Inline Bitumen Emulsification Using Static Mixers

Inline Bitumen Emulsification Using Static Mixers

Jean-Philippe Gingras,† Louis Fradette,† Philippe Tanguy,*,† and Jacques Bousquet‡

URPEI-TOTAL Chair, EÄ cole Polytechnique, Montreal, Canada H3C 3A7, and TOTAL, BP 22, 69360Solaize, France

Highly concentrated bitumen-in-water emulsions were produced with static mixers in continuous mode. Theeffect of the following process parameters on the average droplet size was studied: emulsion flowrate, staticmixers configuration, and surfactant concentration as well as bitumen hardness, concentration, and temperature.Several static mixer configurations were investigated consisting of combinations of SMX (Sulzer ChemtechLtd.) and helical elements with empty sections. The drop size results revealed that the mean droplet sizecould be scaled with the energy or the power draw depending on the static mixer configuration. Moreover,it was shown that the energy draw could capture the effect of emulsion flowrate and bitumen concentrationon the mean droplet size, whereas the power draw (or the specific power) captured the effect of the emulsionflowrate and bitumen temperature. It was also demonstrated that the specific energy, or pressure drop, wasminimized when SMX mixers were inserted after helical mixers.

Introduction

Bitumen emulsions for road surfacing are usually producedwith rotor-stator technologies in continuous mode. Theseemulsions are stabilized oil-in-water dispersions that are liquidat ambient temperature and can be applied by means of a varietyof cold methods. The typical dispersed phase content in theemulsion is 55-70%, and the average droplet size rangesbetween 5 and 50µm, yielding a product with a viscosity lessthan 0.1 Pa‚s particularly easy to manipulate as a pavingmaterial. A challenge related with the production of modernbitumen emulsions is to reach a droplet size of the order of 1µm or less without changing the emulsion formulation.

Gingras et al.1 conducted experimental trials on a bench-scaleunit equipped with an inline rotor-stator and an industrialformulation following the practice in the road industry. Rotor-stator devices are composed of a rotating blade impeller ortoothed crown enclosed in a static cage with small gaps, typicallya slotted cylinder. The results revealed that the droplet size waslimited by the short residence time, which promoted recoales-cence. The coalescing phenomenon was particularly pronouncedwhen the residence time in the dispersing zone was of the sameorder of magnitude as the droplet breakup time and the surfactantadsorption time. The study illustrated the lack of knowledge ofthe hydrodynamics in inline rotor-stator geometries, explainingwhy the droplet size calibration and the process scaleup are oftencarried out by trial-and-error in industry.

To overcome these difficulties, Gingras et al.2 developed aninnovative approach in which a high-internal-phase-ratio (HIPR)emulsion is processed in a coaxial mixer. HIPR emulsions arecharacterized by a network of polyhedral droplets.3 Thisunconventional drop pattern confers viscoelastic and shear-thinning behaviors not seen with classical emulsions. HIPRbased emulsification processes are particularly suitable toproduce small droplets with conventional mixing technology,because their high effective viscosities during emulsificationallow reaching high shear stresses without having to applyextremely large shear rates. This also implies that the viscous

forces are dominant leading to a process operated in the laminarregime. The approach elaborated by Gingras et al.2 consists oftwo steps, namely the formation of the right droplet size bymixing a HIPR emulsion at an optimized composition and awater dilution stage at a concentration allowing emulsionpumpability. The HIPR emulsion is formed by gradually addingbitumen to an aqueous phase, followed by mixing during a fewminutes to reach the desired droplet size range. By comparisonwith the conventional bitumen emulsification process,4-6 thisapproach is characterized by a longer residence time in thedispersing zone, a higher dispersed phase concentration (duringemulsification), and lower shear rates. It should be mentionedthat an important limitation of stirred tank technologies is thesegregated zones that may appear within the dispersion volume.These flow anomalies are a source of nonuniform dropletdistributions throughout the vessel volume, which may yield toa polydisperse end-product and make the process scaleupdifficult. Due to their complex rheology, HIPR emulsions areparticularly prone to segregations and cavern formation.7

It is clear from the above that a compromise must beestablished that combines the advantage of an inline processwhere it is possible to emulsify in a smaller working volumeper unit time and the stirred vessel approach with lowhydrodynamic severity. Static mixers may be used for thispurpose. Indeed, not only do they offer the same advantages interms of space requirement and throughput than rotor-stators,but the added benefits include a shear level significantly lower,8

a larger dispersion volume, and well-established scaleup pro-cedures.9 A wide variety of geometries are available as reportedin a recent review paper on static mixers.10

The objective of this paper is to investigate whether it ispossible to obtain a HIPR bitumen emulsion in the micrometerrange with static mixers in continuous mode and control thedroplet size. Before describing the experimental approach thatwe developed, it seems appropriate to review briefly theliterature on HIPR emulsions in order to put in evidence themain factors governing the droplet size.

The classical approach11,12 to predict the droplet size inlaminar flow, based on the critical capillary number, is notconsidered in this paper because it relies on several conditions(constant deformation forces, absence of droplet interaction,initial spherical shape, etc.) that are quasi-impossible to obtain

* To whom correspondence should be addressed. E-mail:[email protected]. Fax: 514-340-4105.

† EÄ cole Polytechnique de Montre´al.‡ TOTAL.

