Processing, 555 W. Pierce Road, Suite 301, Itasca, IL 60143. Fax: (630) 467-1120. Please include your name, title, location and company affiliation in the response.

And, of course, if you have a process problem you’d like to pose to our readers, send it along and we’ll be pleased to consider it for publication.