You are on page 1of 41

Process Intensifier:

Optimization Using CFD


Part 1
Paper 362c

Pete Csiszar, Black & Baird Ltd., North Vancouver, B.C.


Keith Johnson, Independent Consultant, North Canton, Oh
Post Mixing Optimization and Solutions, Pittsford, NY
’03 AIChE Annual Meeting
Nov 16-21, San Francisco
Post Mixing
Optimization and Solutions
Process Intensification
 High P/V, high shear, small volume, small residence time
 Applications
 High Speed Dispersion of Bentonite
 Ex-situ Bioremediation of Organics
 Rapid Mixing of Water Treatment Polymers
 Preparation of Coatings
 Beverage Industry
 Flotation
 Chemical Extraction
 Series-parallel Reactions
 Oxidation Processes
 Emulsification Applications
 Dry Material Wetting
 Chemical Neutralization
 Mixing of High Viscosity Shear Thinning Fluids
 High P/V, high shear, small volume, small residence time

Post Mixing
Optimization and Solutions
Introduction
Internet Search
 Lightnin Line-Blender
 Radial and Axial impeller designs
 Hayward Gordon In-line Mixer
 Radial and Axial impeller designs
 No systematic study reported on them
 Use CFD to understand and optimize these
pipe mixers

Post Mixing
Optimization and Solutions
Introduction
CFD confirmation using standard mixing
configurations, T=12.5” (317.5 mm)

RP4 radial impeller PBT axial impeller


5” RP4 D/T=0.4 5” 3PBT30 D/T=0.4

Post Mixing
Optimization and Solutions
Experimental Design
Studied 4 Dynamic Pipe Mixers
 Did not consult with the vendors. Data is taken
directly from their respective web sites

LTR HGR LTA HGA


2x 5” RP4 2x 5” RP4 2x 3.5” 3PBT30 2x 5” 3PBT30

Post Mixing
Optimization and Solutions
Experimental Design
All units were studied in a nominal
schedule 40 10-inch pipe (254 mm)
 DO=5 1/8” (130 mm) for LTR and HGR
Q = 1100 GPM (250 m3/hr) – 10” pipe
Q = 650 GPM (148 m3/hr) – 8” pipe
N = 1760 RPM (motor speed)

Post Mixing
Optimization and Solutions
Experimental Design
Post Mixing
Optimization and Solutions
CFD Background
ACUSOLVE GLS-FE
 Rigorous stability and convergence proofs
 Local / Global Conservation operators

 High Performance

 Accuracy - Advective / Diffusive operators

Post Mixing
Optimization and Solutions
CFD Background
 w u

i ,t  u j ui , j  bi   wi ,i p  wi , j ij d

  q,i ui d

   w,t  u j wi , j  qi  m u,t  uk ui ,k  p,i   ki ,k  bi d e


e e

   wi ,i cu j , j d e GLS Terms


e e

  wi  pni   ji n j   qi u j n j d

Minimize error of approximating functions M = O ( h / |V| ) Advective


Hyperbolic/Parabolic Automatic: M = O ( h2 /  ) Diffusive
Stability and Convergence Proven

Post Mixing
Optimization and Solutions Galerkin / Least-Squares
Reynolds number of 40,000
7,200 brick elements; 14,822 nodes
Spalart-Allmaras turbulence model
Advection / Diffusion “continuously” varying

Post Mixing Backward Facing Step Problem


Optimization and Solutions (Advection / Diffusion Example)
Even for this coarse mesh
 Able to predict the two
smaller eddies near the
recirculation corner
 Smallest eddy captured
within a radius of 3-
elements
Predicted reattachment length
= 7.05 (step height)
 Experimental results =
7±0.1

Post Mixing Backward Facing Step Problem


Optimization and Solutions (Advection / Diffusion Accuracy)
These models tended to
converge in the range of
20 to 30 nonlinear
iterations, to a
normalized residual
tolerance of less than
1.0 E-3.
Runs on a 1.8 GHz
laptop computer with
512 MB of memory in
roughly 2 hours.
Runs on a parallel
configuration of two 2.0
GHz PCs with 2.0 GB
memory each, and the
solutions required only
about 30 minutes each

Post Mixing
Optimization and Solutions
Results: CFD Mesh
Lightnin Hayward Gordon

Radials

Axials

Post Mixing
Optimization and Solutions
CFD Solid Shapes
Reduce Assumptions / Approximations
 Eliminate local entry flow assumptions for mixer inlet /
outlet - used long entry exit
 Model size (DOF) not a major issue
 Accurately solves forward / backward facing step
problems
Geometry Idealized
 Sufficient Fluid Mechanics Performance Equivalency
 Eliminates Vendor Conflict / Propriety
ICEM/CFD autohexa extensions for geometry/mesh

Post Mixing
Optimization and Solutions
CFD Modeling Considerations
Validation / Confirmation Approach Defined
 Standard tank configurations run to assess power and
flow characteristics independently with respect to
Industry Data
 Discretization sensitivity considered
General Flow Solution - Defined - (No Turbulence)
 Discretization dependent
 Captures flow separations / eddys
 May produce stable macro / mezzo flow oscillations
 Lower bound power / torque

Post Mixing
Optimization and Solutions CFD Analysis Approach
Turbulence Considerations / Concepts Considered
 Philosophy - “unresolved” eddy diffusion / dissipation /
production
 Intended for “micro” scale turbulence
 Turbulence introduced becomes upper bound to power /
torque
Discrete particle tracking - Turbulent
 Residence Time Statistics
 Mixing Assessments
 Proprietary algorithms based on Eddy Viscosity

Post Mixing
Optimization and Solutions CFD Analysis Approach (Cont)
Power numbers
 RP4, h/D=0.2
 N=360 RPM
 P/V = 5 Hp/1000 gallons (1 kW/m3)

 Z/T = 1, 4 standard, wb/T = 0.1

 Np(CFD) = 2.985

 Np(Lightnin) = 3.4

 Oldshue Proximity Factor = 0.87, Np = 2.958

 CFD Proximity Factor = 0.878

 Conclusion: Oldshue was right!

