Chemical and Materials Engineering, University of Alberta

Results of Dynamic FBRM Measurements in a Confined Impelled Stirrer Tank (CIST) for O/W, W/O, Phase Inversion, and Bitumen

Dewatering Studies

Márcio B. Machado and Suzanne M. Kresta

marcio@ualberta.ca

Process Development & Scale-upJune 28, 2017– Houston, Texas

Introduction

• Mixing

– Multi-phase – liquid/liquid

• Scale-up/down

– Mixing sensitive problems

– What needs to be matched between scales?

– Scaling approaches

• Mixing devices

– Confined impeller stirred tank (CIST)

2

“What do you mean by Well Mixed?”

This is a well mixed CSTR.

Well mixed is not a clear definition, and perfectly mixed in a very short

time is only a safe assumption for perfectly miscible low viscosity fluids.

Coffee mug

3References: Paul et al. (2004); Kresta et al., (2015)

Technical Definition of Mixing

• Mixing has three physical dimensions (concentration, scale, and rate)

• It requires three design specifications:

– Uniformity of concentration

– A specified scale of segregation

– A required rate of mixing, or mixing time, which is often in competition with other rates.

4

Reference: Kukukova et al. (2009)

Identifying Mixing Problems

• Multiphase mixing involves many scales of mixing, frequently mass transfer, and complex physics. Oversimplifying is dangerous.

• Mixing frequently gets worse on scale-up.

• Competing rate processes are frequently sensitive to mixing. Design for local conditions.

5

Liquid/liquid dispersion

• Two (or more) immiscible liquids (or partially soluble)

• Function of turbulence

– Smallest droplets generated by turbulence are around the same size of the Kolmogorov scale

– Satellites drops smaller than Kolmogorov scale may occur

• Droplets are not necessarily spherical

• How to scale-up?

6

See papers by Calabrese or Tavlarides

Principles of robust scale-up

7

40 ft = 12 m, 10 ML

CIJR4 mm,1 mL

CIST120 mm,1 L

Requirements for robust scale-up

• Active flow in the entire vessel (no inactive regions)

– Entire vessel must be fully turbulent (transitional flow is not scalable)

– Regions of high local concentration are minimized (mesomixing)

– Single-impeller stirred tanks require Re > 100 000 to sustain fully turbulent flow at the top third part (Machado et al., 2013)

• Control the Kolmogorov scale

8

Classical scale-up - complications

• Simple scale-up cases

1. Geometric similarity

2. Constant property (power per volume, tip speed…)

3. Empirical design equations (Kawase and Moo-Young, Zwietering, Grenville) – See Paul et al. (2004)

This approach works well for many cases

• Some cases require more robust techniques

– Geometric similarity is not feasible

– Multiple process objectives

– Large scale-up factor

– Local mixing effects9

Sources of scale-up problems

• Reasons why bench scale tests sometimes do not scale

– Non-homogenous energy dissipation

– Different flow regimes at different scales

– Feed plume effects (mesomixing)

– Change of geometry

• Are bench scale test vessels well designed?

– Magnetic stir bars

– Shaker table

– Jar test

– Stirred tank10

Design objectives for new mixers

• Increase uniformity of turbulence over the volume of the vessel

• Allow a wider range of settings for energy dissipation (change impeller, N)

• Increase volume fraction of the mixer in fully turbulent flow

• Reduce mixer volume and eliminate stagnant or inactive zones

Often achieved by miniaturizing the vessel and/or confining the flow (Kresta et al., 2015)

11

CIST – Confined impeller stirred tank

• Fully turbulent to a much smaller Re

• Allows a 1L batch (or less) for additive testing

• Dramatic reduction in inactive volume (5% versus 30% in a stirred tank)

• Allows gravity separation test without material transfer

12

References: Machado and Kresta (2013), Komrakova et al. (2017)

CIST – Confined impeller stirred tank

• Baffled tank filled with 5 or 6 impellers

• High H/T

• Large impeller diameter (D=T/2 or 2T/3)

• Uniform energy dissipation

• Ret as low as 3 000 at the impeller region (20 000 for ST)

• V = 1 L (small footprint and less sample and less waste)

13

T = 7.6 cm

H = 3T

Source: Machado and Kresta (2013)

Mixing energy

• The total mixing energy (J, J/kg) in a mixing vessel is obtained from the total energy dissipated times the mixing time (Machado and Kresta, 2015)

J = ε*tmix

• It is an alternate scaling variable when geometric similarity is not feasible

• If a process is shear sensitive, ε can be decreased and J kept constant by increasing tmix

• Successful scale down from a pipeline to a CIST using J:

– Demulsifier addition tested in CIST and results scaled up to an industrial pipeline

– 50% reduction in chemical usage (Laplante et al., 2015; Chong et al., 2016a; Chong et al. 2016b)

14

Experimental

• Focused beam reflectance method (FBRM)

• Design of the confined impeller stirred tank (CIST)

15

FBRM

• Laser beam directed fluid from in situ probe.

