Post Mixing Optimization and Solutions
Search:

Fermentation

The use of axial flow down-pumping agitators in biological processes

This article is based on Publications #39 , #42 and #44 and has been updated. Although the article does not require any graphics, they will be added soon. The article will also be hyperlinked to other references within the Post Mixing Web Site.



Abstract

The most common impeller style in fermenters is the Rushton turbine and it has been doing a fine job since World War II. In comparison, the use of high-solidity, axial flow down-pumping impellers in fermenters has resulted in a wide variety of both positive and negative results. Much of this comes from the complexity of the biological process and the role mixing plays on fermentations. Model tests were performed in 0.46 and 1.17-meter tanks at Lightnin to try and simulate effects noticed in large-scale fermenters. It is a remarkable result that the high-solidity, axial flow down-pumping Lightnin A315 can achieve the same gas-liquid mass transfer rates with much less shear than the radial flow Rushton turbine (RT). Mixing limited conditions favors the A315 setup, too. Where mixing is not a problem (small-scale and/or high power) the gas-liquid mass transfer results are identical for both radial and axial flow impeller systems. This seems to indicate that the Rushton turbine produces more shear than required to achieve high mass transfer rates. The importance of the shear and flow characteristics of each impeller type is shown. In slender tank configurations, like fermenters, a combination of one lower Rushton turbine coupled with two or more upper hydrofoils, like A315s, produces the best results for temperature and concentration gradient sensitive mediums. This is a result of a good match of shear and overall flow.

Introduction and Background

Biological processes are some of the most complicated processes confronting engineers today. Many parameters influence the outcome of the reaction, and the interrelations of these parameters are not always known. Some of these parameters must be within very critical ranges. After successful testing on small scale, the scale-up can come out just as predicted. Typically, though, the scale-up always brings some new surprises, often resulting in unwanted lower yields, inseparable byproducts, and DO-crashes (DO= dissolved oxygen).

There are many parameters that influence a fermentation that are solely related to the effect of mixing. Many of these parameters affect biological processes much more than chemical processes. Obvious parameters are position of the impeller or impellers, type of impeller, size of impeller and impeller speed. Not so obvious are parameters that are influenced by mixing, such as heat transfer, pumping capacity and mixing times, flow patterns, shear, and energy dissipation. Other non-mixing parameters, though, can directly affect the performance of the mixer, too, such as viscosity, baffling, presence of cooling coils, placement of the nutrient additions, sparge configuration and gas rate.

Fluid and mechanical shear produced by the movement of the impellers are responsible for the dispersing of the inlet air into fine bubbles. Since oxygen has a low solubility, mass transfer of oxygen from the gas phase to the liquid phase is usually considered to be the rate limiting step in fermentations. Rushton type impellers are known to be high producers of shear. Thus, these radial flow-producing impellers are typically found in fermenters.

Rushton impellers are also known to be poor producers of fluid flow for a given power input. Many fermentations must operate within very tight temperature, pH, and dissolved oxygen constraints. The cooler fluid at the wall of the heat transfer surface (jacket or coils) must be quickly distributed throughout the tank contents to achieve a uniform temperature profile and minimize hot spots. The addition of nutrients and acids or bases to keep pH sensitive broths constant is typically added to the fluid surface. Without adequate blending, these additions can cause zones of high concentrations near the oxygen depleted upper section of the fermenter and low concentrations near the lower oxygen rich zone. Axial flow type impellers are well known for producing high amounts of fluid flow for a given power input. Many of them are also typically poor dispersers of gas.

Depending on the mixing parameters involved, a process can appear to be oxygen mass transfer limited or be blending or flow limited. For each process, biological or chemical, the right blend of flow and shear is needed to achieve optimal results. Too much shear is wasted energy and can be detrimental to the "health" of the microorganisms or the product. Too little flow can cause concentration and temperature gradients, which can cause unwanted side-products, too.

For a given amount of power, all impellers produce a certain amount of shear and flow. Tangential flow type impellers and tanks without baffles produce a high amount of tangential flow and very low shear. Axial flow type impellers produce high flow and low shear. The axial flow produces top to bottom turnover, which reduces the mixing time. Radial flow impellers produce low flow and high shear. Finally, spinning disks produce the most shear and very little flow. With the exception of instantaneous reactions, this type of impeller is not encountered often. The knowledge of a given impeller's flow/shear characteristics is important and neither should be overlooked (see discussion about impeller spectrum).

