DESIGN OF A 3 PHASE SLURRY REACTOR FOR SOIL PROCESSING

DESIGN OF A T H R E E PHASE SLURRY REACTOR FOR SOIL PROCESSING

R. H. K L E I J N T J E N S , R. G. J . M. VAN D E R LANS and K. Ch. A. M. L U Y B E N

Kluyver Laboralory for Biotechnology, Delft University of Technology, The Netherlands

For the efflcient treatment of excavated polluted soils (particle sizes in between 1 /im and 4000 fim) in a slurry process, the coarse,

non-polluted, soil fraction has to be removed af the beginning of the process. For this purpose a tapered integrated slurry reactor/

separator was developed: the Dua l Injected Turbulent Separation ( D I T S - ) reactor. In reasonable agreement with the theoretically

derived particle response numbers for the turbulent suspension behaviour of soil, a separation between the fines and the coarse

fraction can be achieved in the D I T S - r e a c t o r . At pilot scaie (4.0 m'), the small particles ( < 700 fim) are suspended in a gas agitated

bulk compartment, while the (settled) coarse particles are fluidized in the bottom compartment. B y way of independent slurry

withdrawal from bottom and bulk, particle separation is achieved.

I N T R O D U C T I O N

Over the last decade, soil pollution has been recognized as a serious environmental problem. New remedial action technologies are therefore being developed and tested in practice'. Among the various techniques under develop­ment are biological reclamation techniques-. The major advantage of these processes is the mineralization of the (organic) pollutant into harmless compounds like bio­mass and carbon dioxide without the destruction of the soil ecosystem.

To obtain the maximum benefits microbial soil decon­tamination offers, a biotechnological slurry process for the aerobic treatment of excavated polluted soils was developed^-'', f n this continuous slurry process, polluted soils are treated in three phase (soil-water-air) suspension reactors. Experiments carried out in a mini-plant have shown that a soil residence time in the order of days is needed to obtain the required decontamination levels^ Regarding this residence time and the fact that the solids hold-up in suspension reactors''' nomally does not exceed 0.2-0.3 m^/m-\ a reasonable process capacity can only be achieved by the use of large reactors. Considering a slurry process with a capacity appropriate to soil processing (10 tons per hour), a total reactor volume of several thousand cubic meters is necessary.

Cost estimates have shown that the operation of an economic, large scale slurry process largely depends on the specific power input (P/VY. This should preferably be minimized to values as low as 50-100 W/m-\ At this low power input, not only particle suspension must be achieved, but also oxygen must be provided for the aerobic microbial decontamination.

In the search for multiphase reactors suitable for this purpose, air agitated slurry reactors were considered most appropriate. Among this class of reactors, the air agitated suspension reactor with a tapered bottom offers the best possibilities to achieve particle suspension, mixing and oxygen transfer at a low power input'^. Such a reactor, called a Pachuca reactor, was originally developed for mineral ore leaching. It consists of a cylindrical tank (ƒƒ 7 ^ 2 - 3 ) with a conical bottom'^ It was reported'-'" that Pachuca tanks at large scale (200-300 m') can be operated with a power input per volume {P V ) as low as

45 W / m \ Due to this moderate power input, a moderate oxygen transfer rate is achieved in Pachuca systems (a factor 10-100 lower than an aerated stirred vessel". Since the microbial oxygen demand for soil processing was shown to be moderate, this aspect does not influence the benefits of the Pachuca concepf*. Practical benefits of air agitated slurry tanks are the low construction and maintenance costs in combination with the absence of corrosion sensitive parts.

From previous work on air agitated suspension reac­tors (mainly bubble columns), it is known that turbulence is the driving force behind particle suspension'^. It was stated that, to keep a sohd phase in suspension, 'the turbulence must be non-decaying and the viscous dissipa­tion due to the particles must be replenished by the turbulence''^ According to this condition, slurry motion has to replenish the potential energy which is lost by the particles due to settling. To quantify the amount of work required to keep particles in suspension, one may consider first the rate of loss of potential energy of a single particle due to settling, g Ap Up, which must be replenished to keep the particle suspended. For a suspension of particles with a spatially distributed solids hold-up £s(.v, v , r ) the power input needed for suspension follows from integration over the reactor volume'':

P'" = —ilg Ap itp ƒ (E. (.v.v.r)) (.v.v.r )]d f ; (1) '/sus 0

In this equation, hindrance effects are incorporated as a function of the spatial solids hold-up distribution''*. The efficiency of suspension, (/sus. gives the fraction of total power used to sustain the solids. Assuming (/sus to be a constant, the power input needed for suspension of a given soil with known Ap. is determined by:

• the solids hold-up distribution, £5(.v.v.r); • the particle size, by way of iip .

