Ultra Scale-Down Device to Predict Dewatering Levels of Solids Recovered

in a Continuous Scroll Decanter Centrifuge

A. G. Lopes and E. Keshavarz-MooreThe Advanced Centre of Biochemical Engineering, Department of Biochemical Engineering, University College London, TorringtonPlace, London WC1E 7JE, U.K.

DOI 10.1002/btpr.1814Published online October 18, 2013 in Wiley Online Library (wileyonlinelibrary.com)

During centrifugation operation, the major challenge in the recovery of extracellular pro-teins is the removal of the maximum liquid entrapped within the spaces between the settledsolids–dewatering level. The ability of the scroll decanter centrifuge (SDC) to process con-tinuously large amounts of feed material with high concentration of solids without the needfor resuspension of feeds, and also to achieve relatively high dewatering, could be of greatbenefit for future use in the biopharmaceutical industry. However, for reliable prediction ofdewatering in such a centrifuge, tests using the same kind of equipment at pilot-scale arerequired, which are time consuming and costly. To alleviate the need of pilot-scale trials, anovel USD device, with reduced amounts of feed (2 mL) and to be used in the laboratory,was developed to predict the dewatering levels of a SDC. To verify USD device, dewateringlevels achieved were plotted against equivalent compression (Gtcomp) and decanting (Gtdec)times, obtained from scroll rates and feed flow rates operated at pilot-scale, respectively.The USD device was able to successfully match dewatering trends of the pilot-scale as afunction of both Gtcomp and Gtdec, particularly for high cell density feeds, hence accountingfor all key variables that influenced dewatering in a SDC. In addition, it accurately mim-icked the maximum dewatering performance of the pilot-scale equipment. Therefore the USDdevice has the potential to be a useful tool at early stages of process development to gatherperformance data in the laboratory thus minimizing lengthy and costly runs with pilot-scaleSDC. VC 2013 American Institute of Chemical Engineers Biotechnol. Prog., 29:1494–1502,2013Keywords: scroll decanter centrifuge, dewatering, ultra scale-down, Baker’s yeast

Introduction

Scroll decanter centrifuge

The scroll decanter centrifuge (SDC) is a continuous cen-trifuge with a simultaneous addition of feed, and dischargeof liquid and solids. This machine can handle feeds withvery high solids concentration up to 60–80% (v/v)1,2 andeven though centrifugal speeds are low when compared withother centrifugal equipments, it still delivers relatively drycakes.3 The mode of operation of a SDC has been previouslydescribed.4–6 The essential parts of a SDC are the rotatingpart, the bowl, and the scroll located inside the bowl, whichrotates in the same direction, but at a slightly differentspeed. The bowl consists of a cylindrical section followed bya conical section, called the beach. The differential speedbetween the bowl and the scroll permits the solids to be con-veyed by the scroll continuously to the conical end of thebowl. The liquid is discharged continuously in the oppositedirection towards the cylindrical end of the bowl. The feedenters the bowl through a stationary pipe into the feed zonelocated in the center of the scroll. When feed is introducedinto the bowl, it is accelerated circumferentially and separa-

tion takes place by centrifugal force. The solids in the feedsuspension are deposited against the bowl wall by centrifugalforce. The scroll movement rejects settled solids through theports at the conical end of the centrifuge and the clarifiedliquids flow to the cylindrical end of the bowl where theyexit over the weir plates.

Ultra scale-down

Ultra scale-down (USD) is a methodology that mimics theoperation of large-scale machinery using small amounts ofprocess material in the laboratory, with typical volumes inthe order of 50–100 mL. As a result, greater number ofexperiments can be performed, with reduced quantity of feedmaterial, while providing design information for scale-up.This is a valuable tool for minimizing the time and costrequired at early stages of bioprocess development.

The use of USD approaches to investigate the clarificationperformance of industrial centrifuges has been subject toextensive research.7–17 The scaling between centrifuges ofdifferent sizes and designs is often undertaken according tothe Sigma theory, which describes the solid–liquid separationperformance of centrifuges, however it does not predict thedewatering performance of a centrifuge. In addition, whendealing with continuous centrifuges such as the SDC, Sigma

Correspondence concerning this article should be addressed toE. Keshavarz-Moore at e.keshavarz-moore@ucl.ac.uk.

1494 VC 2013 American Institute of Chemical Engineers

theory alone is not generally used for scaling up because ofthe dependence of clarification on solids conveyance.