Probably the most underestimated effect of mixing is scale. All impellers provide more than enough flow on small laboratory and pilot-scales. The possibility of flow limitation is never questioned, and thus high shearing radial flow impellers are unquestionably the most often used impellers. Multiple radial impellers cause staging and further decreases the overall flow efficiencies of radial flow impellers, which is not readily noticeable on small scale. Another non-mixing related parameter is the gas volume. Because of stoichiometry, the relative volumetric flow rate of air (vvm) is usually kept constant upon scale-up. The superficial gas velocity (vsg) exiting the sparge pipe is much lower on small-scale as compared to a larger scale. This can often lead from a well-dispersed pilot scale to a gas-flooded large scale. Gas dispersion is almost always worse in the plant.

Upon scale-up, most of the parameters change. Maximum impeller shear rate will go up for radial flow impellers and go down for "pure" axial flow impellers. Shear sensitive cultures will see poorer results upon scale-up in radial flow. Average impeller shear rates decrease for all types of impellers. This is unfortunate, since this type of shear is responsible for the creating of fine gas bubbles and determining the interfacial area important for mass transfer. Flow relative to tank volume always decreases under geometrical similarity because the flow is proportional to impeller speed and impeller diameter to the third power, whereas the power increases with the third power of the impeller speed and the fifth power of the impeller diameter. This is equivalent to longer blend times, and the formation of concentration and temperature gradients.

Shear rates are very difficult to determine, and are thus often neglected. Since shear is the gradient of velocity in respect to space, very finely spaced velocity measurements must be made in order to determine the effect of shear on gas and solid particles of similar dimensions. In other words, a shear gradient determined over a path length of 1 mm is irrelevant to a particle with a 10-Ám size. Likewise, the shear gradient over a 10-Ám path will have no effect on 1 mm particles.

The laser is an ideal tool, not only for achieving very finely spaced velocity measurements, but because it is non-intrusive, too. The differential of the velocity profile emitting from an impeller leads to the maximum and average impeller shear rates. Particles and substances that have long retention times, like the microorganisms, will pass through the impeller many times. They eventually "feel" the effect of the maximum velocity gradient. Gas bubbles may see the impeller only once, and thus the average impeller shear gradient is more important. The integration of all the velocity profiles over the immediate discharge area of the impeller results in the determination of the primary flow. Integration along a plane in the discharge zone of the impeller until the point of flow reversal, results in the determination of total impeller flow of an impeller. The total flow generated by an impeller is a result of the primary flow and the impeller's entrained flow. The ratio of entrained to direct flow is a function of the impeller's diameter to tank diameter ratio. The laser is thus a very useful tool to determine both the flow and shear characteristics of an impeller. Positioning the laser at one point and tracing the flow fluctuations with time, results in the determination of local energy fluctuations or time related shear gradients. High shear impellers also produce high-energy fluctuations.

A newer technology using a laser is PIV or particle image velocimetry. Instead of analyzing just one point at a time, PIV can look at an entire plane in one moment. This technique illustrates the flow pattern in real time. A frame capturing technique and a particle-tracing algorithm allows for the calculation of all velocity vectors in that plane.

Newest generation computational fluid dynamics (CFD) allows for a very good estimation of the local velocities in a mixing environment. Shear rates can be automatically determined, once the flow field has been calculated.

The laser has been used to study the flow and shear characteristics of the Rushton turbine or RT (to be consistent, Lightnin names this impeller RT) and several axial flow impellers including the Lightnin A315. The A3l5 was developed to be a high flow impeller producing moderate shear gradients capable of dispersing higher amounts of gas than the RT under certain conditions. The lasergrams were also used to determine the optimum position of sparge rings. Since blending almost always becomes an issue on scale-up, axial impellers are expected to be an improvement.

The outstanding question to answer is whether the lower shearing characteristics of the axial flow impellers would cause mass transfer to decrease, too. Excellent gas-liquid mass transfer with A315s has been noticed on industrial scale equipment with highly soluble gases, such as in flue gas desulphurization and steam stripping. Here the A3l5 outmatched the RT at very high gas rates where the RT is easily flooded.

