Abstract
Overlapping sprays are used in engineering applications (e.g., spray drying) in order to enhance the agglomeration
processes. Within this perspective, the Spray3D CFD code [1,2] is used to predict three overlapping sprays issued
inside a simplified (cylinder-on-cone) geometry that represents a spray drying chamber of Anhydro A/S, one of the
8 consortium partners that participated in a EU multi-national project on spray drying processes [3]. Induced hot air
at around 150 C and flow rate of 2,633m3/s interacts with feed of solid in suspension at 40 C and leaves the
chamber from two top outlets located at 7 and 12 o'clock respectively (positions in the circumferential direction).
This wall-bounded turbulent flow case is simulated with the use of the standard - turbulence model [4]. The
computational grid is generated using a cylindrical coordinate system in which the cone section is represented by
patching together stepped blocks of unit grid line lengths in the vertical direction. The Spray3D CFD code s in-built
evaporation model for homogeneous fluids is compared with a newly incorporated drying model for fluids with
dissolved solids, which was developed by the EDECAD partners TU-Darmstadt and NIZO [5]. The numerical
calculations are extended from experiments developed by the EDECAD partner University of Bremen [3], at
ambient temperature, which leaves the need for rather difficult and onerous experiments in hotter environments.
*
Corresponding author: paul.watkins@manchester.ac.uk
20th Annual Conference on Liquid Atomization and Spray Systems (Europe) – Orléans (France), 5th-7th September 2005
that immediate problems arose with this gridding inhomogeneous fluids by the partner University of
procedure when representing off-axis inlets and Bremen [11], which required time consuming extention
outlets, i.e., injectors and gas outlets, the effects of into the EDECAD Project purposes so that the initial
which are discussed later. spray data could be adequately processed from the
generated LSS, DSM and PDA data.
occur away from the axis of the flow domain. where, at each new time levels n, the Newton-Raphson
The main reason thought to cause the algorithm suffices to solve for the gas temperature T n ,
disagreements is the spray cone angle, which had to and the temperatures and diameters Td,k and Dd,k of
n n
be limited to 40 in the computations, while in the
experiments was ~62 . The numerical solution using droplet parcels d,k respectively [13]. In Equations (1) to
angles greater than 40 led to code crashing in 30ms (3), md,k denotes the mass of droplet parcels at old time
0
which different boundaries conditions were In Figure 11, the droplets were assumed to contain
implemented (e.g., the 100 C fixed temperature at the 30% solids contents, which caused dry particles to remain
top air outlet, which was assigned on the basis of after the complete removal of moisture. Noting that only
typical spray drying experiments [6,7], and the a few of the injected particles were depleted, the number
highest temperatures residing near the chamber axis of residing particles within the computational domain
where the highest gas tumble motion is present). The became excessive, hence leading to considerable CPU
gas was allowed to flow for 2s real time before the time (35 CPU hours for the ~360ms real time with each
liquid/feed was sprayed and the time counter re- of the 3 nozzles injecting 8 particles per time step of ~10-
initialised for the evaluation of multi-phase flow. 5
s). Unlike in Figure 10 in which the spray hangs at early
sections of the cone region, the presence of solids
contents, as shown in Figure 11, allowed the
droplets/particles to progress further towards the products
collection zone
(a)
(b)