For optimization purposes, the particle size is frequently used as a design parameter. In the mining industries, for instance, the ores are grained in a pretreatment section to an appropriate diameter, mostly dp < 100 pm\ In the processing of excavated polluted soils, grinding is not an

0957-5820/92/S05.00 + 0.00 ( Ins t i tu t ion o f Ci iemicai Engineers

D E S I G N O F A T H R E E P H A S E S L U R R Y R E A C T O R

w t %

incoming soil

f i nes

1 = 0 - 5 3 mm

2 = 5 3 - 1 0 0 u-m

3 = 1 0 0 - 1 8 0 /xm

4 = 1 8 0 - 3 5 5 ixm

5 = 3 5 5 - 5 0 0 Mrn 6 = 5 0 0 - 7 1 0 7 = 7 1 0 - 1 0 0 0 Aim 8 = 1 - 4 mm

c o a r s e f r a c t i o n

| -

I I ^ 1

f r a c t i o n number

Figure 1. Typical composition of the polydisperse (0 <dp <4000 ftm) soil used as feed in the slurry process.

appropriate pretreatment, therefore the heterogeneous, polydisperse character o f soil has to be dealt w i t h . I n Figure 1, a typical particle size d is t r ibut ion o f the excavated soil is depicted. The upper l imi t ((/p = 4 m m ) is determined by the mesh o f the finest sieve used i n the preprocessing o f excavated soil; the lower l i m i t is due to the smallest clay and humic particles ((/p < 1 ^im) that are present. As shown in Figure 1, the soil has significant contr ibut ions o f each f rac t ion in between 0 and 4000 /(m (densities varying f r o m « 2 1 0 0 k g / m ' fo r clay to a;2500 kg/m-' f o r sand).

Due to the different chemical/physical properties o f each f rac t ion '^ , the poflutant is not equally distr ibuted over the different fractions. F r o m previous experiments^ and related literature"' i t was found that more than 90% o f organic pollutants, such as oils and solvents, is preferentially adsorbed to the soil f rac t ion smaller than 100 /<m.

The inhomogeneous d is t r ibut ion o f the pol lutant offers the possibili ty to remove the coarse fractions wi thou t d is turbing the overall decontamination efficiency. T w o options are open f o r this removal:

• separation in a pretreatment unit (e.g. a hydro-cyclone); • separation in the soil slurry reactor i t se l f

The second opt ion is considered beneficial since an integrated separation/reactor system reduces the number o f uni t operations, which results in lower costs. I n addi t ion , the coarse f rac t ion also is exposed to the treatment. The latter adds to the decontamination eff i ­ciency since some o f the (polluted) fines may be adsorbed on to the coarse f rac t ion .

The a im o f the research presented in this paper was the development o f an integrated soil slurry reactor in which the treatment o f polluted fine soil particles is combined w i t h the separation o f the coarse f rac t ion . T o provide a

basis fo r the integrated reactor design, theoretical aspects o f turbulent particle suspension are discussed first.

T H E O R E T I C A L I N V E S T I G A T I O N O F T U R B U L E N T P A R T I C L E S U S P E N S I O N

T w o conditions w i t h respect to turbulent particle suspension can be formula ted:

(1) each particle must be able to f o l l o w the turbulent l iqu id m o t i o n " ; (2) the turbulence must be non-decaying and sufficient in intensity to replenish the continuous energy dissipation due to the settling o f the suspended part ic les '^ ' ' l

The m o t i o n o f particles in turbulent flows has been studied intensively in the last decades'^-"'^"-^'. I n tu rbu­lent flows, particle suspension is a result o f the turbulent l iqu id mo t ion which exerts a drag force on the particle, a force which frequently changes i n direction. Therefore, the interaction between fluid and particle w i l l only result in a complete suspension i f the particle is able to f o l l o w the l iqu id m o t i o n to a large extent. Being a crucial aspect o f turbulent suspension reactors, the mobi l i ty o f a particle in a turbulent flow is taken as the starting point f o r the design o f the slurry reactor/separator.

A n indication o f the abi l i ty o f the particles to f o l l o w turbulent m o t i o n is f o u n d i n the rat io o f a particle response time and a characteristic time for turbulent flow. This rat io (denoted as the particle response number) , facilitates the hydrodynamic description o f turbulent particle suspension since the two mechanisms domina t ing the interaction between fluid and particle are captured in one parameter. I n theoretical studies focusing on particle behaviour under turbulent condit ions, the particle res­ponse number is frequently used as an inertia parameter in computer simulations-''^^.

The particle response time can be calculated f r o m a force balance over a discrete particle in a turbulent flow according to Newtons law o f m o t i o n " :

(2)

Equat ion (2) expresses the rate o f change in time o f the particle velocity relative to the bulk velocity (upi — Uf,). I n the acceleration, bo th the particle mass and the v i r tua l mass o f the displaced l i qu id are considered. The term E F contains the flow resistance, gravity, pressure forces, rotat ional forces and other forces acting on the particle. F r o m these forces, only the flow resistance (drag force) is taken in to account, being the most impor tan t term.