Even though solids dewatering is important in many appli-cations, for instance, in processes where the product of inter-est is secreted by the cells, little work has been undertakento develop small-scale predictive tools, particularly for appli-cations within the biopharmaceutical industry. The majorityof work undertaken concerned sludge dewatering for wastetreatment5,6,18–22 using continuous type centrifuges like theSDC or the gravity thickener, mainly due to their high feedinput and high solids capacity. Initial work on dewatering incontinuous centrifuges was developed to scale-up dewateringperformance in a SDC. It was suggested that scale-up forcompressible solids such as activated sludge required condi-tions of equal centrifugal pressure for equivalent dewatering.This was achieved by maintaining the relative centrifugalforce on scale-down between two geometrically similar cen-trifuges.6 In the context of the biopharmaceutical industry,this criteria was applied to scale-down dewatering levels in aSDC for the separation of homogenized flocculated yeast.Large-scale centrifugal dewatering was successfully pre-dicted using centrifugal tubes in the laboratory by maintain-ing the residence times (based on scroll rates) with constantrelative centrifugal force.23 More recently, an alternative lab-oratory method to predict dewatering levels in semi-continuous centrifuges was developed. This consisted ofmaintaining similar height of settled solids (area of cross-section of the solids collection zone for a constant width)and residence time of the larger centrifuge in the laboratory.Residence times were based in the time a solid particle spentin the formed cake (compaction time) and it was calculatedfrom Sigma theory. Dewatering levels of feeds of high celldensity S. cerevisiae suspensions and E. coli cultures weresuccessfully predicted using this method in a semicontinuouspilot-scale Carrpowerfuge

TM

centrifuge.24

The aim of the present work was to develop a USDapproach using a custom designed USD device, to be usedin a regular bench-top centrifuge, to predict dewatering lev-els of recovered solids achieved in a continuous SDC.Baker’s yeast feed suspensions with a low cell volume frac-tion of 0.1 (v/v) and a high cell density of 0.3 (v/v) weretested in the pilot-scale SDC and in the laboratory, usingready available 2 mL Eppendorf tubes. This was followed bythe verification of the developed USD approach by compar-ing between the levels of centrifugal dewatering obtained inthe pilot-scale SDC and standard laboratory method, to thoseobtained in by the USD device.

Theoretical

Dewatering process in a scroll decanter centrifuge

Two mechanisms by which suspensions are dewatered ina SDC have been previously suggested: drainage and com-pression dewatering.6,25 The drainage dewatering is achievedwhen liquid flows back down the slope of the conical areaof the SDC through the open pores of network formed bythe solids. Compression dewatering occurs when the spacesbetween the settled solids shrink under the action of the cen-trifugal force. The liquid is compressed out of the spacesbetween the solids and up the surface, as the solids structurecollapse under the pressure of the accumulated solids above.Accordingly, the level of compression dewatering increaseswith increased centrifugal force and also with the pressure

generated by the accumulated solids, i.e., the height of thelayer of the settled solids. Compression dewatering onlyoccurs along the cylindrical area of the bowl in a SDC.Drainage dewatering levels depend on the centrifugal forceand also on the slope of the conical area (beach angle) andwould only take place in the conical area of the SDC.

Dewatering key variables

The most significant factors in determining the level of dew-atering in centrifuges have been reported to be: centrifugalforce, residence times, and the settled solids height.6,18,20,21,26

In a SDC the scroll rates determined the solids residence timein the bowl,22 with low scroll rates leading to better dewater-ing due to a greater solids residence time.25 Feed flow ratesalso influenced the levels of dewatering as the settled solidswere continuously re-suspended by incoming flow.4 In addi-tion, the higher the volumetric feed flow rates, the smaller theresidence time of the feed suspension (and hence solids) insideof the centrifuge and the less time it was subjected to the cen-trifugal force, hence the worse the dewatering performance.22

Finally, an increase of the solids concentration in the feedhelped the dewatering process. At higher solids concentration,the pressure applied at the very bottom of the settled solidswould increase due to additional pressure exerted from theaccumulated solids above, explaining the increased dewateringlevels with higher solids concentration in the feed. It wasfound that the compressive force experienced at any point inthe settled solids was dependent upon the height of solidsabove that position.16,19,24,25

USD development

After analyzing the dewatering process in a SDC andidentifying the key variables that influenced dewatering insuch a centrifuge, a method to use a custom designed USDdevice in the laboratory was developed. The USD methodconsisted of spinning the USD device at equivalent residencetimes (Gt) that matched the larger centrifuge; and using afeed concentration that would replicate, in the USD device,the maximum height of wet solids obtainable in the pilot-scale centrifuge.27,28

Equivalent Residence Time (Gt). Equivalent residencetime was described by29:

Gt5x2r0

gt (1)

where x is the bowl angular velocity, r0 is the external bowlradius, g is the acceleration due to gravity, and t is the resi-dence time.

Compression Time (tcomp) and Decanting Time (tdec).The residence time in Eq. 1 was divided into compression time(tcomp) and decanting time (tdec), according to the dewateringoperation. These were the times the USD device would bespun in the laboratory centrifuge. In a SDC, compression timewas mainly dependent on the action of the scroll, which deter-mined the solids residence time inside the bowl30:

tcomp5L

DxP(2)

where L is the length of the bowl, Dx is the scroll angularvelocity and P is the pitch (the distance between scrollblades).

