CHARACTERIZING IMPELLER PERFORMANCE

IN STIRRED TANKS WITH EXAMPLES OF

PROCESS RESULTS

Richard K. Grenville
Director of Mixing Technology

“Mix Thoroughly Until Smooth”

IChemE South Wales Branch

25th April 2017

 TOPICS
• Some words on mixing….
• Impeller geometries:
– Blade shapes etc.
• Re-cap on pump sizing.
• Applying pump rules to impellers in stirred tanks:
– Defining efficiency.
• Shear characteristics:
– Time-averaged velocity gradient.
• Dispersion of immiscible liquids.
• Trailing vortex:
– Look again at dispersion data.
• Conclusions.

2
 IMPORTANCE OF MIXING
• Smith (Trans IChemE., 1990):
– US chemical industry loses $ 10  109 each year due to
poor mixing:
• 1 % increase in yield  ~ $ 1  106.
• One day of down time  ~ $ 1  106.
• Examples:
– Low yields in chemical reactions:
• Change in selectivity on scale-up.
• By-products, MWD, etc.
– Longer than expected batch / cycle times.
– Effluent from WWTP out of compliance.
– Others………………..?
 SCALE
• Lab-scale: Vessel diameter 0.1 m (4 inches)
Operating volume 1 litre (0.25 gall)

• Fine chemicals: Vessel diameter 1.8 m (72 inches)

Operating volume 8000 litres (2000 gall)

• Pharma fermenter: Vessel diameter 6.0 m (240 inches)

Operating volume 300000 litres (80000 gall)

• Bio-fuel fermenter: Vessel diameter 18 m (720 inches)

Operating volume 4 million litres (1 million gall)

• Gasoline storage: Vessel diameter 36 m (1400 inches)

Operating volume 10 million litres (3 million gall)
 6 MILLION GALLON COAL SLURRY TANK

Philadelphia Mixers.

Installed in Mojave
Desert.

Coal slurry pumped to

Power Generation
plant.

Shaft can be raised for

start-up.

Then lowered as bed of

particles is suspended.
 WHAT IS MIXING? (Etchells)
Mixing is the application of

mechanical motion

Mean Flow - Velocity in order to create Turbulence

fluid dynamic effects

Circulation / Blend Time which achieve a Shear

desired process result.

TYPICAL MIXING EQUIPMENT
 HYDROFOILS

Low-solidity Medium-solidity Anti-ragging

Po ≈ 0.30 Po ≈ 0.70 Po ≈ 0.30

“Low shear” impellers (Ducoste et al, AIChEJ, 1997).

“RAGS”
PITCHED & FLAT BLADE TURBINES

Pitched Vertical / Flat

Po ≈ 1.30 Po ≈ 3.00
 DISC TURBINES

Rushton Smith PMSL GDX

Po ≈ 5.00 Po ≈ 3.00 Po ≈ 1.00

“High shear” impellers (Ducoste et al, AIChEJ, 1997).

 HIGH-SHEAR / SAWTOOTH

HSD

www.indiamart.com Po ≈ 0.10 – 0.30

 DEFINING IMPELLER CHARACTERISTICS

P
• Power number: Po = 3 5
ρN D

• Flow number: Q
Fl =
ND3

• Diameter ratio: D
T
D
• Projected blade height:
wP
 POWER INPUT TO FLOWING FLUID
• In pipe flow:
*Note: DH is Pressure Drop (Pa) – not Head (m).

P = QΔH

2
π L U π fL
P = UD × 4 f ρ
2
= ρU3D 2
4 D 2 2D
• First term in equation is determined by geometry
and flow regime:
– Reynolds number, pipe roughness.
– Pipe length to diameter ratio.
– Fittings (elbows, valves, etc.).
Q = 1450 GPM  0.0915 m3 / s
ΔH = 90 ft  269107 Pa
P = 24.62 kW  33.0 HP
φ = 33.0 / 66.4 = 0.497
 AGITATOR - REYNOLDS NUMBER
ρUD
• In pipe flow: Re =
μ

• What are the appropriate velocity and length scale to

use when calculating Reynolds number in a stirred
tank?

• Velocity  Impeller tip speed: UTIP=πND

• Diameter  Impeller diameter: D

• Drop the π: ρND2

Re =
μ
 POWER INPUT BY AN IMPELLER
• In pipe flow: P ∝ ρU D 3 2

• Substitute Tip Speed for Velocity (U) as for Reynolds

number:
3 5
P = XρN D
X is the Impeller Power number, Po. It is determined by
geometry and flow regime:
– Reynolds number
– Impeller diameter / Blade width / Number of blades
– Impeller position (relative to vessel base)
– Baffles
 TURBULENT & LAMINAR REGIMES
• In turbulent regime: Po = Constant

P = PoρN3D5

• In laminar regime: KP
Po =
Re

P = KPμN2D3
FRICTION FACTOR vs. REYNOLDS NUMBER
Turbulent : Po or f = k 1

k2
Laminar : Po or f =
Re

Rushton et al., Chem. Eng. Prog., 1950

 POWER, FLOW & EFFICIENCY
• In pipe flow, the Power Input to the Fluid, Flow
Rate and Pressure Drop are easy to define and
measure.