Fo l lowing H i n z e " , the particles are assumed to react according to the creeping flow (Stokes) regime. This gives a drag coefficient o f Cd = 24/Rep. Us ing this expression the drag force becomes:

F, drag

24 4 dpi Pi: ("pi - "f i )" (3)

I f this drag force is substituted f o r S Fi i n the r ight hand side o f equation (2) , the equation can be wr i t ten i n a differential f o r m having the relative Reynolds particle number as a variable. The solut ion o f this equation relates the change i n the relative particle Reynolds number to a

Trans IChemE, Vol 70, Part B , May 1992

;!6 K L E I J N T J E N S et al.

specific t ime interval. Tak ing a time interval i n which the

Reynolds particle number changes by a factor 2 to be

characteristic (the hal f value t i m e " ) , the f o l l o w i n g rela­

t ion results f o r the response time tpi

0.04 (4)

I n equation (4) , rj( is the fluid dynamic viscosity. The

term between brackets refers to the amount o f mass

considered i n the response time, pp accounts f o r the

particle and fip^ fo r the v i r tua l added mass (amount o f

fluid displaced as the particle moves). The particle

response time can be seen as a measure f o r the particle

inertia in a l iqu id flow.

I t must be noted that only particles smaller than 100

are in the creeping flow regime assumed above. I n view o f

Figure 1, this means that Hinze's der ivat ion is only val id

f o r the fines in the soil. T o derive the response number fo r

the soil particles outside the creeping flow regime, the

standard correlat ion curve f o r Cé=f {Re^) is used to

correct the results^-'.

For the def in i t ion o f a characteristic turbulent time the

rate determining step in the turbulent power conversion

process must be taken in to account. T w o parameters

specifying a class o f eddies at the top o f the turbulent

energy spectrum are considered, the length o f the energy

containing eddy in the system, 4, and its turbulence

intensity, u'. I n an isotropic turbulent flow, the f o f l o w i n g

equation f o r the kinetic energy transfer rate per mass (et)

holds:

(5)

Equa t ion (5) is regarded as 'the first law o f turbulence'

governing the turbulent power conversion^'*. The two

crucial parameters, and 4, are also used to describe

many other turbulent quantities such as the eddy visco­

sity^^. Based on these two parameters, the characteristic

time f o r the rate o f change in a turbulent flow can be

estimated as:

(6)

Equat ion (6) is derived f r o m dimensional analysis". I n

order to estimate values f o r T I , i t is necessary to substitute

u' and 4 by measurable reactor parameters. For this

purpose, the rat io a between the eddy size (4 ) and the

characteristic length o f the system ( L ) is considered.

Generally the size o f the largest eddy is used f o r the

characteristic length o f the system, which may be esti­

mated to be about 0.5 Tïor the tapered f o r m used (80%

o f maximal diameter o f circle in equilateral triangle^''). I t

was f o u n d previously that, i n the suspension reactor, the

eddy length is both a func t ion o f scale and solids hold-up

under the condi t ion o f min imal suspension"*. The eddy

length varied f r o m 4 = 0.15L (lab-scale, 0.4 m-') to

4 = 0.06 L (pi lot plant scale, 4.0 m ' reactor) f o r a slurry

density o f 10 w t % . For a slurry density o f 40 w t % , the

rat io 4 / i varied f r o m 0.05 ( 0 . 4 m ' ) t o 0 . 0 1 7 ( 4 . 0 m ' ) . T h e

rat io a thus appears to be a func t i on o f scale and slurry

densities under min ima l suspension condit ions. In order

to substitute (/' i n equation (6) , the specific turbulent

energy transfer rate, e,, is approximated by the power

input per volume {PjV) divided by the fluid density.

Equat ion (5) therefore reads as:

Pr (7)

The general expression f o r the turbulent time now reads

as a combinat ion o f equations (6) and (7):

,2/3 Pr

PjV

1/3

(8)

Combin ing equations (4) and (8) , the final expression

fo r the response number now reads as:

T p / T , » 0 . 0 4 a - 2 / 3 {Pp + Pp,}d^ {PIV)

'/r Pr 1/3 L 2/3 (9)

I n equation (9) , the response number is spfit in to three

parts. The first part contains a and represents the change

in turbulence characteristics o f the flow. The second part

contains slurry properties, while in the th i rd part , the

reactor size and the specific power input are present.

The significance o f tp/t t in the soil reactor design is

demonstrated by model calculations showing the number

as a func t ion o f the particle diameter (Figure 2) . The

f o l l o w i n g parameter values were used: pp = 2500 k g / m '

(sand), pi = 1000 k g / m ' (water), j ; r=0 .001 Pa.s a n d "

/? = 0.66. The response numbers have been calculated f o r

two small scale reactors, having L equal to 0.01 and 0.1 m ,

and two large scale reactors w i t h L equal to 1.0 and 5.0 m .

Using an average slurry density o f 20 w t % , p under

min imal suspension condit ions was estimated to be equal

to 0.08 at small scale (compare also Reference 27) and

0.03 at p i lo t plant and at large scale.

100

1 0

p/ I

P / V = 50 W / m ^

character is t ic reactor length (m):

0.01

0 .001

O.OO01

O.OOOO 1 10 100 1 0 0 0 1 0 0 0 0

p a r t i c l e s ize ( u m )

Figure 2. Theoretical curves for the particle response number as a function of particle size and reactor scale.