2618 Ind. Eng. Chem. Res.2007,46, 2618-2627

10.1021/ie0611913 CCC: $37.00 © 2007 American Chemical SocietyPublished on Web 03/17/2007

in a real emulsification process characterized by complexhydrodynamics. Moreover, phenomena related to HIPR emul-sification that are still not fully understood (droplet breakupmechanism and dynamic interfacial tension) are required toapply this approach. Therefore, the review will focus on studiesconducted on a scalable emulsification process aiming to identifythe key factors affecting the droplet size.

The literature on inline HIPR emulsification is scarce. Tothe best of our knowledge, no scientific publication focuses oninline HIPR emulsification based on static mixers with theexception of patents. Catafalmo et al.13 described a two-stagestatic mixer process for HIPR emulsions. The first stage of staticmixers is required to premix the immiscible phases while theaverage droplet size scales with the shear rate in the secondstage. The examples given in the patent report the use of SMXelements (Sulzer Chemtech Ltd.), but the inventors suggest thatany commercially available static mixers can be employed forboth stages. For the first stages, they recommend to keep theratio between “the active surface” of the static mixer and theflowrate constant. For the second stage, the following relation-ship is proposed:

In a static mixer, the average shear rate corresponds to14

whereC3 depends on the geometry of the static mixer and canbe determined with the Metzner and Otto approach.15,16

A few articles have been published on the droplet sizecalibration in a larger dispersion volume for HIPR emulsion,mostly in stirred vessels. A detailed summary of these studiesis given elsewhere.2 The objective herein is to review theparameters and models proposed to predict the droplet size.Briceno et al.17 suggested the following relationship when theHIPR emulsion was formed in a tank equipped with a singleRushton turbine:

It was shown thatC4 decreases as the dispersed phaseconcentrationæ increases. In semibatch mode, Gingras et al.2

found that the deformation based on the average shear rate ofthe close-clearance impeller in a coaxial system is the governingparameter. A correlation that takes into account the effect ofthe deformation and the composition was proposed:

It should be noted that the deformation as defined in eq 4depends onæ. Alternatively, Gingras et al.2 also linked themedian diameter with the energy dissipated by the close-clearance impeller for a constant mass of emulsion, using thefollowing:

As for eq 4, the surfactant mass fraction was included alongwith dissipated energy to increase the correlation accuracy.

Adler-Nissen et al.18 developed a close-clearance agitationgeometry to study the effect of rotation speed on the dropletsize in mayonnaise sauce (a HIPR emulsion) operated in batch

mode. Several experiments were performed with the sameagitation time yielding to the following model:

The emulsification studies for inline production of standardemulsions (non-HIPR) in laminar flows can also be helpful toidentify the governing factors that control the droplet size.Karbstein and Schubert19 combined the specific power and theresidence time in the dispersing zone to establish the meanspecific energy (E) in a rotor-stator device:

which was related with the average droplet size in laminar flowby

where C6 depends on the dispersed and continuous phaseviscosities, the interfacial tension, and others parameters notcaptured by the specific energy.

The limited number of publications on this topic includesthe work of Grace14 in which several liquid-liquid dispersionsexperiments at low dispersed phase concentration with Kenicsstatic mixers were conducted. The numerous results presentedgraphically by Grace can be summarized by the followingequations:

whereµc is the continuous phase viscosity.

Grace highlighted the different behavior of dispersions of low(η < 10-1) and high (η > 5) viscosity ratios. At lowη, inagreement with drop breakup models in quasi-static flow,a )1 was found, whereas, forη > 5, a ) 0.5. These exponentswere determined by modifying the flowrate (Q) on the samestatic mixer configuration. Grace also noticed that the lengthof the static mixer and relaxation zone between mixers affectedthe mean droplet size. To keep using eq 9 for the prediction ofd, C7 has to be estimated for each configuration of static mixers.For η > 5, he observed that the experimental points collapsedon a master curve whend was plotted versus the pressure dropin the static mixer of various lengths without a relaxation zoneand that eq 10 was more appropriate withb ) 0.5. Grace foundthat the relaxation zone led to smaller droplets for the sameflowrate or pressure drop, i.e.,C7 or C8 decreased for staticmixer assemblies with relaxation zones.

The effect of process parameters on the droplet size in laminarflow was also studied by Fradette et al.20 for SMX elements.The following equation was proposed:

where the energy dissipated was computed by

It was found thatC9 increased asη decreased andæ increased.Interestingly, the master curve obtained by Grace with pressure

dmed) 11 γ-0.66 (6)

E ) εj × thR ) PQ

(7)

d )C6

E(8)

d ) C7( σµcγ)a

(9)

d ) C8∆p-b (10)

d ) C9E-0.25 (11)

E ) ∆pπD2L4

(12)

d ) C1 +C2

γ(1)

γ ) C3VD

) C34Q

πD3(2)

d32 ) C4(æ)E-0.2 (3)

dmed) C5γcorr-0.62yS

-4.8 (4)

dmed) C6E-0.42yS

-4.2 (5)

Ind. Eng. Chem. Res., Vol. 46, No. 8, 20072619

drop (eq 10) as the predicting variable was more accurate (R2

is larger) when the energy dissipated was taken as theindependent variable. The exponent obtained in this case was-0.6.

The good fit obtained by Grace14 and Fradette et al.20 betweenthe average droplet size and the energy shows that the approachof Karbstein and Schubert19 for rotor-stators is also relevantfor static mixers in laminar flows. This finding is contradictorywith the dependence between the average droplet size and theshear rate reported for HIPR emulsification with static mixers.13

Therefore, many scenarios are conceivable to control the dropletsize in HIPR emulsification with static mixers.