• If the Motor / Shaft Power is measured, the Pump

Efficiency can be quantified.

• How are the Power Input and Flow generated by

an impeller measured?
 MEASURING POWER
• At lab-scale, use DC motor:
P = IV
• Use Wattmeter:
– Direct measurement of power for AC 3-
phase motor.

• Measure current:
– With motor Power vs. Current curve.

• Strain gauges measure

torque:
http://en.wikipedia.org/wiki/Strain_gauge

P
Tq =
2πN
 MEASURING FLOW
• Measure velocities in discharge
from impeller:
– Using Laser Doppler Anemometry or Particle
Image Velocimetry.

• Flow and velocities are time

dependent.
• Calculate mean velocities.
• Define Flow (or Pumping)
Number:
Q = ∫∫ U dA

Q
Fl =
ND3
Brown, MIXING XXII, 2010
 TIME-AVERAGED VELOCITIES

Brown, MIXING XXII, 2010

 TIME-DEPENDENT VELOCITIES

Symmetric flow field Nearly symmetric

→ low vorticity flow field→ low vorticity

Axisymmetric flow field Axisymmetric flow field

→ high +ve vorticity → high -ve vorticity

Myers et al, J. Fluids Eng., 1997

 PLOT Fl vs. Po
1

Rushtons: Fl = 0.63
For axial flow impellers:
Fl (-)

0.1 Fl = 0.77 Po0.33

PBT
N-HYFL
W-HYFL
FBT
RUSH
SAW
HSDs: Fl = 0.04 EE

0.01
0.01 0.1 1 10
Po (-)
 PRESSURE DROP / HEAD
• What is pressure drop in an agitated vessel? What
quantity is equivalent to pressure drop in pipe flow?
P(W) J Nm
2
ΔH(N/m ) = 3
→ 3 or 3
Q (m /s) m m

• In an agitated vessel “DH” is a measure of the

impellers’ efficiency:
– How much energy is input by the impeller to move a
volume of fluid?
PLOT OF EFFIENCY vs. POWER NUMBER
 MASS FLOW EFFICIENCY
• Mass of Fluid Pumped per unit of Energy Input (Fořt et al.,
2010, Chem. Proc. Eng.).
3
FlρND Fl
η= 3 3
= 2 2
PoρN D PoN D

• At equal Power Input per Mass:

Fl D 4/3 -2 / 3
η = 1.18 1 / 3 ( ) (ε T)
Po T
• Plot η versus D/T.
 FLOW EFFICIENCY

Brown, 2010,MIXING XXII and PMSL

 SHEAR RATE

Area = A Force = F
Velocity = V

Gap = y
V
γ =
y
 SHEAR (OLDSHUE (1983))

“Lower shear” “Higher shear”

SHEAR (after OLDSHUE (1983))
SHEAR (after OLDSHUE (1983))

33
 SHEAR RATE
• Shear Rate is a Velocity Gradient:
– Steeper gradient → higher shear rate.

(α - β) VTIP VTIP
γ = =Λ
(ψ - ω) R R

Impeller α β ψ ω Λ Λ / ΛHYDFL Λ / ΛHYDFL

at equal VTIP at equal ε̅
Hydrofoil 0.25 0.14 0.80 0.16 0.17 1.00 1.00

PBT 0.35 0.07 0.65 0.12 0.53 3.11 1.91

Rushton 0.60 0.16 0.12 0.04 5.50 24.4 9.55

 PROCESS RESULT
• Dispersion of Silicone Oil in Water.

• Low volume fraction → non-coalescing:

– Droplet size determined by break-up (Hinze, AIChEJ, 1955).

• Diameters of all impellers are ~50 % of vessel diameter:

– Rushton turbine.
– Pitched blade turbines – 45 and 60 degree blade angle.
– Hydrofoils.
– Sawtooth / High Speed Disperser.

• See also Pacek et al., 1999, Chem. Eng. Sci.