• Backscatter reading.

• Measures chord length distribution (not diameter!).

16

Image from: Mettler Toledo

Instead of considering the difficulty of measuring spherical drops, perhaps we should consider the advantages of collecting a distribution of length scales, just as we do for turbulent eddies.

Chord Length: 0< cl<D …but not all drops are spherical

17

Image from: Mettler Toledo Image from: Kresta, HIM (2004)

Confined Impeller Stirred Tank

18

• CIST with a side port for the FBRM and top port for the PV19

• Extra port for feeding or sample withdraw

• Lid allows for control of air entrainment

• Extra layer of liquid may be added up to the top of the lid

• Several set of impellers can be used – different flow patterns and energy dissipation ranges

Confined Impeller Stirred Tank

19

• CIST operating with a FBRM and a PV19

Results

• Oil-in-water dispersion

• Water-in-oil dispersion

– Phase inversion

• Bitumen de-watering

20

Oil-in-water dispersion

• Canola oil (dispersed phase) in water

• CIST vs. stirred tank

• Different hold ups (10 % or 30 %)

• Liquid draw-down vs. impeller feeding

21

Experimental overview

22

• One set of fluids – no change in interfacial tension(Canola oil with red Leak-Detecting Dye (Kingscote Chemicals) dispersed into tap water)

• 2 Vessels (ST and CIST)

• 3 impeller geometries (RT, PBT and A310)

• In the CIST, H=3T and there are 5 impellers. For the PBT, three configurations tested (DDDDD, DUDUD, UDUDU)

• Feed is done from a surface layer, and at the impeller

• Visual observations (simple video), Chord length measurements (FBRM)

CIST vs. stirred tank

23

Stirred Tank (ST) and Confined Impeller Stirred Tank (CIST)with 10% and 30% hold-up

24

• Water – continuous phase (c)

• Canola oil – dispersed phase (d)

• Φ=10%

• Φ=30%

• Red Leak-Detecting Dye (KingscoteChemicals Inc.)

• No stabilizer

Fluid Specifications

25

Fluid Properties

Water Canola Oil

Interfacial tension

(, mN/m)20.51 ± 0.58

Density (, kg/m3) 1000 914

Viscosity (, Pa s) 1E-03 0.715E-03

Experimental Procedure 1

26

Tank setup per conditions

Liquids agitated by changing N: Tq at each N recorded

Nid determined and P and Re

calculated

First drop of oil separated

Experimental Procedure 2

27

Ncd measured, Pcd and Re calculated

Determine tcd at complete dispersion

Clean and refill the tank.

Set N = Ncd

• In ST, increasing hold-up increases Jcd, but in the CIST, increasing hold-up decreases Jcd

At the end of Part 1:

• Hypothesis: configuration with higher J should result in lower drop size

• Need to measure transient CLD using FBRM

Part 1: Energy for Complete Dispersion (J/kg)Normalized to base case – RT in ST at 10% (17.5 J/kg)8 geometries tested

28

from Bhalerao et al. (2015)

• Wanted to compare surface draw-down with impeller feed to determine if there is a difference in performance.

• Limited to 10% feed in the CIST so that the top impeller is always submerged.

• At this point, we could not do impeller feed in the ST because the required feed times are too short (flowrates are too high) for the pumps we have in the lab.

A new variable in Part 2: feed at the impeller

29

FBRM Probe Position

30

(a) Cumulative, (b) Number and (c) Square Weighted chord length distribution for surface layer CIST-PBT DU-30 at the beginning of mixing process (from t=0)

Evolution of chord length distribution with time for t<teq

31

(a) Cumulative, (b) Number and (c) Square Weighted chord length distribution

for surface layer CIST-DU-30 after equilibrium

Stable CLD for t>teq

32

Results

1. Evolution of mean cl : from completely dispersed (clcd) to equilibrium (cleq)

– Compare vessels, feed conditions, and hold-up

2. Compare the mean chord length to the Kolmogorov scale: where are we on the spectrum of eddies?

3. Scaling condition 1: Power (W/kg)

– power does not consider mixing time, so choose cleq

4. Scaling condition 2: Mixing energy (J/kg) which is (J=P*t/m)

– Better to use equilibrium time or completely dispersed time?