Trans I C h e m E , Vol 70, Part B , May 1992

D E S I G N O F A T H R E E P H A S E S L U R R Y R E A C T O R 87

Since only low power inputs are o f interest, calculations were carried out f o r a specific power input o f 50 W / m ' . Scale effects (normal ly P / F i s larger fo r smaller scales) were not taken in to account. For values o f T p / T t = 1, the particle reaction time is equal to the characteristic turbulent time. I n Hne w i t h the statement made by H i n z e " that, 'tp must clearly be smaller than T I to obtain a high degree o f particle d i f fus ion ' , the cri ter ion fo r the turbulent suspension regime has been taken as i p / t i < 0.1 . For these values, the mobi l i ty o f the particle is large enough to allow fo r a fast adjustment to changes in the turbulent flow. For values o f Tp/t t > 0 . 1 , the inertia o f the particle is not negligible, meaning that the particle is not able to fo l l ow the turbulent flow any more. I t is therefore concluded that fo r practical slurry flows, the settfing regime corresponds

to Tp/x , > 0 . 1 . The model calculations clearly show the influence o f

the particle diameter and reactor scale on the response number. A t smafl reactor scales, the change in regime takes place at a particle diameter o f 150-400 ^ m . A t larger scales the regime change can be expected f o r particles w i th a diameter o f 900-5000 j tm. I t may therefore be concluded that even at large scales, the coarse soil f rac t ion is unl ikely to be kept in suspension (settling regime). The new soil slurry reactor is based on this difference in suspension behaviour.

D U A L I N J E C T E D T U R B U L E N T S E P A R A T I O N R E A C T O R

A reactor was designed i n which the dr iv ing force behind the suspension, turbulent mo t ion , is integrated w i t h the efficient removal o f settled material f r o m the reactor bo t tom. Simultaneous inject ion o f air and slurry at the bo t t om o f a tapered vessel was adopted f o r this purpose. I n the D u a l Injected Turbulent Separation

slurry withdrawal

Figure 3. Schematic representation of the Dual Injected Turbulent Separation (DITS-) reactor.

( D I T S - ) reactor, two compartments can be distinguished (Figure 3):

• a zone w i t h ffuidized coarse material at the bo t tom; • an air-agitated, three-phase suspension o f fine material i n the bulk o f the reactor.

Because o f the tapered shape, the upward superficial slurry velocity rapidly decreases w i t h increasing height. Therefore ffuidization is only achieved f o r the settled particles at the bo t tom. Due to the ffuidization o f the settled particles, blockage o f the inject ion is prevented (air and slurry inject ion proceeds wi thou t d i f f i c u l t y ) .

H a v i n g established two compartments, s lurry can be w i t h d r a w n f r o m these compartments independently. I n this way, the coarse f rac t ion can be continuously separ­ated f r o m the fines, which are transported f r o m the D I T S -reactor to a cascade o f slurry reactors f o r fu r ther treatment. Since settlement problems are not expected, the second stage is equipped w i t h air agi tat ion only. I n

I CÏD-t^

of f gas t rea tment

soil in

6

D. I .T .S reactor! fines

<a>-̂

cascade of! air agi tateci slurry tanks^jgfines

s lurryX A recycleM 1*1

compressed air

•>soil out

remix coarse + fine fractions

Figwe 4. Simplified flowsheet of the biotechnological slurry process for the decontamination of excavated soils.

Trans IChemE, Vol 70, Part B , May 1992

88 K L E I J N T J E N S el al.

Figure 4 a simplified flowsheet is shown o f the integration o f the DITS-reactor and the cascade.

Considering the large scale o f the air agitated cascade reactors, the use o f a series o f large tall conical tanks does not seem opt imal . Therefore a tapered 'channel-like' reactor was designed w i t h a relative low height to w i d t h rat io. I n these reactors, air is injected by way o f a fine o f inject ion manifolds at the bo t tom, along the length o f the reactor. The DITS-reactor , as shown schematically i n Figure 4, is designed as a tapered shaped unit-operat ion in which separation and microb ia l processing are inte­grated. I t becomes clear f r o m the flowsheet that the performance o f the DITS-reactor largely determines the power input needed i n the second stage. A n efficient removal o f coarse material minimizes the power input fo r the cascade o f air agitated reactors. The process econo­mics largely depend on the performance o f the D I T S -reactor.

E X P E R I M E N T A L

Optimization of the Reactor Geometry (Model System)

I n order to determine the op t imal reactor geometry, suspension experiments were carried out i n a gas agitated, flat, scale perspex model o f a tapered channel o f 0.4 m ' (Figure 5, f o r dimensions see Table 1). A flat scale model, w i t h a small depth, was considered appropriate since the ma jo r bulk flow takes place in the vertical reactor plane"*. T o be able to study suspension pheno­mena i n an unambiguous manner a model solid phase (quartz sand, = 300 pm and ps = 2460 k g / m ' ) was used.

I n order to investigate the suspension behaviour at different geometrical configurat ions and volumes, the reactor was equipped w i t h two moveable vertical plates (Figure 5) . Di f fe ren t geometries could be created by

Tdhle I. Dimensions of lapered reactors used.

r o t a m e t e r - l -pressure gouge

a i r i n j e c t i o n

Figure 5. Experimental set-up of t i ie perspex air agitated reactor used for the model experiments. Due to the moveable plates, the geometry and scale can be changed f rom 0.04 to 0.16 m^.

