Pulp fibre suspensions display non-Newtonian rheology,
including a yield stress. Under certain mixer operating
conditions this creates caverns (regions of active mixing)
around the impellers with the cavern size affecting the
extent and quality of mixing attained. Due to the opacity
of pulp suspensions it is not possible to measure cavern
size with direct optical techniques, like photography.
Consequently two non-invasive techniques suitable for use
in opaque media were evaluated for determining the
cavern dimensions: electrical resistance tomography
(ERT) and ultrasonic Doppler velocimetry (UDV). The
agitation of several pulp suspensions in a 38 cm diameter
cylindrical vessel was studied using these methods over a
range of operating conditions. ERT is a non-invasive
technique that images differences in conductivity between
regions in the mixer using voltage measurements made at
the vessel periphery. Cavern measurement by ERT is very
rapid (data are collected within a few seconds) but it
suffers from poor spatial resolution (approximately 5 to
10% of the vessel diameter – from 1.9 to 3.8 cm in our
case). Two methods were evaluated for creating the
conductive environment imaged by ERT – the injection of
saline solution or the addition of small metallic tracer
particles to the region surrounding the impeller. UDV was
used to determine the cavern boundary by measuring the
locations at which suspension velocity fell to zero for
multiple linear paths through the vessel. While UDV
provided better spatial resolution of the cavern than ERT
(about 2 mm), multiple measurements (and consequently
significant time) were needed to build up the profile of the
cavern boundary.
Cavern size as a function of impeller rotation speed is
reported for a range of pulp suspension mixing conditions
(hardwood and softwood pulps, suspension mass
concentrations from 2 to 4%, two impeller offsets from the
wall, and two suspension height-to-chest diameter ratios)
in the 38-cm diameter cylindrical chest. A scaled version
of a commercially available axial flow impeller designed
for use in pulp suspension agitation (the Maxflo,
Chemineer Inc.) was used in the standard side-entering
configuration used for pulp stock chests. Measured cavern
diameters were compared against the axial force model
developed by Ammaullah et al. (1998) for predicting
cavern diameters in non-Newtonian fluids. The
discrepancy between the experimental data and model
predictions were fairly large, although they decreased with
increasing yield stress Reynolds number. The discrepancy
was attributed to the proximity of the impeller to the
vessel walls in the side-entering configuration studied. An
alternative correlation is presented for predicting the
cavern volume in pulp suspensions in this mixing
configuration based on the suspension yield stress

increasing yield stress Reynolds number. The discrepancy
was attributed to the proximity of the impeller to the
vessel walls in the side-entering configuration studied. An
alternative correlation is presented for predicting the
cavern volume in pulp suspensions in this mixing
configuration based on the suspension yield stress

Quantifying the flow and residence time distribution (RTD) in a reactor is critical for predicting reactor performance measures such as yield and selectivity. This work investigated a continuous industrial scale reactor (diameter 11 ft, height 37 ft) employing a 3-stage agitation system with a plunging jet inflow. In operation, the agitation and liquid level are adjusted based on throughput and the specific product being produced. The objective of this work was to quantify RTD under the different operating conditions of liquid level, throughput and agitation. Flow in the reactor was modeled with computational fluid dynamics (CFD). A combination of different modeling approaches was used to better facilitate the CFD simulations. The Volume of Fluid (VOF) method was used to model the plunging jet (gas-liquid) two-phase flow with unsteady-state simulations, while the Multiple Reference Frames (MRF) model was used to model the stirred tank with steady-state simulations. The two modeling approaches were interfaced by using the plunging jet simulation result as the input boundary condition for the reactor flow simulation. The effect of gas entrainment from the plunging jet impingement was accounted for by using the (dampened) velocity profile of the impinging jet. The RTD was obtained from stochastic particle tracking which tracks trajectories and residence times of massless tracers in the reactor. The Random Walk Model was used for dispersion of tracers due to turbulent eddies. A large number of tracers (>10,000) was needed to account for the random effects of turbulence and ensure statistically stable results. The Time Scale Constant in the Radom Walk Model was adjusted to accommodate significant turbulence level differences in discreet regions of the tank. The CFD-predicted flow pattern compared well with lab-scale experiments. The predicted mean RTD was consistent with the bulk reactor turn-over time