4 State of the Art and Recent Developments Associated with Common Mixing Tasks

4.1 Single-Phase Fluids

Figure 5. Axial flow impellers. (a) Chemineer HE3 (Po = ~0.3); (b) Light- nin’ A315 (Po = ~0.85) (from [12] with permission).

Theoretical arguments based on turbulence theory also support the form of Eq. (11) [20], which, with respect to ¯eT, implies all impellers give the same qm at the same scale, and that with increasing scale, qm increases. This increase is a major scale-up issue since the characteristic times of chemi- cal [21] and biological processes [22] are scale independent. Thus, at the bench scale where the mixing time is short, the STR can be considered well mixed. However, at the com- mercial scale, the well-mixed STR assumption is no longer valid and the process performance may deteriorate [21, 22]. In tanks of AR > 1 with multiple impellers, agitator choice becomes important, especially at large scale because of the increase in mixing time. As the aspect ratio increases,    the

mixing time increases and is approximately given by [23]

This task is usually quantified by the mixing time qm, which is the time required after the addition of a tracer for it to be homogeneously distributed throughout the STR. It has been investigated since at least 1953 [18] and in that study, as in many since, an electrolyte solution was added and its con- centration was followed by conductivity probes until it was

In general, multiple axial flow impellers, whether up- or down-pumping, have mixing times approximately half those found with radial flow impellers, the latter giving strong compartmentalization around each impeller while the former give good top-to-bottom motion [12]. However, with multiple impellers, the range of configurations possible

is endless and so are the precise mixing times. However, a new and very promising approach for predicting qm for any configuration in the turbulent flow regime has been recently developed by Liu based on the mean age method via com- putational fluid dynamics (CFD) [24]. This approach itself builds on another one of the outstanding contributions listed in [11], namely the classic paper of Danckwerts on residence time distribution [25].

As the viscosity increases (Re falls to < ~6000, i. e., well above Re leading to an increase in Po), the mixing time increases roughly in proportion to viscosity [23], and this trend is magnified if the fluid is also rheologically complex. In general, ideally an AR = ~1 should be used with large dual impellers, e.g., Intermigs [26]  with  D/T = ~0.65, Po = ~0.7 for the pair, are particularly suitable for such conditions. Recently, large D/T gate-shaped impellers (an example is shown in Fig. 6), which have been used for many years in Japan, have been studied in some depth [27, 28] and compared with other large D/T impellers. The studies suggest that the Sumitomo Maxblend impeller is particular- ly energy efficient with respect to blending in the low Re transition regime with baffles; and without them in the upper part of the laminar regime. However, it also draws a high torque and combined with its large and complex shape means that it has a high initial cost [12].

For laminar blending, baffles are not used, and impellers should have a close clearance from the wall. A wide variety of impellers have been tried and though still used, it was

shown in 1967 [29] that the anchor type is very poor because it produces little vertical motion. The same paper showed that strong vertical flow is obtained with helical rib- bon impeller and it is now accepted as the most energy-effi- cient for many applications [23] though very expensive. A cheaper, modern alternative, though giving a little longer mixing time, is the Paravisc [26]. In the laminar region, for this stirrer and all energy-efficient geometries [23, 26, 29],