Biotechnol. Prog., 2013, Vol. 29, No. 6 1495

The effect of decanting of the compressed solids wasaccounted for by the decanting time, which determined theforce applied to solids to go up the conical slope and out ofthe pilot-scale SDC. Decanting time was calculated by deter-mining the residence time of solids in the solids holding areaof the pilot-scale centrifuge31:

tdec5Vcapacity

Qfeed

5pL½R22ðR2hwÞ2�

Qfeed

(3)

where Vcapacity is the solids volumetric capacity of the centri-fuge (or solids holding volume), Qfeed is the volumetric feedrate, R is the external radius of the centrifuge, hw the weirheight (or pond height), and L is the length of the bowl.

Maximum Compacted Wet Solids Height (hsol). Themethodology to mimic dewatering using the USD deviceconsisted of using a feed concentration that would replicatethe maximum height of wet solids in the pilot-scale centri-fuge. Keeping the same height of solids would mimic theextra pressure applied to solids due to solids build-up insidethe centrifuge which in turn helped to compress the liquidout of the spaces between the settled solid particles. Thiswas calculated based on the volumetric solids capacity of thecentrifuge and in the concentration of solids in the feed(measured as feed cell volume fraction), assuming 100%clarification. It was also assumed that the conical area wasnegligible and that the full shape of the centrifuge wascylindrical.

Dewatering Level. Dewatering level was obtained usingthe following equation16:

%Dewatering51002

�WCW2DCW=dwr

WCW3100

�(4)

where WCW is the weight of the wet solids, DCW is theweight of the dried solids and dwr is the cell dry weight towet weight ratio.

Materials and Methods

All reagents were purchased from Sigma–Aldrich (Sigma–Aldrich, Poole, UK) unless otherwise stated, and were of thehighest grade available.

Preparation of feed material

Dry Baker’s yeast (Sacharomyces cerevisiae) was sus-pended in sodium phosphate buffer (pH 6.5) to give suspen-

sions of the required concentration. Mixtures were agitatedfor �1 h at room temperature (20�C) in a stirred tank.

Pilot-Scale Scroll Decanter Centrifuge

Yeast feed suspensions at varying cell volume fractionswere pumped into a Scroll Decanter P600 centrifuge (AlfaLaval Sharples, Camberley, Surrey, England) at flow ratesranging 60–720 L h21 and differential scroll rates between 2and 55 rpm. The bowl speed was set at constant 6000 rpm(Table 1). Samples from the solids discharge port of the cen-trifuge were taken for assessment of dewatering. All sampleshad a minimum of �10 g of wet solids to ensure accuratemeasurement of dewatering. All samples were taken after atleast five solid volume changes and in triplicate. Operationtemperature did not exceed 20�C.

Standard Laboratory Centrifugal Experiments

These laboratory experiments consisted of spinning centri-fuge tubes at the same range of Gtcomp; and using same feedconcentration, in laboratory centrifuge tubes, as used in thepilot-scale centrifuge. 2.2 mL Eppendorfs with 2 mL samplevolumes were spun in a bench top centrifuge Eppendorf5818R (Eppendorf UK, Cambridge, UK) with a fixed rotorFA-45-30-11 (Table 1). After each centrifuge run, superna-tants were pipetted off and wet cell pellets were taken fordewatering analysis. All experimental points were carriedout in triplicate and at 20�C.

USD device design

To estimate the level of dewatering of the pilot-scaleSDC in the laboratory using small quantities of feed mate-rial, a USD device was developed. This USD device simu-lated two sections of the SDC where two differentdewatering operations occurred: compression and decanting.The decanting dewatering was replaced by the previouslysuggested drainage dewatering6,25 because as yeast suspen-sions were of compressible nature, drainage would notoccur. Instead, once solids were dewatered via compressionin the cylindrical area of the SDC, the movement of thecompacted solids up the conical slope of the SDC due tothe action of the scroll would force the residual liquid rest-ing on top of the settled solids to flow down the slope ofthe conical area, being dewatered by decanting. In order tosimulate the decanting effect, the USD device was designedto keep the slope angle of the conical area of the SDC(slope with 10� angle).

Table 1. Characteristics of a Pilot-Scale Scroll Decanter Centrifuge P600 (Centrifuge (Alfa Laval Sharples, Camberley, Surrey, England) and a

Laboratory Bench Centrifuge Eppendorf 5818R (Eppendorf UK Ltd., Cambridge, UK) Used For: Standard Laboratory Experiments With 2-mL

Eppenforf Tubes; and USD Device Using Different Rotor Types During Compression and Decanting Stages of USD Device Experiments

Pilot-Scale SDC

Vcapacity (L)Qfeed

(Lh21)Nscroll

(rpm)Nbowl, max

(rpm) Dimensions

�2.9 60–720 2–55 6000R (0.075 m); P (0.050 m);

L (0.360 m); hw (0.020 m)Laboratory Method

Tube Type Rotor Type Dimensions2.2-mL Eppendorf tubes FA-45-30-11 r0 (0.100m);USD Device

Experiment Step Holder Type Rotor Type DimensionsCompression Horizontal A-4-62 r0 (0.155 m)Decanting Vertical T-60-11 r0 (0.070 m)

1496 Biotechnol. Prog., 2013, Vol. 29, No. 6

The USD device was composed of three parts (Figure 1):a top part, the main chamber with seals (A1); a detachablebottom part, the collecting chamber (A2); and the mainchambers lid (A3). The main chamber worked as a normalcentrifuge tube when gaps were sealed, simulating the solidsbeing dewatered via compression in the pilot-scale SDC.When main chamber was not sealed, it simulated the slopeof the conical area SDC. The collecting chamber of the USDdevice was used as a collection container for capturing solidsthat went up the slope (when Main chambers gaps were notsealed). The two operations (compression and decanting)were tested in the same device in order to quantify more

accurately the final level of dewatering in the pellet andavoid losses of sample associated with transfer ofsuspension.