– Compare impeller geometries and vessels to group data

– Compare feed conditions

33

• The first and second dots are tcd and teq

• Most of the change in mean drop size happens during the initial dispersion (t<black dot).

• In fact, the change in square weighted cl from cd to eq is less than 10% for most cases tested.

• There is a lot of variation around the mean, even after teq

• What about other configurations?

Mean area weighted (cl32) chord length (ST-SL)

tcd

teq

Evolution of mean chord length

34

Looking in more detail at these two cases, we see that when there is surface feed with the RT, some of the large drops are never broken up.

Both of these cases are in the CIST.

Compare two cases more carefully: PBT-UD-10 and RT-10

0

50

100

150

200

250

300

0 500 1000 1500 2000 2500 3000 3500

cl3

2 (

mic

ron

)

Time (Sec)

SL-UD-10

IF-UD-10 PBT

mean

0

50

100

150

200

250

300

0 500 1000 1500 2000 2500 3000 3500

cl3

2 (

mic

ron

)

Time (Sec)

SL-RT-10

IF-RT-10RT mean

changes

with IF

0

20

40

60

80

100

120

140

160

1 10 100 1000

co

un

ts (

%)

Chord length (µm)

IF-DU-10-N

SL-DU-10-N

PBT

distribution

0

20

40

60

80

100

120

140

160

1 10 100 1000

co

un

ts (

%)

Chord length (µm)

IF-RT-10-N

SL-RT-10-N

RT distribution

Evolution of mean chord length

35

0

200

400

600

800

0 500 1000 1500 2000 2500 3000

cl3

2 (

mic

ron

)

Time (Sec)

RT-30UD-30DD-30A310-30

0

50

100

150

200

250

300

0 500 1000 1500 2000 2500 3000

cl3

2 (

mic

ron

)

Time (Sec)

UD-10DD-10DU-10RT-10

Once the upper impeller is submerged in the oil, there is a dramatic change in variability in the final mean chord length.

Mean cl: The effect of increasing hold-up in the CIST

CIST-10-SL CIST-30-SL

tcd

teq

tcd

teq

Evolution of mean chord length

36

• The range of cl32 (110m<cl32<500m) is much larger than the Kolmogorov scales (9.9m<<26m) and much smaller than the height of the trailing vortices (LI3.5mm) so the drops which dominate the square weighted distribution and the mean chord length fall in the inertial range.

• The number weighted distribution shows roughly half of the drops in the 1-10 micron range, suggesting that mechanisms of break-up are present which act at scales smaller than eddy break-up.

• By selecting cl32 for the scaling comparisons, we are focussing on the turbulent eddy break-up.

Chord length vs. Kolmogorov scale: Where do we fall on the spectrum?

0

20

40

60

80

100

120

140

160

1 10 100 1000

co

un

ts (

%)

Chord length (µm)

sample CLD

number square

37

• Power per mass seems to have little effect on the drop size over the range tested, from 0.1W/kg to 2W/kg.

• Note that this is the power needed to just disperse the liquid, so the range is smaller than we would select for a study of drop break-up.

• Fitting a line to the data on a log-log plot gives an exponent of -0.01.

Scaling with P/m at teq

10

100

1000

0.1 1 10cl

32

-eq

(m

icro

n)

Power/m (W/kg)

10%-CIST

10%-IF

10%-ST

30%-ST

LINE

slope=-0.01

RT-IF

38

• The range of Jcd is wider, from 20-267J/kg

• The slope of most of the data falls on one line, giving an exponent of -0.08 (eight times larger than the power result).

• Using tcd rather than teq brings the RT-IF point into line with the other configurations.

• For the one case where an impeller is submerged in the surface layer, the exponent is dramatically different, at -1.4.

Scaling with Jcd at tcd and clcd

10

100

1000

10 100 1000

cl3

2-c

d (

mic

ron

)

Jcd / m- (J/kg)

10%-CIST10%-IF-CIST30%-CIST10%-ST30%-STSeries5

slope=-0.08

slope=-1.4

30%-CIST

39

10

100

1000

10 100 1000

cl3

2-c

d (

mic

ron

)

Jcd / m- (J/kg)

RTPBT-UDPBT-DUPBT-DD

slope=-0.08

slope=-1.4

• Grouping the data by impeller leads to the same conclusion: for all impeller configurations, the results change dramatically if one impeller is submerged in the surface layer.

• This result is in close alignment with the literature on the effect of geometry on phase inversion during emulsion formation.

