Comparison Between Experimental Data and Computational Fluid Dynamics in Mixing Applications

Authors:?

Damien KERVENNIC and Léna?g SAVARY; Merck?

Sid-Ali OULD KHAOUA and Karim LOUESLATI; EUROCFD

INTRODUCTION
        

In biomanufacturing processes, mixing is ubiquitous.?Examples include buffer and media preparation, stirred?tank bioreactor/fermentor operation, virus inactivation,?DNA digestion, final formulation and beyond. Mixing trials?are sometimes necessary to assess if a technical solution?is adapted for a given process step or to optimize process?parameters. However, trials may not be feasible due to?various reasons:?

? Product is unavailable in suitable quantities

? Product is too expensive

? Safety issues (e.g., Antibody Drug Conjugates, allergenic)

? Availability of equipment?

Computational fluid dynamics (CFD) is a powerful?simulation tool that can be applied to a wide range?of research and engineering problems in many fields,?including aerodynamics and aerospace, weather?simulation, environmental engineering, biological?engineering, fluid flows, and heat transfer. Typically, its?use leads to an improved understanding of the process flow upon which the performance of a product relies,?as well as a reduction in the amount of empirical testing?required.1 Therefore, CFD could be a complimentary or? alternative solution to physical mixing trials. It could also?be an initial step for validation prior to a confirmation??run. To evaluate these options, MERCK collaborated with?EUROCFD on a specific study to compare empirical data?with CFD results.?

About EUROCFD

EUROCFD is a major player of French engineering?companies dedicated to numerical simulation for industry.?EUROCFD deploys its skills in many industrial sectors?such as aeronautics, nuclear, oil & gas, transport, and?biomanufacturing. With best-in-class resources, EUROCFD?conducts research and development programs and?industrial product improvements.

Background

Mixing trials are performed on a regular basis, at MERCK M Lab? collaboration centers. In this particular case, the?objective of the trial was to determine the mixing times?during the final formulation step for a mAb process. As??a mock fluid for the mAb solution, a sucrose solution?was used and a NaCl tracer was introduced to determine?the mixing times at different fluid volumes and impeller?speeds. The process conditions of this study were shared?with EUROCFD to perform a CFD evaluation.

Materials and Methods        

Experimental mixing trials

The Mobius? MIX single-use systems (figure 1) are?particularly suited for final formulation applications as? they provide gentle and low shear mixing environments.?Mobius? MIX 50, 100, and 200 systems were used for?this study. Sucrose and Tween? 20 were dissolved in?reverse osmosis water to obtain final solutions of different?viscosities. Separately, a 4 M NaCl solution was prepared?to serve as a tracer. Following a Design of Experiments?(DoE) (table 1), the Mobius? MIX systems were filled at?specified volumes with the sucrose solutions.?

A conductivity probe was installed to measure the?conductivity of the solution at the surface of the liquid?(figure 2). The impeller was switched on at the given?speed for the test. Once the mixing steady state was?achieved, the tracer was added (1 mL of tracer per?liter of solution) and the conductivity of the solution?was monitored. Once the conductivity was stable, a?sample was taken from the bottom of the Mobius? MIX?system to confirm homogeneity of the full bulk.

Figure 2: Experimental setup in the Mobius? MIX 100 system

The conductivity results were normalized based on the?average conductivity reached at steady state. T95 was?calculated as the mixing time corresponding to the first??time for which all the following conductivity values are??within the 95–105% range of the conductivity increment.?

In the context of the comparison with CFD, three?conditions out of approximately fifty combinations tested?experimentally were selected. The conditions used for?the CFD simulations are summarized in table 2.?

Computational Fluid Dynamics Analysis
        

STAR-CCM+ 2020.1 software was used by EUROCFD?to perform the CFD analysis. The working volume was?divided in a mesh of approximately 6 million of cells and the?mesh further refined near the impeller and conductivity?probe. Mixing in the tank was evaluated by injecting?a virtual tracer (passive scalar) when the flow was?stabilized. The evolution of the concentration of the tracer?was monitored by 26 virtual sampling probes (figure 3)?over 150 to 200 seconds of real time.

RESULTS
        

Physical mixing trials: The results of the experimental trials for the conditions?selected for the CFD simulations are presented in?table 3 and figure 4. At 54 L, a single experiment was?performed at both speeds while the experiment was?performed in duplicate at 64 L (T95 was averaged for?the experiment at 64 L).

Computational Fluid Dynamics Results
        

For the CFD study, mixing time was evaluated based?on the evolution of the tracer concentration in the bulk?solution. T95 was determined when the concentration?measured at the 26 virtual sampling probes was within?± 5% from the target concentration (figures 5 and 6).

The comparison showed good correlation between CFD?and experimental results with a mean difference of 4.9%?(table 4). The model used was especially accurate for?experiments at 150 rpm with a difference below 2%. CFD not only allows for the determination of mixing?times but is also a powerful tool to study fluid behavior.?For instance, the CFD confirmed that the Mobius? MIX?systems provide gentle mixing by calculating the volume?corresponding to dimensionless shear rate (γ/N) ≥ 25,?50, 75 and 100. The results for the three conditions?are summarized in table 5 and show that the volumes?submitted to shear forces are very limited to the zone?around the impeller (figure 7).

The simulations also allowed us to see the velocity iso-contours on different cutting planes (figure 8) or the high?velocity pumping area around the impeller (figure 9).

Finally, CFD allows determination of impeller?characteristics such as the resistive torque that the?fluid exerts upon the impeller, the absorbed power, the?impeller power number, the pumping and circulating?flow rates, and the pumping and circulating numbers. In?the case of the impeller power number, the difference?between the value measured experimentally and the?value determined by CFD was less than 7%.?

DISCUSSION
        

This collaboration confirmed the benefits CFD analysis?could provide for mixing applications. Experimental?conditions were used as a basis for simulations and?the CFD demonstrated that there is a good correlation?between the experimental data and numerical analysis.??In addition to mixing times, CFD could offer a great?panel of fluid behavior visualizations moving from?velocity fields to shear rate or path-lines. It could also?provide impeller characteristics such as power and?pumping numbers. Based on the overall study, CFD?demonstrated that it could be a powerful tool in mixing?applications either as an alternative to actual mixing?trials or as a complementary solution. This study also?highlights the importance of implementing safety factors?in all the steps of a biomanufacturing process including?adding extra filtration areas or additional mixing times??to consider process variability.

REFERENCES        

1 How to - Ensure that CFD for Industrial Applications is?“Fit for Purpose”. NAFEMS, 2010.?

For additional information,??please visit www.MerckMillipore.com??or www.eurocfd.com

06/01/2022