 MULTI-PHASE MIXING
• Turbulent eddies interact with second phase.
• Equate stress / energy of turbulent eddy with
resisting stress / energy.
• For example, in liquid-liquid and gas-liquid
dispersions (Hinze, 1955, AIChEJ):

u' σ σ
ρ(u' ) = k1μ + k 2
2 2/3
(εd) d = k 2
d d ρ

σ 3 / 5 -2 / 5
d = k3 ( ) ε
ρ
d32 vs. ε̅ – All D = ~ T / 2
 RESULTS
• At the same Power Input per Mass:
– Rushton and PBTs make the same droplet size.
– Hydrofoil makes a smaller droplet (~1/2 diameter).
– HSD / Sawtooth makes the smallest (~1/3 diameter).

• Expect PBT to make smaller drops than Hydrofoil if

Hydraulic and Flow Efficiencies and Oldshue’s Shear
Rate correctly account for impellers’ “characteristics”.

• What is missing?
 TRAILING VORTEX

Lattice Boltzmann Large-Eddy Simulation using DMT software from M-Star Simulations
 TRAILING VORTEX

lV = D/x

uV

Turbines: lV ≈ wP / 2
Hydrofoils: lV ≈ wP

http://www.csiro.au/~/media/34F42F86D14645F3B6C3710F4306392E.ashx
 AVERAGE POWER PER MASS
• Simple calculation if impeller Power number is
known.

• For vessel where H = T:

PoN3D5
ε=
(π / 4 )T 3
 MAXIMUM TKE, kmax, & TKEDR, εmax
• Trailing Vortex Kinetic Energy is scaled by Tip Speed
Squared.
• Grenville et al. (2008) showed that, for bladed
impellers:
k max 1/2
2
= 0.104Po (±10%)
UTIP

A=1
3/2 3/2
k max k max D
εmax = A = Ax where : lV =
lV D x
 VORTEX LENGTH SCALE – lV
• Measurements made using Dantec three beam, two
component LDA in back scatter mode with Burst
Spectrum Analyzer (e.g. Lee & Yianneskis (1998)).

D
lV =
x

Rushton turbine: x = 12
Pitched blade turbine: x = 17
A310 Hydrofoil: x = 16
Sawtooth impeller: x = 12
 TVEDR
• Combining:

k max = 0.104Po1 / 2U2TIP

(0.104Po1 / 2U2TIP )3 / 2
εmax = Ax
D

εmax = 1.04xPo3 / 4N3D2

 RATIO
• Ratio of maximum TVEDR to average Power per
Mass: ε (π / 4 )T 3
max
= 1.04xPo3 / 4N3D2 ×
ε PoN3D5

εmax x T3 K
K= = 0.82 1 / 4 ( ) (±15%) Φ=
ε Po D K Rushton
Impeller x Po D/T K F

Rushton 12 4.69 0.48 70.43 1.00

4PBT-45 17 1.16 0.51 75.08 1.07
4PBT 60 17 1.67 0.51 83.58 1.19
A310 16 0.26 0.44 258.6 3.67
Sawtooth 12 0.08 0.51 3318 47.11
d32 vs. ε̅ – All D = ~ T / 2
d32 versus εmax – D = ~ T/2

3/2
k max
εmax = A
lV

-0.4
d32 = 423ε max

RSD = ± 10.6 %
AVG. FLOC LENGTH vs. εMAX
 CONCLUSIONS
• Impellers in stirred tanks are pumps:
– Power input and flow generated can be measured.
• Impeller characteristics are rigorously quantified:
– Hydraulic and Flow Efficiencies.
– Maximum Energy Dissipation Rate in Trailing Vortex.
• Characterization confirmed by Equilibrium Droplet and Floc
Sizes generated with different impellers.
• “Shear”:
– Not strictly correct mechanistic description.
– Conveys concept → size reduction or growth.
• Do we mean “high stress impellers”?
• Need to question “conventional wisdom” – not just for mixing:
– Better measurement, analysis and design techniques.
BACK-UP CHARTS

50
 TURBULENT & LAMINAR REGIMES
π fL
• In turbulent regime: X= = Constant
2D

P = XρU3D 2

16
• In laminar regime: f=
Re

π 16 L Y
X= =
2 Re D Re

P = YμU2D
 ENERGY & POWER
• Kinetic energy  m * U2

• Kinetic energy per volume  ρ * U2

• Kinetic energy per mass  U2

• Characteristic timescale L/U

• Power per mass = Energy per mass / Time  U3 / L

• Also “energy dissipation rate (per mass)”.

52
 POWER INPUT PER UNIT MASS
π fL 3 2
• Power Input to fluid: P= ρU D
2D
π
• Mass of fluid in pipe: M = ρ LD 2
4
• Power Input per Mass:

P (π / 2)fLρU3D U3
ε= = = 2f
M (π / 4)ρLD 2 D

• Characteristic Velocity and Length Scale for pipe flow.