Methodology

The experiments were carried out with a single formulation.The dispersed phase was composed of various batches ofbitumen PG 64-22 (McAsphalt, Canada). Although the bitumengrade was kept the same, the analysis of the batches revealeddifferences likely originating from variable sources of crude oilat the refining stage. The samples were classified in twocategories, referred to as i and ii. Bitumen ii was harder (moreviscous) than bitumen i at room (and the emulsification)temperature. The continuous phase, also called the “soap”,comprised a fatty polyamine surfactant (alkoxylated alkyl(ene)polyamines) provided by CECA (France), tap water, andhydrochloric acid to activate the surfactant hydrophilic head.The final pH of the soap was adjusted between 0.8 and 1. Thesurfactant concentration was estimated by assuming that themass of surfactant was equal to the mass of fatty polyamineadded during the soap preparation, i.e., the water concentrationincluded the tap water and the hydrochloric acid added duringsoap preparation. The concentrations of the three components(bitumen, water, and surfactant) expressed on a weight basisare given in Table 1.

An experimental setup allowing operation in a fully continu-ous mode was designed and built (Figure 1). It comprised threeparts: the bitumen heater, the soap preparation process, andthe emulsification section. As it can be seen in Figure 1, thebitumen and the soap sections can be operated in loop mode.For this purpose, a progressive cavity pump (PCM Moineau)is installed in the soap section, while a gear pump (PulsafeederInc.) is used for the bitumen. Between the pump and the outletof the loop, a Coriolis mass flowmeter (Micro Motion) and athree-way valve are installed in sequence. The three-way valvegives the option of feeding the soap or the bitumen back to thestorage tank or the mixing section. The emulsification takesplace in the mixing section by means of static mixers insertedin “mixing cells” where the bitumen and the soap are fedsimultaneously. This mixing section is designed with sixconsecutive identical cells in which a static mixer assembly (anassembly is composed of several elements) can be inserted. Thedimensions of a mixing cell are indicated in Figure 1. Two typesof static mixers were selected (Figure 2), namely helicalelements (Statiflo International) and SMX elements (Sulzer).The helical element has a geometry similar to the well-knownKM elements from Chemineer Inc. Each SMX is composed ofsix elements having a length equal to its diameter (L ) D )

15.8 mm), while the helical one has four elements with a lengthlarger than its diameter (L/D ) 1.4). The pressure drop wasmonitored by means of two pressure transducers (OmegaEngineering) installed upstream and downstream the mixingcells. The bitumen and mixing section piping, the mixing cells,and the bitumen tank were all provided with heating systems.In our experiments, the emulsification temperature dependedmainly on the bitumen temperature because there was no heatingsystem in the soap loop. To prevent water evaporation, weensured that the temperature at the outlet of the mixing sectiondid not exceed 100°C, as no heat exchanger was installed beforethe outlet of the mixing section. The pressures and flowratesmeasurements were recorded with a data acquisition systemduring the bench-scale unit operation allowing the pressure drop(∆p) and the emulsion flowrate (Qe) measurements by

The total length (L) of static mixer was estimated by

whereNSMX andNH are respectively the number of SMX andhelical elements.

Each trial was started with the heating of bitumen until thedesired bitumen temperature was obtained. To feed the mixingsection, the ratio between the bitumen (QB) flowrate and the

Table 1. Emulsion Composition Ranges Investigated

weight fractions

min max

bitumen (xB) 0.81 0.94surfactant (xS) 0.017 0.053water 0.04 0.13

Figure 1. Schematic representation of the bench-scale unit.

Figure 2. Photograph of the SMX and helical static mixers tested in thiswork.

∆p ) pin - pout (13)

Qe ) QB + QS (14)

L ) NSMXD + 1.4NHD (15)

2620 Ind. Eng. Chem. Res., Vol. 46, No. 8, 2007

total flowrate (QB + QS) was first set at 0.75( 0.05 to avoidhigh initial pressure peaks and the backflow of bitumen in thesoap section. The flowrates were then adjusted at a desired setpoint by modifying the rotational speed of the pumps in orderto increaseQB/QS gradually. Once a steady-state operation wasreached, a sample of about 500 g of concentrated emulsion wastaken in a one-L container. Tap water at 60°C was addeddirectly in the container to dilute the bitumen emulsion to theconcentration commonly used in road surfacing applications.This dilution was performed in batch mode with a commercialhand mixer at low rotational speed. It was verified that thedilution step did not influence the average droplet size (resultsnot shown). The negligible effect of the dilution step on thedroplet size had also been noticed in a coaxial mixer bench-scale unit.2

Few of the drops of the diluted emulsion were dispersed ina solution slightly above the critical micellar concentration.These highly diluted emulsions were analyzed with a MastersizerS (Malvern Instruments) to obtain the droplet size distribution(DSD). Many average diameters can be extracted from the DSDmeasurement. In this article, we selected the volume mediandiameter (dmed) as the basis for the average droplet diameter.The termdmed is the diameter that corresponds to 50% of thevolume cumulative distribution and is a reference mark widelyused among bitumen emulsion practitioners. The polydispersityof the distributions was quantified by the ratio between the meanvolume-weighted (d43) and surface-weighted diameters (d32). Itshould be noted that repeated measurements from the samesample led to DSD values with very small deviation. We alsoverified that the delay between the sampling and the analysishad also no influence on the DSD.