Scale Height Width Depth Injection (m-̂ ) (m) (m) (m) points

0.400 1.180 1.270 0.45 20 4.000 2.700 2.700 1.00 45

varying the sli irry height ( H ) i n combina t ion w i t h the reactor w i d t h (T ). The depth, D, o f the reactor was fixed i n each experiment and equal to the length o f the inject ion groove (0.45 m ) . Gas was introduced at the b o t t o m by means o f an inject ion man i fo ld w i t h equidistant in jec t ion points made o f vertical tubes w i t h an inner diameter o f 2 m m and a mutua l distance o f 2 cm. I n order to circumvent maldis t r ibut ion in the man i fo ld , the cross sectional area o f the man i fo ld was chosen 3 times larger than the total inject ion surface area, as given by the sum o f the inject ion openings^*.

The op t imal geometry was considered to be that geometry requir ing the lowest power input to main ta in a state o f min ima l suspension. The state o f m i n i m a l suspension was detected by visual observation o f the (mono-disperse) particles near the reactor b o t t o m . M i n i ­mal suspension was defined as that state o f suspension in which a collar o f slow slipping sand appeared just above the inject ion man i fo ld . The cri t ical gas inpu t was deter­mined by a stepwise decrease in gas input , s tart ing f r o m complete suspension. Using this method the standard deviat ion i n the measured crit ical gas inpu t was about 5%.

The gas flow was measured at the m a n i f o l d w i t h a rotameter. The pressure registered in the rotameter was used to express the gas flow f o r atmospheric conditions^' . Together w i t h the slurry height and density, the gas flow is used in the w o r k equation f o r rising bubbles, i n order to calculate the power input in the system'".

(10)

I t can be shown that the kinetic con t r ibu t ion ( ( / ) g i / g ) is negligible'". Pressure losses i n pipes and the injector itself are not considered; only the power input present i n the gas phase at the moment o f entrance is accounted f o r i n equation (10).

T o determine the opt imal reactor geometry, particle suspension, in a 10 w t % (0.041 v / v ) slurry, was studied at fou r different volumes, varying f r o m 0.040 m ' up to 0.160 m ' . For each volume, fou r H/Tralios ( 1 , 2, 3 and 4 ) were tested.

DITS-Reactor

T o test the principle o f the DITS-reactor and its separation performance, a 4.0 m ' p i lo t plant (10% o f f u l l scale design) was designed according to the op t imal geometry as determined f o r the model system {H/T= 1) (see Results and Discussion section and Table 1). The reactor was made using steel plates w i t h some perspex view plates. Detailed i n fo rma t ion is given by Kleijntjens"* (available f r o m the corresponding author) . T w o design

Trans IChemE, Vol 70, Part B , May 1992

D E S I G N O F A T H R E E P H A S E S L U R R Y R E A C T O R 89

f \ l

s lur ry pumped t o i n j ec to r

Figure 6. Schematic representation of the solid-liquid separator used in the slurry recirculation loop of the DITS-reactor.

features were crucial f o r the operation o f the D I T S -reactor:

• the l iquid/sol id separator in the slurry inject ion loop, • the dual (g— 1) inject ion man i fo ld .

I t was found that, despite the fact that the slurry was drawn f r o m the top o f the reactor, the dual in ject ion man i fo ld was blocked by the larger particles present in the recirculated slurry. The cri t ical particle size was deter­mined empirically to be about 100 /<m. Therefore a l i q u i d / solid separator was added i n the design.

Liquid!solid separator

The shape o f the l iquid/sol id separator was determined by the fact that i t had to be placed on top o f the p i lo t reactor. This settler posi t ion was chosen i n order to facili tate the return o f the settled f rac t ion to the reactor. The final design was a small and deep, two compartment, settler (height 1.2 m ) w i t h a total volume o f 0.530 m ' . Three sections can be recognized, shown schematicafiy in Figure 6:

(1) Vert ical sieve plate (mesh 1 m m , length 0.7 m , w i d t h 0.4 m , angle to the vertical 5°) . For the removal o f floating fibrous and organic material ; the removed material is directed back to the reactor. (2) Inert ia settler. This section is designed as a pretreat­ment o f the upf low section, removing the heavier sand f rac t ion; the underf low o f this section opens directly in to the reactor; the volume o f this tapered section is 0.160 m ' .

(3) U p f l o w settler. For the removal o f the remaining soil particles w i th a diameter above 100 pm; the underf low is also directed to the reactor; the volume o f this section is 0.370 m\

The slurry flow in the l iquid/sol id separator is schemati­cally depicted in Figure 6. Slurry f r o m the reactor is pumped to the top (a) o f the vertical sieve plate (1). Fibrous material , removed by the sieve (1 m m ) is returned to the reactor (b) , while the sieved slurry flows cont i ­nuously to the inertia settler (c). I n this settler (2) , the larger particles are removed by way o f flow d, while the fines are fur ther transported (e) to a second settler. I n (3)

Figure 7. Three dimensional representation of the 4 m^ pilot DITS-reactor.

particles not sustained by the flow w i f l settle and leave the system (f low f ) . T o maximize the settler efficiency, bo th flows d and f are minimized. The top product o f (3) is pumped straight to the slurry injector (g) . I n Figure 7 a three dimensional representation o f the settler section and the DITS-reactor is given.