USD device experiments

The laboratory experiments using the USD device con-sisted of a bench top centrifuge Eppendorf 5818R (Eppen-dorf UK, Cambridge, UK) and two different holders(designed and costumed built in UCL Engineering work-shop, London, UK): horizontal and vertical. Figure 2shows the dimensions of USD device/holders (Figure 2A),and the positioning of USD device/holders in relation to

Figure 1. Picture of USD device used to mimic dewatering in a pilot-scale SDC, composed of three main parts.

(A1) Main chamber that when completely sealed simulated the compression dewatering in the cylindrical area of a SDC, and when not sealed (gapsof 0.1 mm), simulated the slope in the conical area (10� angle) in a SDC; (A2) Collecting chamber used as a collection container to capture solidparticles that went up the simulated slope, when Main chamber not sealed, and to measure dewatering levels of solids captured; (A3) Main chamberlid used to contain suspension inside the Main chamber while simulating the drainage dewatering experienced in the slope in the conical area.

Figure 2. Dimensions of USD device and holders, and positioning in a laboratory centrifuge.

(A) Dimensions of USD device and holders, designed and built to use in a regular bench top centrifuge Eppendorf 5818R (Eppendorf UK, Cam-bridge, UK). (B) Positioning of USD device in relation to bench top centrifuge axis of rotation during compression and decanting steps.

Biotechnol. Prog., 2013, Vol. 29, No. 6 1497

axis of rotation of laboratory centrifuge (Figure 2B). Dif-ferent laboratory centrifuge rotors were used depending onthe holder used: horizontal holders were attached to aswing rotor A – 4 – 62 and vertical holders used withrotor T – 60 – 11 (Table 1).

The first step of the USD device experiments was desig-nated as the compression step, where USD device was spun(Main chamber with gaps sealed) perpendicularly to theaxis of the laboratory centrifuge. The positioning of theUSD device was achieved by using the horizontal holder(Figure 3A). This simulated the compression dewatering ofsolids due to the rotation of the bowl’s centrifugal force.The result was a pellet of solids settled in the bottom ofthe Main chamber and supernatant on top. The second stepwas the decanting step, when seals in the Main chamberwere removed and USD device was positioned parallel toaxis of bench top centrifuge. The positioning of the USDdevice was achieved by using the vertical holder (Figure3B). Here the compressed pellet of solids moved up thesimulated slope and were directed into the Collecting cham-ber for subsequent dewatering level quantification. Thissimulated the scroll force being applied to the compactedsolids up the slope of the conical area and out of the SDC.Figure 4 shows pictures of the stages of USD device proce-dure to predict dewatering levels of the pilot-scale SDC.

Maximum height of wet solids (hsol) in the USD device

In order to mimic the maximum height of wet solids ofthe pilot-scale centrifuge in the USD device, Baker’s yeast

suspensions of varying cell volume fractions between 0.01and 0.75 (v/v) were spun in a bench top centrifuge Eppen-dorf 5818R (Eppendorf UK, Cambridge, UK). The USDdevice containing 2-mL sample volume was spun for up to90 min at RCF 3000g until height of wet solids was con-stant. The supernatant was pipetted off and final height ofwet solids for each feed concentration measured. All experi-mental points were carried out in triplicate and at 20�C.

Wet and dry cell weights

To assess the dewatering levels of wet solids, wet (WCW)and dry (DCW) weights were quantified. Empty collectingchambers of USD device were predried and preweighed.After centrifugal experiments and removal of supernatant, todetermine the WCW, chambers containing wet pellets wereweighed. To obtain the DCW, the same cell pellets, weredried at 100�C until a constant weigh was achieved. TheDCW and WDC were obtained by subtraction of the weightof predried collecting chambers of USD device.

Results and Discussion

Height of wet solids in USD device (hsol)

The USD device required to use a feed concentration thatwould mimic the height of wet solids of the pilot-scale SDC.The feed concentration to be used in the USD device toachieve the required height of solids was achieved by plot-ting Baker’s yeast feed cell volume fraction as a function ofthe final height of wet solids in 2 mL USD device (Figure

Figure 3. Pictures of USD device and holders placed in regular bench top centrifuge with different centrifuge rotors.

USD device and holders were designed, built and customised to be used in bench top centrifuge Eppendorf 5818R (Eppendorf UK, Cambridge, UK).(A) Horizontal holder with USD device in swing out rotor A-4-62 (compression step); (A1) Horizontal holder with USD device; (B) Vertical holderwith USD device in rotor T-60-11 (decanting step); (B1) Vertical holder with USD device.