Jcd scaling grouped by impeller

30%-CIST

40

Conclusions – oil-in-water dispersion

• The CLD shows that almost half of the drops are smaller than the Kolmogorov scales, while the weighted mean falls conveniently inside the inertial range. More work is needed on the mechanisms which lead to formation of these very tiny drops.

• At the just-dispersed condition, cl32 scales well with Jcd , in contrast with

– Ncd and tcd which showed no trends

– energy required to completely disperse the second phase, which showed conflicting trends depending on geometry

– P/m at equilibrium, which showed no effect on cl32

• The hypothesis that mixing energy is a useful scaling variable is upheld for the moment.

41

Water-in-oil dispersion

• Number averaged droplet counts

• Effects of the position of the liquid interface relative to the impeller blade

42

Interface at 2mm below lowest impeller

Interface at 2mm above lowest impeller

Continuous Phase Canola Oil Canola Oil

Dispersed Phase Water Water

Water Hold-up 4.77 wt% 9.91 wt%

Average Energy Dissipation, ɛ (W/kg)

0.25 to 4 0.5 to 2

9

Experimental Conditions

43

Number averaged droplet counts –experiments with 4.77wt% of water

44

- Maximum is at the limit detection of the equipment

- Kolmogorov scale is larger than 15 µm for all cases

- What is generating these small droplets?

Dispersion of 4.77 wt% of water in canola oil

• Interface: 2mm below the lowest impeller

• Lower energies do not have peaks, but it takes longer for them to get to equilibrium 10

45

Comparing the position of the interface

46

• Energy dissipation of 2 W/kg

• Peak is higher when the interface is located above the impeller

Dispersion of water (10 wt%) in canola oil

47

• Most part of the readings are on the order of 2-3 µm (smaller than Kolmogorov scale) – limit detection of the equipment

• The initial dynamics of the dispersion are effected not only by the hold up, but also by the placement of the interface relative to the Rushton impellers.

• At the initial stages of the dispersion, the continuous phase (canola oil) may be pumped towards the dispersed phase (water).

• So canola oil may be initially dispersed into water.

12

Conclusions – water-in-oil dispersion

48

Bitumen de-watering

• Bitumen contains water and solids that need to be removed

• Chemical demulsifier is used for that

– Demulsifier concentration

– Mixing (energy dissipation and mixing time)

– Injection concentration (pre-mixing, demulsifierdilution)

49

Experimental Setup

50

Sampling

Sampling Point

Height Below Interface

Relative Height

FBRM 30 mm 0.13

Z1 52 mm 0.23

Z2 96 mm 0.43

Z3 140 mm 0.62

Z4 184 mm 0.82

rendering by Jairamdas, 2016 51

liquid height = 225 mm

Main results

52

- Demulsifier concentration is the most important variable as expected

- Correct levels of mixing energy (energy dissipation x mixing time) can cut the amount of demulsifier by half.

- Right mixing conditions also improve settling

- Pre-dilution of the demulsifier (lowering injection concentration) significantly improves performance –avoid high local concentration

FBRM Fouling

Conclusion: Fouling must be accounted for in LQ froth

53

Sapphire repellency treatment (Aculon) is very effective

FBRM – Concluding remarks

- The FBRM measures chord length instead of drop size. This seems physically meaningful when placed in the same frame as the spectrum of length scales in turbulent flow

- Air entrainment is a major factor that has to be carefully addressed

- We tested several fat soluble fluids as continuous phase –Conosol, Silicon and canola oil

- Canola oil provides the most stable results

- We are trying n-heptane and mineral oil

- Saphire repellency treatment was very effective for bitumen experiments

54

Acknowledgments

Several people worked on the results: Dr. John Nychka, Runzhi(Anna) Xu, Colin Saraka, Akshay Bhalerao, Fatemeh Safari Alamuti, Tanya Varma, Khilesh Jairamdas, Alexandra Komrakova

55

Handbook of Industrial Mixing & Advances in Industrial Mixing

56

References• Bhalerao, A., Alamuti, F.S., Kresta, S.M., Machado, M.B., Komrakova, A.E., 2015. Liquid Drawdown in a Stirred Tank and Confined Impeller Stirred Tank (CIST),

AIChE annual meeting Salt Lake City, Utah, USA.

• Calabrese, R.V., Chang, T.P.K., Dang, P.T., 1986a. Drop breakup in turbulent stirred-tank contactors. Part I: Effect of dispersed-phase viscosity. AIChE Journal 32, 657-666.