Fermentations are, as mentioned earlier, very complicated. Early industrial experiences with upper axial impellers and a lower RT showed in some cases excellent results and in other cases poorer results. Since all small-scale mixers produce enough flow, the best comparisons came from the industry. Typically, the large-scale fermenters are not fitted with state-of-the-art equipment and so the effect of the mixers is solely judged on yield, the time to achieve that yield and/or, DO-profile. The underlining mechanism for that improvement is not always so clear. Apparently, some of the poorer results were due to poorly designed (in respect to gas handling) axial impellers that were flooded upon scale-up. The design of the A315 mitigated these problems, and the added effect of high blending rates usually improved the process by decreasing the mixing time.

Experimental

Model studies in water/air made it possible to get a better "picture" of the effects of mixing and blending. In equipment similar to fermenters, oxygen mass transfer studies were performed using a modified steady-state sulfite method.

To avoid the often criticized chemical enhancement factor in typical sulfite studies, no catalyst was used, and great care was taken to remove all possible chemical enhancers and consumers. Copper sparge rings were found to enhance the oxygen uptake, where as the slightest drop of oil from the gearbox caused the reaction to nearly cease. Since catalysts were not used, the oxygen concentration in the liquid could not be assumed to be zero, and was monitored. The oxygen concentration was never found to be zero, with values from 0.5 to 1.5 ppm being typical depending on the level of agitation. The sulfite alone was not capable of absorbing the entire amount of oxygen. This "slower" sulfite reaction eliminated the enhancement factor and allowed for a simpler mathematical representation of the chemical process. The dissolved water concentrations were measured with Yellowspring probes and the off gas was funneled into a drying column to remove the partial pressure of water and then directed to an oxygen detector. A mass balance was made and the rate of oxygen transfer was determined by means of mass flow meters and these concentration measurements.

The small-scale 0.46 m diameter tank was made of plexiglas so that the mixing effects could be seen. The tank had four baffles and a flat bottom. Depending on the number of impellers studied (from 1 to 3) the liquid level was adjusted to represent "reasonable" industrially acceptable geometries. In other words, the impellers were not placed unreasonably close to each other, or too close to the bottom or surface. The large 1.17 m tank was made of stainless steel. All impellers and metal internal parts like the sparging devices were made of stainless steel.

Like fermentations, this reaction had to be pH controlled to give consistent and reproducible results, and thus base was added when needed. Toward the beginning of the reaction the addition rate was high, because of the buffering effect of the sulfite-sulfate pair, and toward the end the pH was more sensitive to the addition rate.

Results

As expected, the blend times of single mounted impeller configurations were very similar for the same power. Differences were only in seconds. Also, the mass transfer coefficients, kLa, of the impellers were almost identical. This result indicates that the lower shear gradients of the A3l5 were sufficient to produce similar oxygen transfer rates as the RT. The additional shear of the RT was wasted. At low power levels, the A3l5 was significantly better than the RT (see Publication #44). At lower levels, the flow or pumping capacity of the A315 is much better than the RT and so even on small scale, an apparent improvement was noticed due to blending. At higher power levels, the kLa performance merges and then at even higher power levels, there is an advantage to be seen for the RT. The merging can be a result that the two systems have approached equal mixing times. The further increase in power increases the shear of the RT proportionally more than the A315 and since blending is no longer limiting, the mass transfer limitation begins.

Dual impellers showed similar results. Here two RTs, two A3l5s or one of each were tested. Again, dual A3l5s were best at lower powers or in the flow limited regime. The dual RTs were worst because they were flooded. But as the power was increased, the kLa values merged. Again, this indicates that the RT produces more shear than it needs to accomplish the mass transfer. This time the curves did not cross. Perhaps, the curves would. Further testing is needed at the higher power levels.

Interesting results came from the triple configurations. Based on the blending tests, where 3xRTs actually took 3 times longer to blend compared to 3xA3l5s, the 3xA315 configuration was expected to be the best mass transfer impeller. Again, the A3l5 configuration was tops at low power levels. But soon the 3xRT and the lower RT upper 2xA3l5s configurations surpassed the 3xA315s configuration significantly. It was surprising that the lower RT upper 2xA315s geometry was best.

