• DECEMBER 2019

CFD simulation of stirred

tank mixing processes
using Ansys
Sravan Kumar Nallamothu
Outline

• Engineering challenges
• Ansys capabilities:
‐ Case studies
‐ Democratization
• Summary

Animation of tracer mixing in a stirred tank

Acidic components (blue-green) mix into the bulk of the vessel

Animation: Ansys Ensight + Ansys VRXperience

4
Significant Challenges Exist To Operate Mixing Equipment

Scale-up from lab to production scale Liter → 1,000s of liters

• Scale-Up Time Hours → weeks (operating conditions)

Limited lab time and testing capacity Weeks of delay

• Yield Material cost Pharma API costs more than GOLD

• Time to Market Material change Change over time and cleaning

Operating cost
•
Power consumption
New Products Overall system performance System sizing

System commissioning and control Operation

 Quantified Impact Of Simulation On Efficient Mixing Equipment
Utilization

Production Time
Performance Revenue
Reduction

1000x 40% 1%
Acceleration of Reduction in culture Increase in concentration
simulation time with development time2 can add millions of dollars
reduced-order models1 of revenue per batch3

1 Jatale et al, AICHE 2020

2 http://storage.ansys.com/webinar/171017-f910.mp4?utm_medium=SW20IBPMixing
3 Engineering estimate based on process/cost knowledge
Simulation Capabilities Required To Realize These Benefits
Multiscale, multidomain and multiphysics simulation for mixing equipment
Single Phase Mixing Systems

Gas-Liquid Systems & Bioreactors

Solid Dissolution (Solid–Liquid Systems)

Intermeshing Mixers

Static Mixers

Clean-In-Pace (CIP)

Democratization and System-Level Modeling

7
Physics in Agitated Vessels

• Single-phase flows
• Velocity field
• Turbulence
• Gas-liquid flows
• Bubble size distribution

Increasing complexity
• Mass Transfer
• Vortex prediction
• Liquid-solid flows
• Solid suspension
• Particle dissolution
• Powder mixing
• Liquid-liquid flows
• Micro-mixing
• Emulsion
• Reacting flows
• Product selectivity
• Mass transfer
• Crystallization
Flow field generated by LA310 impeller

8
Physics in Agitated Vessels

• Single-phase flows
• Velocity field
• Turbulence
• Gas-liquid flows
• Bubble size distribution

Increasing complexity
• Mass Transfer
• Vortex prediction
• Liquid-solid flows
• Solid suspension
• Particle dissolution
• Powder mixing
• Liquid-liquid flows
• Micro-mixing
• Emulsion
• Reacting flows
• Product selectivity
• Mass transfer
• Crystallization

9
Single Phase Flow Analysis
Shear rate characteristics of reactor Zonal Residence Time Distribution
Numerical data
Quantity Zone 1 Zone 2 Zone 3
• Dissipation rate
• Shear rate Zone Ht (m) 1.01 2.20 3.37
• Power number Total Visit No, 109k 180k 71 K
• Blend time Average RTD(s) 7.11 5.56 16.01
• Flow number
Zone3
• Kolmogorov length
Zone 2

Zone 1

Field data at every location

Blend time

Velocity vectors in an unbaffled reactor

Blend time data at different locations in reactor

10
Single Phase Analysis

Stratified Improved

11
 Modeling Scale-up of Dissolution

Instantaneous concentration
• Scale dissolution of Propylparaben from 2L to 4,000L
vessel
• Determine operating conditions that predict similar
dissolution profiles at both scales
CFD
• Geometric similarity of reactors Lab
• Ansys Fluent used to determine conditions where
P/V is maintained at all scales
Offset due to clumping at low rpm Typical results
“In this work, we have successfully demonstrated
that scale-up of solid dissolution can be done
using a systematic process of lab-scale
experiments followed by simple CFD
modeling to predict commercial-scale
experimental conditions”