1498 Biotechnol. Prog., 2013, Vol. 29, No. 6

5). The resulting correlation was used to mimic the calcu-lated height of wet solids of the pilot-scale centrifuge. Byknowing the height of solids experienced in the pilot-scalecentrifuge, and using the correlation, the feed concentrationto be used in the 2 mL USD device could be easily chosen.

Pilot-scale scroll decanter centrifuge

The dewatering performance of Baker’s yeast suspensionswas investigated in a pilot-scale SDC. Baker’s yeast feedconcentration of 0.1 (v/v) (�33 g L21 DCW) (Figure 6) wasused to test the low cell density suspension; and the 0.3 (v/v) (�100 g L21 DCW) cell volume fraction feed (Figure 7),to test a high cell density suspension. A range of scroll ratesand volumetric feed flow rates were tested in the pilot-scale

SDC, which were then translated into a function of the Gt.All graphs were found to have levels of dewatering increas-

ing with Gt until reaching a maximum value (�66.5%) for

all conditions measured. This was expected as the more pres-

sure applied to the settled solids due to an increase in the

relative centrifugal force and/or residence time, would result

in a higher level of solids compression and consequently

higher level of dewatering.18,20,26

The percentage of dewatering was assessed for varying Gtvalues which were separated into compression time (tcomp)

and decanting time (tdec). Varying equivalent compression

times (Gtcomp) were measured at constant equivalent decant-

ing times (Gtdec) (Figures (6A and 7A). Similarly, the effect

of varying Gtdec on dewatering was measured at constant

Gtcomp (Figure 6B and 7B). The effect of Gtcomp however

was seen to have a more prominent and rapid effect upon

the dewatering levels than Gtdec. This was expected as litera-

ture reports that Baker’s yeast suspensions is of a compressi-

ble nature and would primarily dewater via compression.6,26

The reason for the slower, more gradual impact of Gtdec

upon dewatering was because the decanting process only

removed residual liquid that was left on top of the settled

solids.

The effect of an increase in the solids concentration in the

feed was previously reported in literature16 with an increase

in dewatering performance due to the extra-pressure exerted

on the accumulated solids. This effect was not observed in

the present pilot-scale SDC experiments as the maximum

dewatering level of �66.5% was reached by both Baker’s

yeast feeds, irrespective of the solids concentrations used.

Because of the continuous operation of the SDC, the solids

first needed to accumulate inside the cylindrical area of the

centrifuge before the scroll was able to carry the settled sol-

ids up the slope of the conical area and out of the centrifuge.

The height of the settled solids was maintained thereafter

during the continuous operation of the SDC (continuous

removal of the settled solids by the scroll movement) and

hence so was the maximum dewatering levels reached.

Pilot-scale SDC, lab method, and USD device

After pilot-scale trials, and the development of the USDdevice and methodology, the latter dewatering results werecompared with the ones obtained in the pilot-scale SDC

Figure 5. Correlation between maximum height of wet solidsin the USD device and Baker’s yeast feed cell vol-ume fraction (v/v). Linear regression fitted to thedata set with R2 > 0.98.

Figure 4. Pictures of stages of USD device procedure to mimic dewatering in a SDC.

(A) 2 mL of feed loaded onto USD device Main chamber with seals on; (B) compression step where suspension was spun in USD device (with bothseals on) and using an horizontal holder (USD device placed perpendicular to the axis of rotation of the laboratory centrifuge); (C) removal of super-natant; (D-E) decanting step where the main chamber (with both seals off) and settled solids were spun in a vertical holder (USD device placed par-allel to the axis of rotation of the laboratory centrifuge); (F) movement of settled solids into the Collecting chamber of the USD device where solidswent up the 10� angle slope of the USD device were spun using horizontal holder; (G) solids recovered in Collecting chamber of USD device formeasurement of dewatering level.

Biotechnol. Prog., 2013, Vol. 29, No. 6 1499

and standard laboratory method. The percentage of dewa-tering was assessed for varying Gt values which, as previ-ously shown, were separated into equivalent compressiontime (Gtcomp) (Figures 6A and 7A) and equivalent decant-ing time (Gtdec) (Figures 6B and 7B), for Baker’s yeastfeeds of 0.1 (v/v) and 0.3 (v/v) cell volume fractions,respectively. It can be seen from these graphs that thetrend in dewatering data relating to the USD device wasalmost identical to the pilot-scale SDC, where dewatering

increased until reaching a maximum value of �66.5%, forboth feed solids concentrations analyzed. The dewateringtrend as a function of Gtcomp from the USD device wasalmost identical to the pilot-scale SDC, where dewateringtrends differed by � 610%, for both Baker’s yeast feedconcentrations tested. This was also the case for the dew-atering trend as a function of Gtdec. The only exceptionwas seen when Gtdec < 0.15 3 106 s and at constantGtcomp 5 0.023106 s, for the lower Baker’s yeast feed