Injector

T o be able to inject air and recycled slurry simulta­neously at the b o t t o m o f the tapered reactor, a special in ject ion man i fo ld was developed. I n the m a n i f o l d , slurry is distr ibuted across vertical in ject ion points, i n a similar fashion to that described f o r the gas injector used in the model system (Figure 5) . The ma jo r problems w i t h the slurry inject ion were particle settlement and maldis t r ibu­t ion . The latter was circumvented by choosing a cross sectional area 3 times larger than the to ta l in ject ion surface area^^. The slurry m a n i f o l d was designed f o r a slurry feed in which all particles above 100 ;tim are removed. The gas-liquid injector was developed w i t h 40 equidistant injecdon points, each constructed as an inner air jet , surrounded by outer slurry jets'".

Since only high slurry densities were o f interest i n the development o f the DITS-reactor , its performance at 30 and 40 w t % soil hold-ups was invesdgated. First , i t was empirically determined which combinadon o f gas and slurry inject ion flows resulted in the simultaneous exis­tence o f a fluidized bed at the reactor b o t t o m , and a turbulent suspension o f fines in the bulk . H a v i n g achieved two compartments, the gas and slurry in jec t ion flow rates were fur ther minimized to obtain a m a x i m u m size bo t t om compartment (visual observation). Subsequendy, sam­ples were taken f r o m the bed (15 1 f r o m a drain-cock at 0.05 H at the centre line o f the vertical side) and the bulk (60 1 at height 0.65 H). Particle size dis tr ibut ions were established by wet sieving o f the slurry samples, using a set o f sieves wi th decreasing mesh (see Figures 8 and 9) .

R E S U L T S A N D D I S C U S S I O N

T o establish an opdmal reactor geometry, sixteen experiments were carried out focusing o n the min ima l

Trans IChemE, Vol 70, Part B , May 1992

9 ü K L E I J N T J E N S et al.

I ' i ^ bottom

I I bulk

30wt%; 4.Om'

1 = 0 - 5 3 um 2 = 5 3 - 1 0 0 Mm

3 = 1 0 0 - 1 8 0 Mm

4 = 1 8 0 - 3 5 5 Mm

5 = 3 5 5 - 5 0 0 Mm

5 = 5 0 0 - 7 1 0 Mm

7 = 7 1 0 - 1 0 0 0 Mm

8 = 1-4 mm

J L 4 5 6

f ract ion number

Figure 8. Composition of the bottom and bull? compartment for the 4.0 m ' DITS-reactor, having a solids hold-up equal to 30 wt%.

3 0 0

W / m ^

2 4 0

1 8 0

P / V

1 2 0

6 0

0

O.OO

1 0 W t % sand

-

+ ...

" " ^ - ^ +

— 1 - - H / T = 1

- - A - - H/T=2

—e- - H/T=3

+• H/T=4

0 . 0 4 0 . 0 8

reactor

0 . 1 2 0 . 1 6 0 . 2 0 3

voluine ^ Figure lU. Measured power input per volume (PjV) in the model system (gas agitation only) as function of geometry and scale.

1 = 0 - 5 3 Mm

2 = 5 3 - 1 0 0 Mm • 18 ' 3 = 1 0 0 - 1 8 0 Mm

4 = 1 8 0 - 3 5 5 Mm

w t % 5 = 3 5 5 - 5 0 0 Mm

6 = 5 0 0 - 7 1 0 Mm 36 '7 = 7 1 0 - 1 0 0 0 Mm

8 = 1-4 mm

40 wt%; 4.0 m=

J E L f ract ion number

Figure 9. Composition of the bottom and bulk compartment for the 4.0 m ' DITS-reactor, having a solids hold-up equal to 40 wt%.

power input at different geometries o f the tapered model reactor. The specific power input is given i n Figure 10. The Figure shows that at H/T= 1 the specific power input is m in ima l f o r each volume.

T o investigate the separadon performance o f the DITS-reactor , experiments were carried out at a scale o f 4.0 m ' . Four items concerning the reactor performance were studied:

• the inject ion flow rates and the power input ; • the injector performance;

separadon performance; the sol id / l iquid separator.

A gas inject ion flow rate o f min ima l 0.47 m ' / m i n and a slurry inject ion flow rate o f m i n i m a l 0.14 m ' / m i n were measured to achieve fluidization o f the coarse materials and suspension o f the fines simultaneously f o r a solids hold-up o f 30 w t % . For the 40 w t % slurry, the gas and slurry flow rates were 0.43 m ' / m i n and 0.15 m ' / m i n , respectively. The cri ter ion f o r min ima l flow o f bo th gas and slurry was defined in line w i t h the min ima l suspension cri ter ion: no stagnant zones were allowed just above the in jecdon man i fo ld . Tak ing the gas input responsible f o r the turbulent agitadon o f the bu lk , equat ion (10) was used to determine the power input i n this compartment. F o r the 30 w t % and 40 w t % slurries power inputs o f 191 W and 206 W were calculated corresponding to specific power inputs (P/V) in the bulk compartment o f 48 W / m ' and 51 W / m ' , respectively.