Application of Modeling to Scale-up Dissolution in Pharmaceutical Manufacturing by Venkat Koganti (Pfizer Global Research
and Development) AAPS PharmSciTech, Vol. 11, No. 4, December 2010

12
Scale up analysis

• Objective: Evaluate local energy

dissipation rate as scale up criterion for
geometrically scaled vessels
• Number of reactors: 4
‐ 20, 200, 2,000, 20,000 liter
• Operating Conditions:
‐ 20Litre reactor is run at 63 rpm
‐ RPM for other reactor selected such as to obtain same
local energy dissipation rate Baffled vessel with two A315 impellers

P Po N 3 D 5 Po N 3 D 5 Po 3 2
Assume   = = = N D    N 3 2
D
Vimp D 2
D 2

 Dw D
4 4 4

13
Simulation Results – Flow and Blend Time
20-Liter 200-Liter
20-
• Types of simulations done Liter
‐ Flow
‐ Blend Time
‐ Exposure Analysis (TDR, Shear Rate) 2000-Liter 20000-Liter

‐ Zonal Residence Time Distribution (RTD)

Power Numbers:
20L 200L 2000L 20000L
RPM 63 37.7 22.6 13.6
CFD Upper Po 0.85 0.84 0.83 0.82
Observations:
CFD Lower Po 0.5 0.49 0.48 0.48 • Vendor specified “single impeller” power
Table 1: Power Number for both impellers number is 0.75
Blend Time
20L 200L 2000L 20000L
– Upper impeller close to that
Blend Time 25.8s 41.9 61.2s 111.8s – Lower impeller draws lower power
Table 2: Blend time for different sizes
• Blend time increases with scale

14
Simulation Results – Exposure Analysis and Zonal RTD
Exposure Analysis: Shear Rate

Observations:
‐ Cells/Particles exposure to high shear rate
decreases with increase in reactor size
‐ Dissipation rate profiles are identical
‐ “Normalized” zonal residence time
behaviors similar for all reactor sizes
Zone1 Low intensity at high reactor volumes
Zonal Residence Time Distribution: Zone 1
Exposure Analysis: Dissipation Rate

Zone3

Zone2

Zone1

Time taken for the study < 1 day

Zonal Residence Time Distribution – Similar distribution at different scales Dissipation Rate Exposure Analysis – Almost Identical environment

15
 Fluid flow of shear thinning material in stirred tank

• Venneker et al.2,
‐ Flat bottom tank
‐ 6 bladed Rushton turbine
‐ Full baffled condition
• Mesh
‐ Polyhedra cells with boundary layers
• Operating conditions
‐ T = 0.627 m
‐ 0.1% Blanose
• K = 13.2e-3, n = 0.85
‐ Rotational Speed
Mesh used in this study
• 3.8 rev/sec
‐ Turbulent flow
• RKE model
2Bart C.H. Venneker1, Jos J. Derksen2, Harry E.A. Van den Akker,Turbulent, flow of shear-thinning liquids in stirred tanks—The effects of Reynolds number and flow index, chemical engineering research and design 88 (2010 ) 827–843

16
Flow patterns: Velocity, strain rate & viscosity distribution

Center Plane Impeller Plane Center Plane Center Plane

Velocity distribution Velocity distribution Strain rate distribution Viscosity distribution

17
Velocity profile comparison

Measurement
location

18
Physics in Agitated Vessels

• Single-phase flows
• Velocity field
• Turbulence
• Gas-liquid flows
• Bubble size distribution

Increasing complexity
• Mass Transfer
• Vortex prediction
• Liquid-liquid flows
• Micro-mixing
• Emulsion
• Liquid-solid flows
• Solid suspension
• Particle dissolution
• Powder mixing
• Reacting flows
• Product selectivity
• Mass transfer
• Crystallization
Gas dispersion by LA315 impellers

