Session 2 - Theme: Experimental Techniques & Results (2)

Laminar mixing is critically important in industries dealing with high‑viscosity materials, food, and cosmetics, yet conventional anchor impellers operating at very low Reynolds numbers inevitably generate distinct poorly mixed regions between the blades and the central shaft. To address this long-standing limitation, we examined whether rotational reciprocation could induce chaotic mixing and thereby enhance mixing performance. Chaotic mixing was successfully achieved when the Reynolds number was not excessively low, although isolated mixed regions still appeared, with their positions and shapes varying dynamically with the impeller motion. To eliminate these regions, we further investigated asymmetric anchor geometries and found that circumferential asymmetry, achieved by unequally spacing the blade attachment angles, was effective in suppressing poorly mixed regions. CFD analysis revealed that excessively narrow inter‑blade angle produced stagnation near the shaft, and that overly large reciprocation amplitudes weakened chaotic mixing. Finally, we demonstrated that a three‑blade anchor impeller with a 120‑degree inter‑blade angle, reciprocated at an amplitude of 90 degrees, could completely eliminate poor mixing while maintaining unsteady chaotic flow. Chaotic mixing induced by rotational reciprocation has rarely been explored in previous anchor‑impeller studies, and the use of symmetry breaking to improve mixing performance provides a new and practical design guideline.

In various mixing operations in tanks, achieving good mixing without an impeller is sometimes necessary. These include cases, such as metallurgical applications where the liquid is a molten metal, and cases where an inhibitor has to be rapidly blended in to avoid a runaway reaction due to impeller failure. In such cases, gas is often used to induce bulk flow and mix the contents of the tank. The present study investigated the use of gas to induce mixing in tanks without impellers by measuring the liquid phase mixing time using a decolourisation technique. In all cases the gas was introduced using a vertical dip pipe from the top of the tank. Experiments were conducted at three different scales, using tanks with diameters of 0.45, 0.61 and 1 m (roughly 67, 160, and 730 litres, respectively), all un-baffled and with torispherical bases. The effects of the off-bottom clearance and eccentricity of the dip pipe were quantified, as well as that of the gas flow rate, using air and water as test fluids. The effect of liquid height in the tank was also examined, and preliminary work has been carried out on the effect of viscosity using glucose solutions. It has been found that gas injection is an effective method for mixing liquids. The mixing time reduces as the power input due to the buoyancy of the gas increases, though scale, geometric parameters, and liquid viscosity can have a significant effect. Different scaling parameters and correlation approaches have been assessed.

Power input for two phase gas-liquid processes in stirred tanks is usually predicted using empirical correlations with no well-adopted mechanistic models. Accurate prediction of gassed power is critical to ensuring correct power input for industrial processes such as gas-liquid mass transfer. A new mechanistic model is proposed for the ratio of gassed to ungassed power (PG/PU) based on the dimensionless numbers Ae (Aeration Number) and Fr (Froude Number). The model uses a local force balance to establish maximum gas cavity size, a local mass balance to determine equilibrium gas cavity size at a given gas flow rate and proposes a direct theoretical relationship between the gas cavity size and PG/PU. To evaluate the predictive accuracy of the model, power measurements were performed in an air/water system for Rushton, Smith, and R135 Gasfoil impellers in cylindrical vessels across three scales (24”, 34”, 48”). The model was then fitted to the data with impeller specific coefficients, covering Aeration numbers from 0 to 0.7 and Froude numbers from 0.1 to 3.0. The model correctly captures key observed behaviours across all three scales and is shown to be more accurate than available empirical models. Detailed results and correlation for the Rushton impeller are presented. The mechanistic basis improves robustness and furthers understanding of the mechanisms behind gassed power input for different impellers. By improving prediction of gassed power, the model allows for better and more robust industrial designs, ensuring the desired power input is delivered to processes that rely on it.