Figure 7. Dewatering performance of a pilot-scale SDC, laboratory method and USD device as a function of equivalent residencetimes (Gt) with Baker’s yeast feed with 0.3 (v/v) cell volume fraction. Pilot-scale runs using a scroll decanter centrifugeP600 (black symbols) performed at flow rates ranging from 60 - 720 L/h and scroll rates of 2 - 55 rpm; and the bowl speedwas set at constant 6000 rpm. Laboratory experiments (white symbols): Laboratory method using 2 ml Eppendorf tubes (2ml sample) with the same feed concentration as pilot-scale SDC; and USD device designed and costumed built to operate ina bench top centrifuge Eppendorf 5818R with Baker s yeast feed with 0.4 (v/v) cell volume fraction. (A) Pilot-scale SDCand USD device at varying equivalent compression times (Gtcomp) measured at constant equivalent decanting time (Gtdec) of:(�, �) 0.363106 sec. Laboratory method at varying Gtcomp. (B) Pilot-scale SDC and USD device at varying Gtdec measuredat constant Gtcomp of: (�, �) 0.303106 sec. Regression line of the best fit to each representative data set using double rec-tangular hyperbola function (R2 > 0.97): (—) Pilot-scale SDC; (….) Laboratory method; (---) USD device.

Figure 6. Dewatering performance of a pilot-scale SDC, laboratory method and USD device as a function of equivalent residencetimes (Gt) with Baker’s yeast feed with 0.1 (v/v) cell volume fraction. Pilot-scale runs using a scroll decanter centrifugeP600 (black symbols) performed at flow rates ranging from 60 to 720 L h21 and scroll rates of 2–55 rpm; and the bowlspeed was set at constant 6000 rpm. Laboratory experiments (white symbols): Laboratory method using 2 mL Eppendorftubes (2-mL sample) with the same feed concentration as pilot-scale SDC; and USD device designed and costumed built tooperate in a bench top centrifuge Eppendorf 5818R with Baker’s yeast feed with 0.3 (v/v) cell volume fraction. (A) Pilot-scale SDC and USD device at varying equivalent compression time (Gtcomp) measured at constant equivalent decanting time(Gtdec) of: (�, �) 0.3 3 106 sec; (~, ~) 0.23106 sec; (W, �) 0.13106 sec. (~N) Laboratory method at varying Gtcomp. (B)Pilot-scale SDC and USD device at varying Gtdec measured at constant Gtcomp of: (~, ~) 0.023106 sec; (�, �) 0.053106

sec. Regression line of the best fit to each representative data set using double rectangular hyperbola function (R2 > 0.97):(—) Pilot-scale SDC; (….) Laboratory method; (---) USD device.

1500 Biotechnol. Prog., 2013, Vol. 29, No. 6

concentration tested. At low Gt (both Gtcomp and Gtdec)solids were not fully settled having more trapped liquidinside (lower dewatering levels). This makes the analysisusing the USD device procedure more susceptible to var-iations due to difficulties of fluid transfer and also thepotential loss of sample, which in turn affects the accu-racy of the dewatering prediction.

The standard laboratory method used the same feed concen-tration as the pilot-scale SDC in the 2-mL Eppendorf tube, in abench-top centrifuge. It could be seen that using this method,dewatering trends were only predicted within � 6 10% whenhigh concentration of Baker’s yeast was used. As previouslyreported,27 this only occurred because the concentration ofBaker’s yeast of 0.3 (v/v) used in the 2-mL Eppendorf tube,resulted in a similar height of settled solids as experienced inthe pilot-scale SDC. In addition, using solely 2-mL Eppenforftubes, the effect of decanting dewatering, represented as Gtdec,could not be accounted for. The USD device, however, wasable to both take into account the two types of dewatering thattook place in a SDC and accurately predict the dewatering per-formance of the pilot-scale SDC, particularly for feeds withhigh Baker’s yeast cell concentrations.

Figure 8 summarizes averaged maximum levels of dewa-tering achieved in the pilot-scale SDC and results obtainedfrom laboratory predictions using both the standard labora-tory method and the USD device. This graph also presentsthe calculated P values using t student test to assess signifi-cant differences between laboratory predicted values andpilot-scale results. Using the standard laboratory method forlow concentrations of Baker’s yeast feed resulted in a Pvalue < 0.05 which meant that results were statistically dif-ferent from the pilot-scale SDC. However, as previouslydescribed, at higher feed concentrations the laboratorymethod reached similar levels to that of the SDC (P value �0.05). The developed USD device could accurately predictmaximum dewatering levels achieved in the pilot-scale SDCfor both concentrations of Baker’s yeast analyzed, whichwas highlighted by the P value of � 0.05.