19
Gas-liquid Flows

Gas dispersion by Chemineer CD-6 impeller

• Challenges
– Prediction of efficient mass transfer
– Gas distribution throughout the tank
– Power draw by impeller for sparging
– Limiting maximum shear rates
– Scale-up process to large reactor Short circuiting at
baffles
• Benefits of CFD
‐ Detailed information for the flow field and
shear rate characteristics
‐ Prediction of gas holdup
‐ Mass transfer quantities:
• Estimation of local and global Kla Picture
• Interfacial area
‐ Power dissipation can be obtained to study
effect of gas
Velocity vectors (left) Gas distribution (right)
‐ Gas distribution through out the vessel calculated in a tall fermentor
‐ Simulate different conditions for scale up
studies

20
Gas holdup and mass transfer in stirred gas-liquid tank

• Experiments were done by Laakkonen1

• Geometry:
• Reactor volume: 194 Lit
• 6 blade Rushton-Turbine
• Operating conditions
• Angular velocity: 390 rpm
• Gas flow rate: 0.7 vvm
• Mesh
‐ Polyhedra mesh Image showing the mesh used for stirred tank
with ring sparger
• Population balance method to predict An image of experimental setup with
sintered sparger2
bubble size distribution
1. Laakkonen M, Alopaeus V, Aittamaa J. Modelling local bubble size distributions in agitated vessels. Chem Eng Sci. 2007;62;721-740
2. http://lib.tkk.fi/Diss/2006/isbn9512284618/

21
Bubble size distribution

390 rpm, 0.7 vvm

D
Bubble size distribution on Iso-surface 3.3
of gas volume fraction 10%
3.3 E
4.1
4.5
C
2.9
2.6

B
2.5
2.3

A
3.2
-- Experiment
2.7
-- Simulation

22
Gas holdup comparison

Simulation results match very well with

experimental measurements

Simulations slightly over-predicted gas holdup

at higher gas flow rates

23
Dissolved Oxygen

Normalized
dissolved oxygen
concentration at a
point near the
liquid level

Contours of oxygen mass fraction on center plane along with gas volumetric distribution

https://www.aiche.org/conferences/aiche-annual-meeting/2018/proceeding/paper/297c-cfd-modeling-oxygen-dissolution-bioreactors-mass-transfer-and-population-balance-study-stirred

24
 Design Of Experiments (DOE)

• DOE & response surface

‐ 10 design points
‐ Total run time: 10 hours!

Comparison of solution speed scale-up with

Response surface of mass different mesh densities
transfer coefficient as function of
gas flow rate and agitation speed

25
Gas-Liquid flows: Vortex Prediction

• Surface vortex formation is often seen as

undesirable effect that needs to be
avoided
‐ Air entrainment
‐ Improper solid mixing
• Unbaffled vessels are needed in specific
applications to avoid dead zones, minimize
impurities and avoid cleaning issues with
high viscosity fluids
‐ Baffles break the tangential velocities and lead
to better axial mixing
• Predicting the surface vortex behaviour in
unbaffled and baffled tanks is required to
decide the operating range for the
agitation speed, shaft angle and impeller
offset etc.

26
 Geometry

• Mixing tank geometry1:

• Cylindrical tank
• 6 bladed Rushton Turbine impeller

• Mesh:
• Polyhedra: 89K
‐ Dynamic mesh adaption (PUMA)

1. Jennifer N. Haque, Tariq Mahmud, Kevin J. Roberts, and Dominic Rhodes, Modeling Turbulent Flows with Free-Surface in Unbaffled Agitated Vessels
Ind. Eng. Chem. Res., 2006, 45 (8), 2881-2891

27
Free surface profile: Comparison

Animation of liquid volume fraction

with PUMA method for automatic
Experimental data points are measured from haque et. al. mesh adaption

28
Physics in Agitated Vessels

• Single-phase flows
• Velocity field
• Turbulence
• Gas-liquid flows
• Bubble size distribution

Increasing complexity
• Mass Transfer
• Vortex prediction
• Liquid-solid flows
• Solid suspension
• Particle dissolution
• Powder mixing
• Liquid-liquid flows
• Micro-mixing
• Emulsion
• Reacting flows
• Product selectivity
• Mass transfer
• Crystallization
Suspension experiments & CFD (Bakker et al, 1994)