• Calabrese, R.V., Wang, C.Y., Bryner, N.P., 1986b. Drop breakup in turbulent stirred-tank contactors. Part III: Correlations for mean size and drop size distribution. AIChE Journal 32, 677-681.

• Chong, J.Y., Machado, M.B., Arora, N., Bhattacharya, S., Ng, S., Kresta, S.M., 2016a. Demulsifier Performance in Diluted Bitumen Dewatering: Effects of Mixing and Demulsifier Dosage. Energy & Fuels 30, 9962-9974.

• Chong, J.Y., Machado, M.B., Bhattacharya, S., Ng, S., Kresta, S.M., 2016b. Reduce Overdosing Effects in Chemical Demulsifier Applications by Increasing Mixing Energy and Decreasing Injection Concentration. Energy & Fuels 30, 5183-5189.

• Coulaloglou, C.A., Tavlarides, L.L., 1976. Drop Size Distributions and Coalescence Frequencies of Liquid-Liquid Dispersions in Flow Vessels. AIChE Journal 22, 289-297.

• Coulaloglou, C.A., Tavlarides, L.L., 1977. Description of Interaction Processes in Agitated Liquid-Liquid Dispersions. Chemical Engineering Science 32, 1289-1297.

• Komrakova, A.E., Liu, Z., Machado, M.B., Kresta, S.M., 2017. Development of a zone flow model for the confined impeller stirred tank (CIST) based on mean velocity and turbulence measurements. Chem Eng Res Des submitted.

• Kresta, S.M., Etchells, A.W., Dickey, D., Atiemo-Obeng, V.A., 2015. Advances in Industrial Mixing: A Companion to the Handbook of Industrial Mixing. Wiley.

• Kukukova, A., Aubin, J., Kresta, S.M., 2009. A new definition of mixing and segregation: Three dimensions of a key process variable. Chemical Engineering Research and Design 87, 633-647.

• Laplante, P., Machado, M.B., Bhattacharya, S., Ng, S., Kresta, S.M., 2015. Demulsifier Performance in Froth Treatment: Untangling the Effects of Mixing, Bulk Concentration and Injection Concentration using a Standardized Mixing Test Cell (CIST). Fuel Processing Technology 138, 361-367.

• Machado, M.B., Bittorf, K.J., Roussinova, V.T., Kresta, S.M., 2013. Transition from turbulent to transitional flow in the top half of a stirred tank. Chemical Engineering Science 98, 218-230.

• Machado, M.B., Kresta, S.M., 2013. The confined impeller stirred tank (CIST): A bench scale testing device for specification of local mixing conditions required in large scale vessels. Chem Eng Res Des 91, 2209-2224.

• Machado, M.B., Kresta, S.M., 2015. When Mixing Matters: Choose Impellers Based on Process Requirements. Chemical Engineering Progress 111, 27-33.

• Paul, E., Atiemo-Obeng, V.A., Kresta, S.M., 2004. Handbook of Industrial Mixing: Science and Practice. John Wiley and Sons, Inc., Hoboken, New Jersey.

• Tsouris, C., Tavlarides, L.L., 1994. Breakage and coalescence models for drops in turbulent dispersions. AIChE Journal 40, 395-406.

• Wang, C.Y., Calabrese, R.V., 1986. Drop breakup in turbulent stirred-tank contactors. Part II: Relative influence of viscosity and interfacial tension. AIChEJournal 32, 667-676.

• Zhou, G.W., Kresta, S.M., 1996. Impact of tank geometry on the maximum turbulence energy dissipation rate for impellers. AIChE Journal 42, 2476-2490.

57

Extra slides

58

Video – DUDUD PBT in CIST at φ = 10%first 27s of 270s

59

Small oil droplets and large droplet in background

PV19 images – oil-in-water dispersion

60

Oil droplets in background and foreground

PV19 images – oil-in-water dispersion

61

Small droplets on surface of larger droplet

PV19 images – oil-in-water dispersion

62

Complex droplets

PV19 images – oil-in-water dispersion

63

Flow field• CIST operating

with 5 Rushton impellers

• Same flow behaviour in all jets leaving the impellers

• 5 recirculation zones between impellers

64

Power per volume comparison

• When Rushtonsare used, CIST is 13 times more homogeneous

• When A310s are used, CIST is 6 times more homogenous

• # of impellers forces the energy dissipation to be well distributed

65

Ϯ Zhou and Kresta (1996)

Impeller *εmax/εaverage

for the CISTεmax/εaverage

for stirred tankϮ

Rushton 10.3 138

A310 7.8 46

Intermig 4.8 N/A

* Impeller sketches: Hemrajani and Tatterson in Paul et al., 2004.