Conclusions

The scroll decanter centrifuge (SDC) is used extensivelyin wastewater technology, chemical industry, mineral treat-ment, etc. It is relatively low cost and has the capacity ofhandling high volumetric feed rates, high cell density feedsand producing relatively dry solids. However this centrifugehas been somewhat neglected by the biopharmaceuticalindustry because for reliable prediction of the dewateringperformance in such a centrifuge, tests using the same kindof equipment at pilot-scale are required, which are time con-suming and costly. To alleviate the need of pilot-scale runs,a novel ultra scale-down (USD) approach using a USDdevice was developed to be used in a regular bench-top cen-trifuge. This USD device simulated two different dewateringoperations that occur in a SDC: compression, which takesplace in the cylindrical area of the SDC; and decanting thattakes place in the conical area of the SDC. To simulate thedecanting dewatering, the USD device was designed tomaintain the same slope angle of the conical area as thepilot-scale SDC. This USD device used reduced feed vol-umes (2 mL) and a feed concentration that would mimic theheight of wet solids experienced in the pilot-scale SDC. Tovalidate the USD device, feeds of Baker’s yeast suspensionswere run in a pilot-scale SDC and the level of dewateringcompared. Dewatering levels achieved were plotted againstequivalent residence time (Gt), obtained from scroll ratesand feed flow rates used at pilot-scale, and differentiated asequivalent compression (Gtcomp) and decanting (Gtdec) times,respectively. Standard laboratory experiments using 2-mLEppendorf tubes employing comparable feed concentrationto that of the pilot-scale SDC were only able to predict thedewatering levels of the pilot-scale when high concentrationsof Baker’s yeast feed were used, and did not account for thedecanting dewatering in a SDC. The dewatering trend as afunction of Gtcomp from USD device matched the pilot-scaleSDC by 610%, for both Baker’s yeast feed concentrationstested. Pilot-scale dewatering trends as a function of Gtdec

however, only matched the USD device results when highequivalent residence times (both Gtcomp and Gtdec) wereexperienced. This effect was especially prominent at lowBaker’s yeast suspensions.

The USD device was found to give an accurate predictionof maximum dewatering levels achieved in the pilot-scalescroll decanter centrifuge in all cases studied, namely lowand high Baker’s yeast feed solids concentrations (P value �0.05).

The USD device successfully predicted the level of dewa-tering in the recovered solids of the pilot-scale SDC, particu-larly for high solids feed concentrations, and simultaneouslyaccounted for all variables that influenced dewatering in aSDC. Therefore this USD device provided an effective wayof determining centrifugation conditions that maximizeddewatering levels during processing of high cell densityyeast cultures that allowed for faster process development,targeting key process variables, requiring only millilitrequantities of material.

Acknowledgments

The authors acknowledge the Engineering and PhysicalSciences Research Council (EPSRC) and the InnovativeManufacturing Research Center (IMRC) for funding.

Figure 8. Summary of averaged maximum dewatering levelsachieved in a pilot-scale SDC, laboratory methodand USD device with Baker s yeast feed cell volumefractions of 0.1 (v/v) and 0.3 (v/v). T-student testused to obtain p-values between the pilot-scale SDCand the USD device. (Pilot-scale) P600 scrolldecanter centrifuge (Alfa-laval Sharples, Camberley,Surrey, UK); (Lab method) Laboratory methodusing 2.2 ml Eppendorf tubes with same feed con-centration as pilot-scale centrifuge; (USD device)USD device with 2ml feed sample.

Biotechnol. Prog., 2013, Vol. 29, No. 6 1501

Nomenclature

DCW = Dry cell weight (g)

Dx = Scroll differential angular velocity in a SDC (rad

s21)

dwr = Cell dry weight to wet weight ratio (g g21)

g = Acceleration due to gravity (m s22)

Gt = Equivalent residence time (s)

Gtcomp = Equivalent compression time (s)

Gtdec = Equivalent decanting time (s)

hsol = Height of wet solids (m)

hw = Weir height or pond height (m)

L = Length of the centrifuge bowl (m)

Nscroll = Scroll rate (rpm)

Nbowl,max = Maximum rotational speed of the bowl (rpm)

P = Pitch or the distance between scroll blades in a SDC

(m)

Qfeed = Volumetric feed rate (m3 s21)

r0 or R = External bowl radius (m)

SDC = Scroll decanter centrifuge

T = Residence time/spin time in the centrifuge (s)

tcomp = Compression time (s)

tdec = Decanting time (s)

USD = Ultra scale-down

Vcapacity = Solids volumetric capacity of the centrifuge (pond

volume) (m3)

WCW = Wet cell weight (g)

x = Bowl angular velocity (rad s21)

Literature Cited

1. Atkinson B, Mavituna F. Biochemical Engineering and Biotech-nology Handbook. New York: Palgrave Macmillan/Nature Press;1983.

2. Hatti-Kaul R, Mattiasson B. Isolation and purification of pro-teins. Biotech Bioproc Ser. 2003;27.

3. Medronho RA. Solid–liquid separation. In: Hatti-Kaul R, Mat-tiasson B, editors. Isolation and Purification of Proteins. BiotechBioproc Ser. 2003;27:131–190.

4. Faust T, Gosele W. Investigation of the clarification process indecanter centrifuges. Ger Chem Eng. 1986;9:136–142.

5. Leung WWF, Shapiro AH. Dewatering of fine-particle slurriesusing a compound-beach decanter with cake-flow control. FiltrSep. 1999;36:49.