Besides the flow rates, a ma jo r parameter relevant to the performance o f the injector was the pressure, mea­sured at the man i fo ld . Bo th the gas and slurry manifolds were moni tored separately. The gas injector was operated at an overpressure o f about 0.4 bar. T o inject the slurry, a pressure build-up o f about 0.7 bar was required. Tak ing the pressure build-up due to the slurry height i n the reactor as 0.3 bar, the pressure loss i n the gas injector was about 0.1 bar. I n the slurry injector, the pressure loss was about 0.4 bar. Since the pressure in the slurry m a n i f o l d itself d id not increase dur ing operation, bo th the removal o f the particles and the fibrous materials was in accord­ance w i t h requirements. I f the reactor was operated wi thou t a sofid-l iquid separator, pressure increased w i t h i n 2 hours up to 3-5 bar, fo l lowed by complete blockage o f the man i fo ld .

T o investigate the separadon capacity, w i t h respect to the coarse f r acdon in the 4.0 m ' p i lo t plant , samples were taken f r o m the bo t t om and the bulk compartment slurry and analysed f o r their particle size d is t r ibut ion . I n Figures 8 and 9 the particle distr ibutions o f bed and bulk

Trans I C h e m E , Vol 70, Part B , May 1992

D E S I G N O F A T H R E E P H A S E S L U R R Y R E A C T O R 91

are given f o r 30 and 40 w t % , respectively, while in Figure 11 the corresponding cumulative particle distr ibudons are depicted. For bo th slurry densities, the d i s t r ibu t ion in the bulk samples differs f r o m the d is t r ibut ion i n the bed samples. Particles larger than 700 nm are absent in the bulk samples, while the fines are hardly present i n the bed. The difference in the bo t tom curves is at t r ibuted to the difference in accumulation o f the most coarse f rac t ion at the bo t tom o f the bed at different slurry densities, while the sampling poin t is f ixed.

I t may be concluded that the separadon i n the D I T S -reactor meets the design purpose: the larger uncontami-nated particles which require a large power input f o r suspension may be removed, while only a negligible amount o f the contaminated fines leaves the system untreated.

T o determine the settler efficiency, the slurry flows pumped to the settler were considered (note po in t a i n Figure 6) . F o r 30 and 40 w t % , these slurry flows were 0.273 and 0.402 m ' / m i n respectively and the correspond­ing settler efficiencies (injector flow rate/slurry w i t h ­drawal flow rate) were 0.51 and 0.37. W i t h regard to the removal o f the solids, the settfing performance was as expected; no particles above 100 ^ m were detected in the overf low at po in t g. The latter demonstrates that the upf low pattern i n the second settler suited its purpose.

Experimental Suspension Criterion

T o compare the experimental suspension behaviour in the p i lo t plant w i t h the theoretical predicdon, the results given i n Figure 11 are compared w i t h the predicted values given i n Figure 2. This comparison can be made since the expenmental P/V (48 and 51 W / m ' ) is close to the P/V used i n the model calculations (50 W / m ' ) . The regime change in the p i lo t plant is predicted at </p « 1000 pm (note the curve f o r a characterisdc length o f 1 meter in Figure

w t %

60

40

20

10 100 lOOO 10000

p a r t i c l e s i ze (m ic romete rs )

Figure 11. Cumulative particle distribution of bed and bulk. Data of Figures 8 and 9.

2) . This predict ion comes close to the dpKlQQ / i m at which the experimental regime change takes place. Regarding the diff"erence between the theoretical and the measured transidon particle size, i t should be recognized that the theory was derived f o r one single particle i n a turbulent flow, while experiments were carried out f o r dense slurries.

C O N C L U S I O N S

F r o m the theoretical and experimental results derived i n this paper the fo l l owing can be concluded:

• Due to the negligible amount o f po l lu t i on adsorbed to the coarse soil f rac t ion , i t is beneficial to separate this f rac t ion f r o m the polluted fines (^/< 100 /xm) at the beginning o f the slurry process. Integrated separation and soil processing can be achieved in a tapered D u a l Injected Turbulent Slurry (DITS-)reactor . Pi lot scale experiments (4.0 m ' ) showed that, f o r a 30 and 40 w t % slurry density the coarse soil fracdons {d^ > 700 ^<m) can be separated f r o m the fines.

• Assuming that 'a h igh degree o f particle d i f fus ion can only be obtained i f the particle response dme is smaller than the turbulent t i m e ' " , the theoretical regime transi­t ion f r o m turbulent suspension to settling is taken at Tp / t i « 0 . 1 . This cr i ter ion is conf i rmed by the experiments carried out in the 4 m ' DITS-reactor .