The current design and optimization of mixing processes for non-Newtonian fluids in large industrial-scale vessels rely mainly on CFD simulations or down-scaled experiments. Available experimental techniques are mostly limited to point measurements at fixed positions inside the reactor. Contrary, advanced techniques that provide distributed data, such as PIV/PTV, are hardly adjustable in industrial reactors. Recently, Lagrangian Sensor Particles (LSP) were proposed and tested in industrial scale reactors. These particles follow the flow and log data for pressure, acceleration and spin rate. In this work, the mixing of shear-thinning CMC-solution of different concentrations in a 150 L stirred tank reactor, equipped with a 3-blade PBT running between 150 and 300 rpm is studied numerically and experimentally. Eulerian based CFD simulations are validated with LSP measurements. The LSP data are evaluated by averaging the counts of axial upwards/downwards velocity components, resulting in a mean averaged axial velocity profile along reactor axis. In comparison, in the Euler based simulation the axial velocity components are evaluated by a weighting function. The results indicate a root mean squared error (RSME) between 0.02 m/s and 0.05 m/s with an average velocity between 0.13 m/s and 0.25 m/s. Additionally, the evaluated mean circulation times from simulations are in good accordance with experimental data. Increasing deviations between simulations and measurements at higher CMC concentration (i.e. higher viscosity), reveal a shift to laminar-dominated flow, which cannot be captured by the current model assumptions of the CFD setup.

Efficient removal of volatile components from viscous fluids under vacuum is a critical operation in large‑scale agitated vessels. In batch reactors, this operation proceeds through boiling, transition, and surface‑evaporation stages, with the last being governed by gas–liquid mass transfer at the free surface. This study quantifies mass transfer under surface‑evaporation conditions in large stirred tanks, where the removal of volatile components becomes extremely time‑consuming. A practical measurement approach was established by analyzing the dissolution of a volatile species from the gas phase into the liquid, monitored via transient gas‑phase pressure. The volumetric mass transfer coefficient (kLa) was obtained directly from pressure‑decay profiles. To enhance transfer, a gate‑type impeller was installed so that its blades intersected the free surface, promoting surface renewal and interfacial turbulence. Performance was evaluated using kLa and the power consumption per unit liquid volume. Scale‑up behavior was examined using vessels with diameters from 0.3 to 2.2 m. Importantly, by formulating a dimensionless correlation that explicitly accounts for the increase in gas–liquid interfacial area, we demonstrate a predictive method for scale‑up that consistently captures observed trends across the examined scales. These findings provide a practical basis for the design and scale‑up of vacuum‑operated mixing systems handling viscous fluids.

Continuous antisolvent precipitation has emerged as a scalable method for producing lipid nanoparticles (LNPs), which is known to be a micro-mixing limited process. Several microfluidic devices and meso-scale devices have been developed for antisolvent precipitation. The device design and operating parameters influence local micro-mixing and thereby affect the quality attributes of LNPs – particle size, polydispersity and encapsulation efficiency. In the present work, we demonstrate the use of coiled pinched tube and vortex-based cavitation device operated in a loop configuration for continuous antisolvent precipitation of LNPs. The results are compared with once through operation of commercial microfluidic devices and coiled pinched tube. In a once-through configuration, micro-mixing is tightly coupled with capacity. In loop configuration, the micro-mixing is controlled by flow rate through recirculation loop and therefore independent of net flow rates of solvent and antisolvent streams (capacity). This offers flexibility of adjusting the throughput as per the requirements without compromising mixing efficiency. Furthermore, the use of loop configuration helps in highly concentrated LNP suspensions (up to 50 mg/mL) without jeopardising the limits on size and polydispersity which is not possible in conventional once-through configuration. These experimental results are modelled with two and three environment engulfment models to simulate interactions of micro-mixing and kinetics of LNP formation in the aforementioned devices. The presented approach of loop configuration offers effective pathway to achieve desired size and productivity of LNPs, over a broad range of feed lipid concentrations.