6. Gosele W. Scale-up of helical conveyor type decanter centri-fuge. Ger Chem Eng. 1980;3:353–359.

7. Ambler CM. The evaluation of centrifuge performance. ChemEng Prog. 1952;48:150–158.

8. Ambler CM. The theory of scaling up laboratory data for thesedimentation type centrifuge. J Bioch Microb Tech Eng. 1959;1:185.

9. Boychyn M, Doyle W, Bulmer M, More J, Hoare M. Laboratoryscaledown of protein purification processes involving fractionalprecipitation and centrifugal recovery. Biotech Bioeng. 2000;69:1–10.

10. Boychyn M, Yim SSS, Shamlou PA, Bulmer M, More J, HoareM. Characterization of flow intensity in continuous centrifugesfor the development of laboratory mimics. Chem Eng Sci. 2001;56:4759–4770.

11. Boychyn M, Yim SSS, Bulmer M, More J, Bracewell DG,Hoare M. Performance prediction of industrial centrifuges usingscale-down models. Bioproc Biosyst Eng. 2004;26:385–391.

12. Mannweiler K, Hoare M. The scale-down of an industrial discstack centrifuge. Bioproc Biosyst Eng. 1992;8:19–25.

13. Maybury JP, Mannweiler K, Titchener-Hooker NJ, Hoare M,Dunnill P. The performance of a scaled down industrial discstack centrifuge with a reduced feed material requirement. Bio-proc Biosyst Eng. 1998;18:191–199.

14. Maybury JP, Hoare M, Dunnill P. The use of laboratory centrif-ugation studies to predict performance of industrial machines:studies of shear-insensitive and shear-sensitive materials. Bio-tech Bioeng. 2000;67:265–273.

15. Vesilind PA. Estimation of sludge centrifuge performance. JSanit Eng Div. 1970;96:805–818.

16. Salte H, King JMP, Baganz F, Hoare M, Titchener-Hooker NJ.A methodology for centrifuge selection for the separation ofhigh solids density cell broths by visualisation of performanceusing windows of operation. Biotech Bioeng. 2006;95:1218–1227.

17. Tustian AD, Salte H, Willoughby NA, Hassan I, Rose MH,Baganz F, et al. Adapted ultra scale-down approach for predict-ing the centrifugal separation behavior of high cell density cul-tures. Biotechnol Prog. 2007;23:1404–1410.

18. Chu CP, Lee DJ. Experimental analysis of centrifugal dewater-ing process of polyelectrolyte flocculated waste activated sludge.Water Res. 2001;35:2377–2384.

19. Landman KA, White LR, Buscall CR. The continuous-flowgravity thickener—steady state behaviour. AIChE J. 1988;34:239–252.

20. Lee SJ, Chu CP, Tan RBH, Wang CH, Lee DJ. Consolidationdewatering and centrifugal sedimentation of flocculated acti-vated sludge. Chem Eng Sci. 2003;58:1687–1701.

21. Leung WWF. Dewatering biosolids sludge with varigatedecanter centrifuge. Trans Filtr Soc. 2001;1:38–44.

22. Corner-Walker N. The dry solids decanter centrifuge: capacityscaling. Filtr Sep. 2000;37:28–32.

23. Rumpus J. Scale down of recovery and dewatering in industrialcentrifuges. PhD thesis. University of London, 1998.

24. Tustian AD. An engineering study of key interactions within theprocess for antibody fragment production. Eng D thesis. Univer-sity College London, 2008.

25. Ward PN. The characterisation of biological suspensions with aview to selecting and optimizing dewatering processes. PhD the-sis. University of London, 1989.

26. Leung WWF. Torque requirement for high solids centrifugalsludge dewatering. Filtr Sep. 1998;883–887.

27. Lopes AG, Keshavarz-Moore E. Prediction of centrifugal dewa-tering of Baker’s yeast suspensions and P. pastoris broths intwo pilot-scale centrifuges using novel ultra scale-downapproach. Biotech Bioeng. 2012;109:2039–2047.

28. Lopes AG, Khan N, Liddell J, Keshavarz-Moore E. An ultrascale-down approach to assess the impact of choice of recombi-nant P. pastoris strain on dewatering performance in centrifuges.Biotechnol Prog. 2012;28:1029–1036.

29. Belter PA, Cussler EL, Hu W-S. Centrifugation. In: Belter PA,Cussler EL, Hu W-S, editors. Bioseparations Downstream Proc-essing for Biotechnology. Wiley; 1988: 47–72.

30. Wiesmann U, Binder H. Biomass separation from liquids bysedimentation and centrifugation. Adv Bioch Eng/Biotechnol.1982;24:119–171.

31. Records A, Sutherland K. Decanter Centrifuge Handbook.Elsevier Science and Technology; 2000.

Manuscript received Feb. 15, 2013, and revision received Aug. 18,

2013.

1502 Biotechnol. Prog., 2013, Vol. 29, No. 6