Results

The effect of six experimental parameters (Table 2) on theemulsion droplet size was investigated for a total of 74 runs.Figure 3 shows the mixing cells configuration (MCC) testedfor various values ofQe andxB (other experimental parameterswere initially kept constant). These MCCs were chosen amongthe 36 ) 729 possibilities (a cell could be empty or occupiedby a SMX or a helical mixer). Many configurations werediscarded due to the insufficient deformation forces created atthe maximum HIPR emulsion flowrate allowed on the setup.A MCC in which a helical mixer is inserted in each cell (a totalof 24 elements) is an example of such an inadequate MCC:for Qe ) 350 kg/h andxB ) 0.9, an average droplet size of 25µm was obtained. It was established that at least two SMXmixers (12 elements) have to be inserted in an MCC in orderto fall in an acceptable range ofdmed. These observations are inagreement with a CFD work in which key hydrodynamicparameters were computed for SMX and helical mixers geom-etries. Rauline et al.8 showed that the average shear rate providedby a SMX element is twice that of a helical type. Simulatingviscous fluids flowing in the laminar regime, they also obtainedthat, respectively, 2 and1/2 and 4 helical elements have to beused to reach the same level of distributive mixing and stretchingof a single SMX element. The configurations shown in Figure

3 were selected to highlight the effect of the number of elementsand the combination of helical and SMX mixers on the dropletsize. The MCC G for which most of the runs were conductedwill be taken as the reference MCC in the result analysispresented below.

The influence ofQe on the mean diameter was captured bythe pressure drop measurement. For the range ofQe and xB

reported in Table 2, the energy dissipated in the static mixers(evaluated with eq 12) is plotted versusdmed in Figure 4 forMCC A-J. The dashed line in Figure 4 is fitted on the data ofreference MCC with a power trend showing thatdmeddecreasesasE increases. Although the range ofxB values investigated israther large, the experimental points collapse on the same curvefor the reference MCC. This result shows that, for a given cellconfiguration, the average droplet size can be controlled by theemulsion flowrate and the bitumen concentration via the energydraw (or the pressure drop) measurements. However, the valuesof dmed obtained for a given level of energy for two differentMCCs are not necessarily similar. AtE ≈ 175 J, a plateau isreached even if more SMX elements are added to increase theenergy dissipated. When the number of SMX elements (NSMX)is significantly higher that of the helical elements (NH), e.g.,MCC I and J, larger values ofdmed are obtained for the same

Table 2. Range or Levels of Experimental Parameters Investigated

experimental parameters range or levels

emulsion flowrate (Qe) 90 - 370 kg/hemulsion bitumen concentration (xB) see Table 1mixing cells configuration (MCC) A-Jbitumen hardness i and iisoap surfactant concentration (yS) 26, 29, and 31 (wt %)bitumen temperature (TB) 90 and 97°C

Figure 3. Mixing cells configurations (MCCs) investigated.

Figure 4. Influence of the energy dissipated (E) on dmed for ten mixingcells configurations: fitting curve on MCC G (+).


energy consumption, by comparison with the cases where thereare two or less SMX elements per helical element in the mixingcells, e.g., MCC C and G.

The same set of experimental data is plotted versus the powerdraw in Figure 5. The power draw in the static mixer has beenestimated by∆pQ. It can be seen in Figure 5 that the dropletsize diameter decreases as the power is increased. The datacorresponding to the configurations without empty cells (MCCC and MCC G-J) collapse on the same curve, and the diametersobtained fall below the experimental points of the otherconfiguration for which larger values ofdmed are obtained forthe same power level.

Figure 6 shows that the polydispersity reaches a plateau whenE > 100 J. As for the effect of the power ondmed, the MCCswithout empty cells collapse on the same curve when thepolydispersity is plotted versusE or P (see also Figure 7). TheMCC with empty cells characterized by a lower range ofEyielded larger polydispersity. For the same level of energy, itis interesting to see that configurations A, D, and E led todistributions with lowerdmed values and larger polydispersityby comparison with the reference configuration.

Given that the maximum outputs of the pumps were reachedin our trials, the results indicate that the design of mixing cellslimits the median droplet size to around 1.7µm. The cause ofthis limitation will be discussed below.

The effect ondmed of the parameters that were kept constantin the results presented above, i.e., bitumen hardness, soapsurfactant concentration, and bitumen temperature, could be seenin Figures 8and 9. For the same level of energy or powerdissipated (obtained with various combinations ofQe and xB

on the reference MCC), the emulsification of harder bitumenleads to larger droplets and the effect ofyS corresponds to thesame findings as with the stirred tank emulsification,2,21 i.e.,

Figure 5. Influence of the power dissipated (P) on dmed for ten mixingcells configurations: fitting curve on MCC G (+).

Figure 6. Influence of the energy dissipated (E) on the polydispersity forten mixing cells configurations: fitting curve on MCC G (+).

Figure 7. Influence of the power (P) on the polydispersity for ten mixingcells configurations: fitting curve on MCC G (+).

Figure 8. Influence of the energy dissipated (E) on dmed for differentcombinations of bitumen hardness (i or ii), soap surfactant concentration(yS), and bitumen temperature (TB) on MCC G: fitting curve on first series(+).

Figure 9. Influence of the power dissipated (P) on dmed for differentcombinations of bitumen hardness (i or ii), soap surfactant concentration(yS), and bitumen temperature (TB) on MCC G: fitting curve on first series(+).


increasing yS allows the formation of smaller droplets. For theeffect ofTB, the gap between 97°C (lozenges) and 90°C (“X”symbols) for the same bitumen hardness andyS ) 0.28 is notas significant as the previous parameters especially in Figure9. An explanation for this finding is given below. It was alsofound that the bitumen hardness,yS and,TB had no significanteffect of the polydispersity (results not shown).