Some of these trends appear to be easy to explain. kLa was also found to be affected by the impeller to tank diameter ratio (D/T) of the RT, whereas the A3l5 showed basically no effect. The Rl00 maximum was found to be at 0.35. Tests showed that all A315s performed better than RTs when the D/T > 0.5 or D/T < 0.3. The key to understanding this again sits with the flow and shear relationship.

The most interesting fact that came out of these tests were that there was no instance where either impeller outperformed the other by more than 15% with the exception of the mixed triple configuration. This indicates that the performance of the A315 is virtually as good as the RT if designed properly. Retrofitting, i.e. equal torque, impeller speed and power doesn't always result in the optimum performance for the A315. Especially, when the RT already had the optimum D/T. It must be also pointed out, that the sulfite test is not a very sensitive test in respect to blending. These tests showed that practically the same kLa could be achieved with a lower radial shearing impeller.

The fact that all A315 configurations were best at low power levels does indicate that the A315 may be superior to the RT in medium and high viscosity fermentation broths. Here the blending is extremely important. With non-Newtonian broths, the dampening effect of the broth to the flow will cause the difference in blending characteristics to be much greater. The Lightnin A320, which looks like a three-bladed A315, has proven to have the most axial thrust, even in the transitional flow regime of any impeller. Since the A3l5 was designed to have a much higher flooding characteristic, the A315 impeller will probably show better results in the transitional flow regime, too.

Conclusions

The lower RT upper dual A315 configuration proved to be the best configuration when based on mass transfer alone. This is an indication that one type of impeller alone cannot achieve the optimum blend of flow and shear. Since most fermentations add the gas only at the bottom, the majority of the gas is consumed in the lower third oxygen rich zone of the tank. Here shear is more important than flow. The shear of the upper RTs was totally wasted with the 3xRT set-ups and thus caused the kLa comparison at the same power not to be an optimum. Replacing the upper RTs with A315s caused the entire setup to rotate faster for the same torque and power. This caused higher local shear rates in the bottom radial impeller section. The upper A315s, because they weren't flooded, pumped the oxygen starved media to the oxygen rich bottom and enhanced the blending and thus increased kLa. Apparently, in the 3xA315 configuration, the lower A315 impeller could not achieve enough shear. It is also possible that the combined effect of three A315s, which are "feeding" each other, produced so much more axial flow that the gas bubbles were just propelled out of the tank without enough retention.

From this study, the importance of shear and flow has been demonstrated for fermenters and other chemical and biological reactors.

For highly soluble gasses, gas-handling capabilities are extremely important and the A3l5 has been shown to be unsurpassed.

The correct mixer solution for gases of low solubility is more difficult. Other factors begin to be important, too, especially in large scale. Processes that are suspect of being highly sensitive to temperature and concentration gradients will benefit from A3l5s on large scale, where on small scale no difference will be noticed. For those fermentations that are less sensitive to gradients, good gas liquid mass transfer is very important. For fermenters and slender chemical reactors the optimum appears to be a high shearing RT in the vicinity of the gas inlet with upper A3l5s to provide unflooded axial flow. Even though the oxygen may be completely consumed, 79% of the original bubble is still nitrogen. Axial flow impellers not having a high solidity ratio will tend to flood and the pumping and mixing will cease.

For those genetically engineered animal and cell cultures, which yield more product in the same volume, 3xA315s and other high solidity hydrofoils may be the best choice. kLa may be 15% less than the equivalent RT selection, but the lower impeller dictates the flooding characteristic of the system. The A3l5 can distribute four times the gas at the same power as compared to the RT at low power levels and still blend better. Since this seems to be the trend, the quantitative effect of viscosity on kLa is important. Several studies have shown that kLa is not affected by viscosity until the viscosity is above 10-25 cPs (mPas). As viscosity increases beyond that, kLa has been shown to decrease rapidly. The exponent on apparent viscosity is -0.7. This means that at 600 cPs, the kLa is an order of magnitude less than at 1 cPs. As viscosity increases, the process regime will be in the transitional to laminar regime. On small-scale the flow regime may be in the laminar regime. There mixing will be most critical and since all impellers pump in a radial direction there will be little difference between radial and axial impellers. In the transitional regime an all-axial flow impeller system should give the best results.



Contact webmaster with questions or comments about this web site.
Copyright 2003 - 2013 Post Mixing Optimizations and Solutions, LLC
Last modified: February, 2013
Web Design by Sandy Gifford, Graphic Design, Web Development, and Illustration