29
Solid Suspension Modeling

100 rpm Clear liquid 150 rpm 220 rpm

• Challenges
‐ Solid suspension is a key concern for:
• Solid catalyzed reactions
• Crystal growth
• Dissolution
‐ Uniform solid suspension Higher solids
concentration
‐ Power prediction with the presence of solids
Solid profile at different agitation speeds. Ref: IchemE, 2008

Benefits of CAE
Iso-surface and contours showing – Detailed information for the flow field
solid distribution and cloud – Shear rate characteristics
height in a conical based vessel
• Many crystals can be damaged by exposure to
regions of high shear
– Predict Just Suspension Velocity
– Predict the solids concentration profile through out
the vessel

30
Validation of cloud height: Problem Description

• Reactor details
‐ Vessel Diameter: 0.61m
‐ Liquid Level: 0.915m
‐ Impeller Diameter: 0.2m
‐ Clearance from Bottom: 0.15m and 0.39 m
‐ Tank Bottom: Torispherical
• Material Properties
‐ Liquid Density: 1000 kg/m3
‐ Liquid Viscosity: 0.001 Pa-s
‐ Solid Density: 2630 kg/m3
‐ Particle Diameter: 180 micron
• Operating Conditions
‐ Solid Concentration: 10% wt and 15% wt
‐ Agitation Rate: 150 RPM to 450 RPM in the steps
Gohel, Shitanshu et al., CFD Modeling of Solid Suspension in a Stirred Tank: Effect of Drag
of 50 RPM Models and Turbulent Dispersion on Cloud Height, IJChE, 2012, 956975

31
Effect of Turbulent Dispersion Force

With TDF Without TDF

Single Impeller, 150 RPM, 10% wt Loading

32
Cloud height comparison

Plot for 10% wt Loading Plot for 15% wt Loading

Hc/H
Hc/H

N/Njs N/Njs

Gohel, Shitanshu et al., CFD Modeling of Solid Suspension in a Stirred Tank: Effect of Drag Models and Turbulent Dispersion on Cloud Height, IJChE, 2012, 956975

33
Particle dissolution: Design space study

• Particles dissolving in a stirred tank

reactor.
‐ Particle size distribution
‐ Dissolution rate dependent on particle size
and external concentration
‐ Dissolution and convection time scales.
• Predict mixing time as a function of
‐ Initial particle size distribution (0.1 – 2 mm)
‐ Saturation concentration (150 – 3000 kg/m3) • Cylindrical Tank
‐ Particle density (150 – 3000 kg/m3) − Capacity=720L
− Tank diameter = 40”
• Mesh
− 100K Polyhedral elements

34
Physics of dissolution: Mass transfer model

• Mass transfer from salt particle to solution

m = kA(c sat − c)
There are many correlations available in literature
• k is mass transfer coefficient for mass transfer coefficient. One can adjust the rate constants
to match experimental data
• Calculation of k
• Sherwood number for dissolution of sphere1

d p4 / 3 1/ 3 
Re p = ; Sc =
/ D
kd p
Sh = = 2.0 + a * Re 0p.5 Sc1/ 3 Where
D

a = coefficient to be tuned
For cylindrical tank, 0.4 value is used in literature
Pure diffusion + Convection Lower value means rate is closer to diffusion only

1Sano et al., Mass transfer coefficients for suspended particles in agitated vessels and bubble columns, Journal of chemical engineering of Japan, Vol. 7, No. 4, 1974

35
Results from CFD

Fast dissolution

Slow dissolution

36
Response surface generation

37
Democratization and
System Level Modeling
Democratization and System Level Modeling

Geometry
&
Meshing

Mixing
Post
Solver modeling
processing
tool

Design
space
study

39
Democratization and System Level Modeling

• Ansys provides key features to address these challenges:

‐ Ansys has developed a Mixing Wizard that removes the burden of having to be a CFD expert to run
mixing analysis. It utilizes Ansys Fluent as the workhorse with its full capabilities.
‐ Reduced-Order Models (ROMs) can be easily created for mixing simulation that can be easily
consumed by plant operators and this eliminating the need for CFD knowledge, training, or
simulation-ready hardware.
‐ ROMs can also be integrated within Ansys Twin Builder to represent a bigger system with additional
components for digital commissioning and operator training.