The above results seem to indicate that either the energy orthe power could be used as a variable to predict the dropletsize for a given MCC. However, it is not clear whether theyare the unique factors or other governing factors like the specificenergy; the shear rate could be considered to predict the averagedroplet size. To address these issues adequately, a statisticalanalysis has been carried out.

Average Droplet Size Modeling

The modeling strategy employed is described in Figure 10for the 53 trials in which the bitumen hardness,yS, andTB havebeen considered as constant. In a first attempt,dmedis correlatedwith three variables: the governing factor candidate, the MCCs,and thexB. Afterward, the model obtained is analyzed to verifyif the variables are statistically significant. If an MCC orxB isnot significant, we can infer that the variable is captured by thegoverning factors. The best-case scenario would be that agoverning factor includes the effect of modifying the MCC,xB,and other critical process parameters on the average droplet size.

In addition to energy and power, four other governing factorcandidates are considered based on models proposed in theliterature to predict the droplet size. They are all listed in Table3. Because most of these parameters are extracted from studyinvolving mechanical agitation technologies, the second columnshows the particular expression of these parameters for staticmixers. Four of these parameters can be estimated with thepressure drop measurements. The pressure drop in a pipe (withor without static mixers) depends on the friction factor (f). Inlaminar flow,f is inversely proportional to the Reynolds number(Re). Therefore, the product off andReyields a parameter thatdepends on the static mixer geometry. This parameter can beexpressed in a dimensionless form by the following:

The replacement of∆p by the ratio deduced from eq 16 givesthe dependence of the listed governing parameters on the staticmixer geometry and dimensions (D, L, and KP), the total

emulsion flowrate (Q ) Qe), and the effective emulsificationviscosity (µ ) µeff). These relations are given in the third columnof Table 3.

The governing factor candidates for which the estimation isnot based on pressure drop measurements (γ andγ) depend onthe geometrical constant of the static mixerks. The ks valuescommonly found in the literature22 for SMX and helical elementsare 40 and 22. If a single type of mixer is installed in the mixingcells, γ and γ can be easily determined. When both types ofmixers are combined in the mixing cells, the shear rate isestimated as the maximum shear rate

if at least one SMX mixer is inserted in the mixing cells and

otherwise.The deformation is estimated as the total deformation caused

by each element. By combining the relationship of Table 3 andeq 15,γ is evaluated by

The range of each governing factor candidate is reported in thelast column of Table 3.

A covariance analysis had to be performed, because MCCwas a categorical factor andxB and the governing factor werecontinuous variables. A covariance model23 combines categoricalfactors and quantitative variables to predict the quantitativeresponse variable. For the conditions described above, thecovariance model is written as

whereâ0 is the overall mean andâ1 andâ2 are the regressioncoefficients.

The indicesi, j, andk stand for theith level of MCC (betweenA and J), jth observation of theith level, andkth governingfactor listed in Table 3. The logarithmic form of the quantitativevariables was chosen because most models and scaling laws towhich our results will be compared have power law expressions.Moreover, a good fit was obtained with a power law trend inFigures 4 and 5. A crucial assumption of covariance models isthe constant slopes for each level of the categorical factors. Foreq 20, this means thatâ1 andâ2 are similar for each MCC. Thecovariance analysis for the model of eq 20 was performed witha statistical software package (STATISTICA). The covarianceanalysis of eq 20 was performed successively for each governingparameter with the strategy of Figure 10, and Table 4 highlightsthe key results. The effect of MCC could also be seen in Figure11 in which thedmed was evaluated (with the 95% confidencelimit) for the average governing factor andxB for eachconfiguration.

The best-case scenario mentioned above was not obtained inthis work, because no governing factors captured the effect of(all) MCCs andxB on dmed. On the other hand, for a givenconfiguration, it was shown that the energy (E) or the specificenergy (E) could be advantageously taken as governing param-eters, because they captured the effect ofxB and yielded the

Figure 10. Strategy for the identification of the key factors affecting themedian droplet size.

KP ) D2∆pµVL

) πD4∆p4µQL

(16)

γ ) 404Qe

πD3(17)

γ ) 224Qe

πD3(18)

γ ) 45NSMX + 22× 1.4NH (19)

[ln(dmed)] ij ) â0 + MCCi + â1([ln(GFk)] ij - ln(GFk)) +

â2([ln(xB)] ij - ln(xB)) (20)


highestR2. The regression coefficientâ1 (or log-log slope) ofE or E obtained in this work can be compared with those foundin the literature (Table 5). For the same formulation of thebitumen emulsion, the coefficient found with static mixers ishigher than the one obtained in the coaxial mixer. Thediscrepancy between the values reported in Table 5 can be

attributed to the fact that several factors related with theemulsification process affect the exponent of the energy. Despitetheir slightly lowerR2, the P and ε are good predictors for agiven configuration when the dispersed phase concentration isadded as a complementary factor in the model. The scalingfound between the mean droplet size and the deformation in

Figure 11. Equation 20dmed prediction for the mixing cells configurations A-J for the (a) energy, (b) specific energy, (c) power, (d) specific power, (e)shear rate, and (f) deformation as governing factors (GF).