 GRENVILLE et al. (2008)
1

kmax / UTIP2 = 0.1037 Po0.5444

r2 = 0.9595
k max / U TIP (-)
2

0.1

RT
HYDs
PBTs
SAW

0.01
0.01 0.1 1 10

Po (-)
 FLOCCULATION
• Removal of fine particles (sub-micron) from liquid:
– Water and Waste Water Treatment.
– Polymer processing.
• Flocculation increases particle size:
– Agglomerates of many particles.
– Easier to remove by sedimentation or filtration.

• Final floc size determined by equilibrium between growth and

breakage.
• Time affects floc shape and density:
– Breakage and re-flocculation.

• Effect of mixing / shear rate?

 LITERATURE
Stein & Camp, 1943, J. Bos, Soc. Civ. Engrs.
Defined Shear Rate – G.

ε 1 /2 P 1 /2 1
G=( ) =( ) [= ]
ν Vμ τK
Commonly used in Water Treatment industry.
Scale-up of Jar Tests.

http://www.nesc.wvu.edu/pdf/dw/publications/ontap/2009_tb/jar_testing_dwfsom73.pdf

Benz, 2007, Chem. Eng. Prog.

“G-value has no legitimate use in designing or specifying
agitators”.
 LITERATURE
Koh et al., 1983, Chem. Eng. Sci.
Defined effective Shear Rate:
Couette flow Laminar and Taylor Vortices:
2 2 1 /2
ε Vort 1 / 2
GEff = (G Lam +G )
Vort G Vort =( )
ν

Coufort et al., 2007, Chem. Eng. Proc.

For both inertial and viscous sub-range expect:
L Floc ∝ ε -1 / 4

ε̅ not sufficient to predict agglomeration phenomena.

Knowledge of local hydrodynamics is needed.

Kresta, 1998, Can. J. Chem. Eng.

 SPICER, 1997, PhD THESIS
• Measured Floc Length versus Time:
– Equipment:
• Tank: T = 0.15 m
• Impellers: Rushton, PBT and Hydrofoil D = T/3
• G-value: 15, 25 and 50 s-1
• Particles: Polystyrene d = 870 nm φV = 1.4 x 10-5
• Flocculant: Aluminum sulfate hydrate 0.5 g / lt.
– Initially mixed at G = 300 s-1 to remove agglomerates.

• FSD measured with optical microscope and video camera.

– Images digitized and recorded with Image Analysis software.

• Plot Number Average Maximum Floc Length versus Time:

– Longest dimension of Floc.
MECHANISM OF FLOC GROWTH

Primary particles collide and agglomerate.

Long flocs break.
Small fragments collide and reorganize.
Become more compact over time.
NAMFL vs. TIME

λK, AVG =141 μm

 NAMFL vs. TIME

λ̅K = 200 μm
 NAMFL vs. TIME

λ̅K = 258 μm
NAMFL vs. TIME – EQUAL εMAX

204 μm ± 0 %

187 μm ± 4.4 %

λK, MAX = 55 μm
 EXPONENT x = -0.35
• Exponent, x, on εMAX is -0.35:
– Upper 95 % CL = -0.313 Lower 95 % CL = -0.386
– t-Stat = -22.72 P-value <<< 0.05
• Coufort et al., 2007: x = -0.250
– Floc size in inertial and viscous sub-ranges.
• Shinnar, 1961 (and others): x = -0.250
– Coalescence controlled liquid-liquid dispersions.
• Calabrese et al., 1986 (and others): x = -0.400
– Break-up controlled liquid-liquid dispersions.
• Schulze et al., 2000 (and others): -0.250 > x > 0.400
– Dependent on balance between break-up and coalescence.

• Equilibrium floc size dependent on break-up and growth.

 WHAT DO WE MEAN BY SHEAR?
• Tilton, 2016, Private Communication:
– Shear (high or low) is generally used terminology:
• Conveys the general concept – (in)efficient disperser.
• Not strictly the best mechanistic description.
– Need Energy dissipation rate in region where desired mixing process is
controlled:
• ε̅ for “global” processes, e.g. blend time.
• εMAX for “local” processes, e.g. drop break-up, flocculation, micro and
mesomixing (using a dip pipe).
• High shear impeller → high UTIP and small wP.
• Define “Turbulent (Corrsin) Shear Rate”:

u′ k 1MAX
/2
1/3
ε 1 /3
MAX 1
γ C ~ = → u′
~ (ε MAXl0 ) → γ C ~ 2 / 3 [∝ ]
l0 l0 l0 τC
 d32 versus γC – D = ~ T/2
k 1max
/2

γC =
l0

d32 = 17305 γ C-1.2

RSD = ± 11.0 %