40
Democratization and System Level Modeling

• Mixing Wizard automates:

‐ Geometry creation
‐ Meshing
‐ Solution setup
‐ Postprocessing and reporting
• Tabulated data
▪ Torque, power
• Contours
▪ Flow, turbulence, Kolmogorov mixing length
• X-Y-plots
▪ Blend time, RTD, exposure analysis
• Animation
▪ Blend time

41
Democratization and System Level Modeling

• Comparison between flow field 3D CFD ROM

obtained by 3D CFD and the
corresponding flow field obtained by
running the ROM. 3D CFD takes hours
while ROM takes seconds. ROM results
are visualized in Ansys Fluent on same
mesh.
• An alternative way to consume ROM is
through the standalone and light ROM
viewer, which enables real time results
visualization by changing model inputs
(e.g., rpm, liquid properties). Only
process knowledge (not CFD expertise)
needed.

42
Resources

• MixingWizard in Appstore
‐ https://catalog.ansys.com/product/5c5af1ce
393ff6123869b583/mixing-guided-proc
• Tutorials for stirred tank mixing
‐ Ansys Learning Hub
(https://jam8.sapjam.com/groups/LP7HIZTia
hvkzzBFjYbUsa/overview_page/HnmPLV7e7
5evzFhqIuSCF9)
‐ Currently 7 tutorials on stirred tank and 2
tutorials on static mixers
‐ More tutorials are being created

43
Summary

• Quality stirred tank mixing is central to

success to bringing products/therapies
to market quickly
• Ansys offers easy to use tool that can
simulate various complex physical
processes that occur in stirred tanks
• Accurate and comprehensive
simulation capability avoids costly
mistakes and improves yield
• Democratization of simulation
bestpractices through automation and
reduced order modeling (ROM)
provides a quicker way of addressing
the issues with reactor design

44
 Join the Ansys conversation!

Read and comment.

Find the Ansys blog:

ansys.com/blog

45
Questions?

To ask a question, click on the Q&A button at the

bottom of the screen to display the Q&A window:

The recording will be available in about

one week on the Ansys resource library:
ansys.com/resource-library

46
Appendix
Physics in Agitated Vessels

• Single-phase flows
• Velocity field
• Turbulence
• Gas-liquid flows
• Bubble size distribution

Increasing complexity
• Mass Transfer
• Vortex prediction
• Liquid-solid flows
• Solid suspension
• Particle dissolution
• Powder mixing
• Liquid-liquid flows
• Micro-mixing
• Emulsion
• Reacting flows
• Product selectivity
• Mass transfer
• Crystallization Oil in water emulsion

48
Emulsions

• Agricultural:
‐ Pesticides
• Industrial:
‐ Lubricants, surface coatings, roads,
highways, paints, polishes
• Cosmetics:
‐ Lotions, foundation creams, cold creams,
shaving creams,…
• Food:
‐ Margarine, ice cream, nutraceutical drinks
• Pharmaceuticals
‐ Vitamin drops, milk of magnesia…
Droplet diameter contours on center plane

49
Immiscible liquid mixing

• Kerosene in water dispersion

‐ Kerosene is added to water in stirred tank in
20% volume ratio, which will form a
dispersion with a droplet size distribution
governed by flow dynamics at given
operating conditions
‐ No surfactants
‐ Population balance model is used to predict
drop size distribution
Drip size distribution as function of loading

51
Physics in Agitated Vessels

• Single-phase flows
• Velocity field
• Turbulence
• Gas-liquid flows
• Bubble size distribution