Table 3. Governing Factor Candidates, Calculation Method for Static Mixers, and Ranges Investigated in this Work

GF candidatesexpression forstatic mixers

developed relationshipfor laminar flow

range testedin this work

specific energy E ) ∆p E ) 4KPµeffQeL/πD4 4.73× 106-26× 106 J m-3

specific power ε ) 4∆pQe/πD2L ε ) 16KPµeffQe2/π2D6 1.16× 106-24.7× 106 W m-3

power P ) ∆pQe P ) 4KPµeffQe2L/πD4 12.5-161.7 W

energy E ) ∆pπD2L/4 E ) KPµeffQeL2/D2 48-285.5 Jdeformation γ ) ksL/D 786-1620shear rate γ ) ksV/D ) ks4Qe/πD3 356-1447 s-1

Table 4. Key Results of the Covariance Model of Equation 20

significant effect

GF in eq 20 GF MCC xB R2 â1 [-95% CL;+95% CL] â2 [-95% CL;+95% CL]

specific energy yes yes no 0.97 -0.64 [-0.69;-0.59] n/apenergy yes yes no 0.97 -0.64 [-0.69;-0.59] n/apspecific power yes yes yes 0.91 -0.28 [-0.31;-0.24] -1.09 [-1.84;-0.34]power yes yes yes 0.91 -0.28 [-0.31;-0.24] -1.09 [-1.84;-0.34]shear rate yes yes yes 0.82 -0.44 [-0.52;-0.34] -1.84 [-2.88;-0.80]deformation yes no yes 0.34 -0.67 [-0.39;-0.95] -2.33 [-0.61;-4.05]


the coaxial mixer cannot be applied to predict accuratelydmed

in the static mixer given that a very lowR2 is obtained. Thelarger R2 corresponding to energy-related factors can beexplained by the fact thatµeffQe is directly proportional to theshear stress expression in laminar flow (µγ). This latterexplanation is based on the direct dependence between theaverage shear rate and the fluid velocity for static mixers.8

For configurations A-G (group 1), the mean level of energy(120 J) yielded admed in a close range except for configurationB in which the SMX mixers were put before the helical mixers(see the square area in Figure 11a). The same observations canalso be made for configurations C and G-J (group 2) consider-ing the power or the specific power (see the square area in Figure11c and d). This can be interpreted as if there was a criticalthreshold below which thedmedscales with the energy and abovewhich thedmed scales with the power or specific power. Forthe MCC above this threshold (group 2), decreasing the dropletsize by adding more static mixers is not an adequate approach.By analogy with the kinetics of emulsification in stirred vessels,we explained this behavior by an equilibrium state betweenrupture and coalescence reached above the threshold. Thisassumption is strengthened by the fact that the specific poweris commonly used to predict the drop size in a stirred tank underthe equilibrium state.24

Interestingly, the specific energy cannot be taken as thegrouping governing parameter for the MCC in group 1. Thisshows that before an equilibrium state is reached, the averagedroplet size scales withL-1.2 rather thanL-0.6. Nevertheless,the results shown in Figure 11b can be very useful from apractical standpoint. The energy cost for a continuous emulsi-fication process with static mixer depends on the pumps’ powerconsumption. As for static mixers, it can be evaluated by theproduct of the pressure drop and the flowrate. In mostapplications, the flowrates are set to reach the desired throughputof emulsion with a drop size target. Energy saving can beachieved by minimizing the pressure drop across the mixers.The specific energy in static mixers corresponds to the pressuredrop across the MCC as indicated in Table 3. Among the MCCstested, MCC C should be chosen in this case, because a smallerdroplet was predicted for the same level of∆p (see Figure 11b).Obviously, this choice will be adequate if the energy or thepower dissipated is sufficient to reach the desired droplet size.An inspection of Figure 11b reveals that the energy-efficientMCCs are those in which helical mixers are in front of SMXmixers andNH ≈ NSMX. This shows the benefits of adding low-shear mixers before high-shear mixers for HIPR emulsification.Given that the helical mixer is less efficient than the SMX onefor dispersion applications in laminar flow, this result demon-strates that a premixing step, characterized by a low shear rate,is a key aspect for HIPR emulsification with static mixers andis in agreement with the two-stage approach proposed byCatafalmo et al.13 As it can also be seen in Figure 11b forconfiguration B, the decrease in pressure drop expected bycombining helical and SMX mixers can be lost by insertingthe helical mixers after the SMX ones. These features could be

explained by the fact that a progressive shear stress favors theformation of smaller droplets for HIPR emulsification. Theexperimental plan was not designed to assess the effect ofrelaxation volume on the droplet size. Nevertheless, trialsperformed on MCC D and E (at constant bitumen hardness,yS,andTB) could be compared for this specific aspect. As can beseen in the various graphs of Figure 11, the confidence intervalsoverlap showing that there is no significant difference betweenan MCC with no relaxation zone and an MCC with two non-consecutive relaxation zones ondmed.