Post Mixing
Optimization and Solutions
Results: Power Number
Power numbers
 3PBT30, h/D=0.25
 Np(CFD) = 0.55 OB/D = same as HGA
 Np(CFD) = 0.57 OB/D = same as LTA

 PF=1.044: Agrees with Oldshue, again!

 Np(4PBT45, h/D=0.2) = 1.27

 Nagata: sin(angle)1.2 Np(4PBT30, h/D=0.2) = 0.63

 Shaw: Np(4PBT30, h/D=0.2)=0.58

 Nagata: 77.5% of a 4-bladed impeller


 Np(3PBT30 h/D=0.2) = 0.45-0.48
 Nagata: h/D = 0.2 to 0.25 = an increase of 21%
 Np(3PBT30 h/D=0.25) = 0.54-0.58

 Conclusion: Nagata was right!

Post Mixing
Optimization and Solutions
Results: Power Number
Post Mixing
Optimization and Solutions
Results: Power Number
These small units can agitate up to 1.584
Million Gallons (6 Million Liters) per day (at
1100 GPM (250 m3/hr))

Post Mixing
Optimization and Solutions
Results: Power
85  P/V  715 Hp/1000 Gallons
17  P/V  143 kW/m3
Post Mixing
Optimization and Solutions
Results: P/V
Rule-of-thumb: Impeller generated flow should be at
least 3 times the pipe throughput.
Not one of these devices complies.
Even the LTA appears to be doing some mixing at 650
GPM, which has R = 28% or about 1/4th the pipe flow
rate.
LTA seems to have lost its mixing ability at 1100 GPM.
Perhaps the rule-of-thumb for Process Intensifiers is
that impeller generated flow should be at least 1/4th
the pipe throughput.

Post Mixing
Optimization and Solutions
Results: Impeller Flow to Throughput
Default max-min pressure fields

Post Mixing
Optimization and Solutions
Results: Pressure Drop
Common scale pressure fields

Post Mixing
Optimization and Solutions
Results: Pressure Drop Normalized
Post Mixing
Optimization and Solutions
Results: Velocity Vectors
Post Mixing
Optimization and Solutions
Results: Velocity Vectors
Post Mixing
Optimization and Solutions
Results: Velocity Vectors
Post Mixing
Optimization and Solutions
Results: Velocity Vectors
Post Mixing
Optimization and Solutions
Results: Velocity Distribution
Post Mixing
Optimization and Solutions
Results: Flow Visualization
Post Mixing
Optimization and Solutions
Results: Flow Visualization
LTA:
650
GPM

Post Mixing
Optimization and Solutions
Results: Tracer Study
LTA:
1100
GPM

Post Mixing
Optimization and Solutions
Results: Tracer Study
LTR:
1100
GPM

HGA:
1100
GPM

Post Mixing
Optimization and Solutions
Results: Tracer Study
Post Mixing
Optimization and Solutions
Results: Residence Time Distribution
Post Mixing
Optimization and Solutions
Results: Residence Time Distribution
LTA: 1100 GPM
 Single Input, 1750 RPM
 Single Input, 0 RPM
 Multiple Inputs, 1750 RPM

Post Mixing
Optimization and Solutions
Results: Residence Time Distribution
Post Mixing Results: Comparison to Non-
Optimization and Solutions Newtonian Fluid
This report demonstrates the versatility of using CFD to
model and understand a complex mixing device such as the
Process Intensifier.
Previous use of CFD often meant very long computing time
and it was often quicker to do the experiment. Not any more.
ACUSOLVE was successfully able to determine the power
number of the impellers within 1% of reported values without
the use of fudge factors on a repeatable basis.
 Must be right if it says that Oldshue and Nagata were right!
This demonstrates that the ACUSOLVE CFD code
formulation and its adherence to fundamental physics are
extensible to handle the arbitrary geometric structures and
flow conditions of inline mixers.
Solutions consistent with general fundamental
understandings of these mixer classes. However, past
conventional wisdom concerning assumed internal details,
clearly challenged by detailed CFD results.
Post Mixing
Optimization and Solutions
Conclusions
Four configurations studied, yielding insights for mixing improvements.
For example, tracer inlet location sensitivity, impeller locations, pumping
direction, size, speed.
All examples demonstrated under sized impeller capacity for specified
flow. Part 2 will talk about impeller optimization for Process Intensifiers.
Specific optimizations are clearly a function of application, fluid rheology,
and mixing needs.
Provides a substantial platform for further wide ranging parameter study
for specific application optimization.

Post Mixing
Optimization and Solutions
www.postmixing.com
Evidence of the speed and
accuracy of Acusolve CFD
 Paper given last night from 5:27 PM to 6:00
PM
 Computational time = 90 minutes (Laptop)
 A Novel Mixing Technology Provides Benefits
in Alumina Precipitation, Ian C. Shepherd*,
Clive Grainger, CSIRO Australia
 T = 14 m, Z = 40 m, conical bottom, V  6158
m3
 Upper Oversized RT
 D/T=0.30, w/D=0.333, h/D=0.29
 Settling velocity = 0.126 m/s
 Upward (red) flow = 0.3 m/s
 Downward (blue) flow = 0.15 m/s
 Resulting Np = 4.7 (fully baffled  7.5)
 Resulting Power = 230 kW
 Resulting P/V = 0.037 kW/m3 = 0.18 Hp/1000
gallons

Post Mixing
Optimization and Solutions

You might also like