Increasing complexity
• Mass Transfer
• Vortex prediction
• Liquid-solid flows
• Solid suspension
• Particle dissolution
• Powder mixing
• Liquid-liquid flows
• Micro-mixing
• Emulsion
• Reacting flows
• Product selectivity
• Mass transfer
• Crystallization Crystallization in a stirred tank

52
Crystallization

• Challenges
‐ Well designed operating parameters
Number density at outlet
‐ Super saturation
‐ Uniform suspension
‐ Multiphase heat and mass transfer
‐ Restrictive particle size and shape
specifications Volume fraction of lactose crystals

‐ Narrow crystal size distribution • Benefits of CAE

‐ Micro-mixing effects • Flow characterization at different scales
• Just suspended speed
• Population balance modeling to account for nucleation,
growth and breakage
• Predict crystal size distribution
• Super saturation history/profile
• Reduce batch time by using optimal conditions

53
Clean-In-Place(CIP)

• Key challenges
‐ Clean the vessel using the least amount of
cleaning agent to reduce cost and
environmental footprint within the shortest
amount of time to reduce equipment
downtime.

• Simulation challenges
‐ Very wide length scales; vessel size is typically
in meters (possibly even tens of meters) vs.
spray nozzle dimensions (and hence cleaning
jet size) that is typically in mm or less
‐ Nozzles could have complex motion
(combination of rotation around different axis)
to ensure coverage. Rotation can either be Cleaning water jet within a vessel
flow-driven or motor-driven. colored by velocity magnitude,
courtesy Spraying Systems

54
Clean-In-Place(CIP)

0.50
0.45
0.40

Total Pressure (MPa)

0.35
0.30
Exp
0.25
0.20
0.15 CFD
0.10
0.05
0.00
0.00 1.00 2.00 3.00 4.00 5.00 6.00
Distance (m)

PUMA efficiently capturing Cleaning jet velocity Comparison between recorded

cleaning jet formation magnitude variation impact pressure vs. distance
between CFD and experiments1

1 Sami et al, “Using CFD to Simulate Mixing Tank Clean-In-Place Process”, NAFEMS 2019 conference

55
Crystallization- BaSO4 Precipitation Example

• Semi-batch process
• BaSO4 precipitation
• Transient behaviour
• Better understanding 150 mm

• CFD as a validation tool k

• Effect of impeller
m
• Effect of feed location 50 mm

150 mm

56
Flow Field in Precipitator

• The flow field affects scalar mixing significantly and its interaction with crystallization
kinetics

Flow pattern in the precipitator at 150rpm

57
Transient Behaviour of Supersaturation

0.5 seconds 10 seconds 100 seconds 165 seconds

• The high S zone remains similar in shape and volume over the batch period
• The zone defines a segregated region where micromixing and mesomixing
effects are significant

58
Transient Behaviour of CSD (1)
Crystal number concentrations
1.E+14

• The S level remains low with perfect mixing

Crystal number concentration (log), 1/m3

1.E+12

assumption so that nucleation is by a 1.E+10

150rpm 300rpm 600rpm

heterogeneous mechanism 1.E+08

• The local S level can be as much as 2000 due 1.E+06

Perfect Mixing

to non-perfect mixing (captured by CFD) and 1.E+04

homogeneous nucleation occurs 1.E+02

0 50 100 150 200

• The predicted d32 also increases linearly with

Time (s)

the operating time apart from the initial stage 14

600rpm
‐ d32 behaves nonlinearly with different agitation 12 150rpm

speed

Mean crystal size, d32 (mm)

10

8 Perfect Mixing

6
300rpm
4

2
Mean crystal sizes
0
0 50 100 150 200
Time (s)

59
Supersaturation vs. Impeller Speed

• Impeller speed affects both the peak level of supersaturation and the extent of the
supersaturated zone

N = 150 rpm at 10 seconds N = 600 rpm at 10 seconds

60