For the soap surfactant concentration (yS), the bitumentemperature (TB) and the bitumen hardness, the same approachwas taken to verify if these experimental factors could becaptured by the governing factors. Even if the soap surfactantconcentration and the bitumen temperature can be treated asquantitative factors, they are considered as categorical factorswith respectively three and two levels (see Table 2). To scanthese governing factors, the emulsion flowrate and the bitumenconcentration were varied on the reference MCC. This impliesthat γ ) constant,E ) E, andP ) ε for each run. In order todetermine if these process parameters could be captured by thegoverning factors candidates, the strategy similar to the oneshown in Figure 10 was performed with a covariance analysis.The results summarized in Table 6 demonstrate that the bitumenhardness and theyS need to be added as complementary factorsin the prediction model. ForTB, the power and shear ratecaptured the effect of modifying the bitumen temperature,whereas more energy was required to reach the samedmed at alower bitumen temperature. This apparent discrepancy can beexplained by the fact thatTB affects two parameters with anantagonist effect on the droplet size: the effective viscosity (µeff)and the dispersed phase viscosity (µd). As observed for the effectof the bitumen hardness, an increase ofµd (without changingTB) caused an increase ofdmed (see Figures 8 and 9). Theexpected benefits of increasing the effective viscosity bydecreasing the emulsification temperature are counterbalancedby the effect of increasing the dispersed phase viscosity.

In summary, a single model cannot be proposed due to theeffect of the mixing cells configurations. Before the equilibriumstate, the following equation is proposed:

and at the equilibrium state

where C10 and C11 also depend on the bitumen hardness.Alternatively, P can be replaced byε in eq 22. The 95%confidence limits of the coefficients of eqs 21 and 22 arereported in Table 4.

It should be recalled that HIPR emulsions with the sameformulation, composition, and temperature as those of this studywere produced in a coaxial mixer.2 Despite the significantlylower shear rate in the vessel, smaller droplets (dmed_min) 0.4µm) were obtained. The combination of an extended residencetime in the dispersing zone and a low shear rate which enhance

Table 5. Log-Log Slopes for Scaling Relation between the AverageDroplet Size and the Energy or Specific Energy Found in theLiterature and in this Work

emulsification parameters log-log slopes

inline rotor-stator in laminar flow19 1static mixer (helical) in laminar flow14 0.6static mixer (SMX) in laminar flow20 0.25coaxial mixer/HIPR emulsion2 0.42stirred-tank/HIPR emulsion17 0.2this work 0.64

Table 6. “Capture Matrix” of the Bitumen Hardness, yS, and TB forThree Governing Factor Candidates

captured by

process parameters energy power shear rate

bitumen hardness no no noyS no no noTB no yes yes

dmed) C10(yS,TB)E-0.64 (21)

dmed) C11(yS)P-0.28xB

-1.09 (22)


time dependent phenomena such as the droplet breakup or thesurfactant adsorption is the primary approach to obtain smallerdroplets.

Conclusion

The aim of this article was to evaluate the capabilities ofstatic mixer technology to produce micrometer range HIPRbitumen emulsions in continuous mode and identify the keyfactors controlling the average droplet size. Several static mixerconfigurations were considered based on SMX elements or acombination of helical and SMX elements. The results showedthat in the range of the experimental parameters tested, themedian volume diameters were included between 1.7 and 4.1µm. A covariance analysis was carried out in which the mixingcells configurations were treated as a categorical factor. It wasshown that energy and power (or the specific power) areappropriate governing factors if the mixing cells configurationstested were divided in two groups. This transition was attributedto an equilibrium state in which the droplets are not affectedby factors involving the residence time in the dispersing zone,i.e., energy or specific energy. If the equilibrium state is notattained, the energy dissipated is the adequate factor to calibratethe average droplet size, and otherwise, the power dissipated(or the specific power) is the right factor.

The main conclusion of this work is that the transpositionfrom a stirred vessel to a continuous inline emulsificationprocess at HIPR conditions cannot be based on the shear stressespecially if the residence time differs by a few orders ofmagnitude. A more conservative approach would be to decreasethe residence by 1 order of magnitude or less and use very low-shear mixing conditions especially at the beginning of theemulsification process, i.e., in a premixing step. A loopconfiguration could be designed for this purpose as it has beendone successfully in other emulsification processes.25

Acknowledgment

The authors would like to thank Matthieu Lelie`vre for histechnical assistance.

Nomenclature

Latin Symbols

C1,C2,...,C11 ) constants (-)d ) droplet diameter (m)d32 ) mean surface-weighted diameter (m)d43 ) mean volume-weighted diameter (m)dmed ) median volume diameter (m)D ) conduit diameter (m)E ) energy draw (J)E ) specific energy (J/m3)f ) fanning friction factor with inserts (-)ks ) shear rate constant (-)L ) overall static mixer length (m)N ) number of static mixer elements (-)p ) fluid pressure (Pa)P ) power draw (W)Q ) flowrate (m3/s)tR ) residence time (s)T ) temperature (°C)V ) fluid velocity (m/s)x ) weight fraction in the emulsion (kg/kg)y ) weight fraction in the soap (kg/kg)

Subscripts

B ) bitumencorr ) corrected by the dispersed phase contente ) emulsioneff ) effectiveH ) helicalin ) static mixers section inletout ) static mixer section outletS ) surfactant

Superscripts

a, b ) log-log slope

Greek Symbols

γ ) deformation (-)γ ) shear rate (s-1)â1,â2,â3 ) regression coefficients (-)ε ) specific power (W/m3)η ) viscosity ratio (-)µ ) viscosity (Pa s)σ ) interfacial tension (N/m)τ ) shear stress (Pa)æ ) dispersed phase volume fraction (-)

Dimensionless Number

Reynolds number,Re) FVD/µpressure drop number,KP ) Re× 2f ) D2∆p/µVL ) πD4∆p/

4µQL

AbbreViations

DSD ) droplet size distributionGF ) governing factorH ) helicalHIPR ) high-internal-phase-ratioMCC ) mixing cells configuration.

Literature Cited

(1) Gingras, J.-P.; Tanguy, P. A.; Mariotti, S.; Chaverot, P. Effect ofprocess parameters on bitumen emulsions.Chem. Eng. Process.2005, 44,979.

(2) Gingras, J.-P.; Fradette, L.; Tanguy, P. A.; Jorda, E. Concentratedbitumen-in-water emulsification in coaxial mixers.Ind. Eng. Chem. Res.2007, in press.

(3) Princen, H. M. The structure, mechanics, and rheology of concen-trated emulsions and fluids foams. InEncyclopedic handbook of emulsiontechnology; Sjoblom, J., Ed.; Marcel Dekker: New York, 2001; Chapter11, p 243.

(4) Marchal, J. L.; Yeadon, G.; Lepert, A.; Mourand, J.; Rauline, A. J.M.; Lapie, T.; Marchal, J. Bitumen emulsion with low average particle sizeis prepd. by feeding bitumen, water and surfactant through static mixers.Patent EP283247-A2, 1988.

(5) Durand, G.; Poirier, J. E.; Richard, F.; Lenfant, M. ManufacturingProcess and Emulsions Properties.Proceedings of the 2nd World Congresson Emulsion, Bordeaux, France, September 23-26, 1997.

(6) Poirier, J. E. Les e´mulsions de bitume et la matiere molle, que reste-t-il a comprendre?Proceedings of the 3rd world congress on emulsion,Lyon, France, September 24-27, 2002.

(7) Cuellar, I.; Bullon, J.; Forgarini, A. M.; Cardenas, A.; Briceno, M.I. More efficient preparation of parenteral emulsions or how to improve apharmaceutical recipe by formulation engineering.Chem. Eng. Sci.2005,60, 2127.

(8) Rauline, D.; Tanguy, P. A.; Le Blevec, J.-M.; Bousquet, J. Numericalinvestigation of the performance of several static mixers.Can. J. Chem.Eng.1998, 78, 389.

(9) Kruijt, P. G. M.; Galaktionov, O. S.; Andersen, P. D.; Peters, G. W.N.; Meijer, H. E. H. Analysing mixing in periodic flows by distributionmatrices: mapping method.AIChE J.2001, 47, 105.

(10) Thakur, R. K.; Vial, Ch.; Nigam, K. D. P.; Nauman, E. B.; Djelveh,G. Static mixers in the process industries-A review.Chem. Eng. Res. Des.2003, 81, 787.


(11) Taylor, G. I. The viscosity of a fluid containing small drops ofanother fluid.Proc. R. Soc. London, Ser. A1932, 138, 41.

(12) Taylor, G. I. The formation of emulsions in definable fields of flow.Proc. R. Soc. London, Ser. A1934, 146, 501.

(13) Catafalmo, V.; Shiveley, T. M.; Blum, G. L.; Lipic, P. M.;Desmarais, T. A. Apparatus and process for in-line preparation of HIPEs.Patent WO0127165, 2001.

(14) Grace, H. P. Dispersion phenomena in high viscosity immisciblefluid systems and application of static mixers as dispersion devices in suchsystems.Chem. Eng. Commun.1982, 14, 225.

(15) Metzner, A. B.; Otto, R. E. Agitation of non-Newtonian fluids.AIChE J.1957, 3, 3.

(16) Arzate, A.; Reglat, O.; Tanguy, P. A. Determination of in-lineprocess viscosity using static mixers.Flow Meas. Instrum.2004, 15, 77.

(17) Briceno, M.; Salager J. L.; Bertrand, J. Influence of dispersed phasecontent and viscosity on the mixing of concentrated oil-in-water emulsionsin the transition flow regime.Chem. Eng. Res. Des.2001, 79, 943.

(18) Adler-Nissen, J.; Mason, S. L.; Jacobsen, C. Apparatus for emulsionproduction in small scale and under controlled shear conditions.FoodBioprodrod. Process.2004, 82, 311.

(19) Karbstein, H.; Schubert, H. Developments in the continuous

mechanical production of oil-in-water macro-emulsions.Chem. Eng.Process.1995, 34, 205.

(20) Fradette, L.; Tanguy, P. A.; Choplin, L. Liquid/Liquid viscousdispersions with a SMX static mixer.Chem. Eng. Res. Des.2007, in press.

(21) Nunez, G.; Marin, R.; Ventresca, M. L. Emulsion formation systemand mixing device. Patent US5399293, 1995 .

(22) Fradette, L. EÄ tude de dispersion dans un me´langeur statique SMX.Ph.D. dissertation, EÄ cole Polytechnique de Montre´al, Canada, 1999.

(23) Neter, J.; Wasserman, W.; Kutner, M. H.; Nachtsheim, C. J.Appliedlinear statistical models; McGraw-Hill Irwin: Boston, 2005.

(24) Chen, H. T.; Middleman, S. Drop size distribution in agitated liquid-liquid systems.AIChE J.1967, 13, 995.

(25) Brocart, B.; Tanguy, P. A.; Magnin, C.; Bousquet, J. Design ofin-line emulsification processes for water-in-oil emulsions.J. DispersionSci. Technol.2002, 23, 45.

ReceiVed for reView September 11, 2006ReVised manuscript receiVed February 7, 2007

AcceptedFebruary 16, 2007

IE0611913


Inline Bitumen Emulsification Using Static Mixers

Documents

Transcript of Inline Bitumen Emulsification Using Static Mixers