• For a 40 w t % slurry density, the experimentally measured power input f o r gas in the DITS-reactor was 206 W . Since the power input f o r the recycfing o f the slurry is only about 10% o f this value"*, this results i n the economically attracdve low P/ V f o r the reactor bu lk o f about 50 W / m ' .

• As opt imal geometry f o r the tapered reactors used in the soil slurry process, H/T= 1 was found .

N O T A T I O N

c, drag coefficient D depth of tapered reactor, m d diameter, m F force, N

g gravitational constant, 9.8 m/s' H slurry height in reactor, m L characteristic length of reactor, m

u length of energy containing eddy, m

p pressure, Pa p power, W

power input, W PIV specific power input, W/m^ Re^ relative Reynolds particle number dp pr (i/pi — t time, s T maximal width of tapered reactor, m u velocity, m/s u' turbulent liquid fluctuation velocity, m/s

c,. particle volume, m '

Vr reactor volume, m ' X coordinate in the depth of the reactor, m

y coordinate in the width of the reactor, m z coordinate in axial direction, reactor height, m

Greek .symbols a turbulence ratio /e/L

P virtual coefficient

y scale and particle size dependent coefficient ',z) local solids hold-up, ra' solids/m' reactor

turbulent kinetic energy transfer rate, W/kg dynamic viscosity, Pa.s

Trans I C h e m E , Vol 70, Part B , May 1992

92 K L E I J N T J E N S et al.

(/sus suspension efficiency (j> volumetric flow rate, m'/s p density, kg/m' hp density difference between solid and liquid phase, kg/m-Tp particle response time, s T, characteristic turbulent fluctuation time, s

Super- and subscripts drag drag force action at particle e eddy f fluid phase, water g gas phase, air i specific direction inj injector " atmospheric conditions p particle s solid phase sl slurry 00 in infinite medium, used for terminal settling velocity

R E F E R E N C E S

1. Wolf, K, , Van den Brink, W. J, and Colon, F, J, (eds), 1988, Conlaminated Soil '88, Second Int TNO-BMFT Conference on Conlaminaled Soil (Kluwer Academic Publishers, Dordrecht),

2. Soczo, E, R. and Staps, J. J, M . , 1988, Review of biological treatment techniques in the Netherlands, in Conlaminaled Soil '88, Wolf, K . , Van den Brink W. J. and Colon, F, J. (eds) (Kluwer Academic Publishers, Dordrecht), 663-670,

3. Kleijntjens, R. H . , Meeder, T. A„ Geerdink, M , J. and Luyben, K . Ch. A . M . , 1990, Design of a slurry process for the decontaminadon of excavated polluted soils, in Conlaminaled .loil '90, Arendt, F., Hinsenveld, M . and Van den Brink, W. J. (eds) (Kluwer Academic Publishers, Dordrecht), 997-998.

4. Kleijntjens, R. H . , 1991, Biotechnological slurry process for the decontamination of excavated polluted soils (PltD Tliesis, Delft University of Technology).

5. Kleijntjens, R. H . , Smolders, A . J. and Luyben, K . Ch. A . M . , 1989, Technological and kinetical aspects of microbial soil decontamina­tion in slurry reactors on mini plant scale, Proceedings of the 3rd Inlernational Conference NATO/CCMS Pilol Study, Montreal, November 1989.

6. Nienow, A . W., 1968, Suspension of solid particles in turbine agitated baffled vessels, Chem Eng Sci, 23: 1453-1459.

7. Hallett, C. J., 1981, Oxygen mass transfer in Pachuca tanks {PhD Thesis, Imperial College of Science and Technology, University of London).

8. Bos, P., Huber, T. F., Luybenm, K Ch A. M . and Kuenen, J. G., 1988, Feasibility of a Dutch process for microbial desulphurization of coal, Resources, Conservation and Recycling, 1: 279-291.

9. Merriman, A. D. , 1958, A Dictionary of Metallurgy (McDonalds Evans, London).

10. Lamont, A . G. W., 1958, Ai r agitation and Pachuca tanks. Can J of Cliem Eng, 36: 153.

11. Joosten, G. H . E., Schilder, J. G. M . and Janssen, J. J., 1977, Chem Eng Sci, 32: 563-566.

12. Narayanan, S., Bhatia, V. K . and Guha, D . K . , 1969, Suspension of solids by bubble agitation. Can J of Chem Eng, 47: 360-364,

13. Soo, S. L . , 1955, Statistical properties of momentum transfer in two phase flow, Cliem Eng Sci, 5: 57-67.

14. Barnea, E. and Mizrahi, J., 1973, Part 1, General correlation for

fluidization and sedimentation in solid multi-particle systems, Chem EngJ, 5: 191-199.

15. Brady, N . C , 1984, Tlie Nature and Properties of Soils, 9th edition (Macmillan Publishers Comp, New Y o r k ) .

16. Werther, J. and Wilichowski, M . , 1990, Investigation of the physical iTiechanisms in the purification of soils by washing processes, in Conlaminaled sod '90, Arendt, F., Hinsenveld, M . and Van den Brink, W. J. (eds) (Kluwer Academic Pubhshers, Dordrecht), 907¬920.

17. Hinze, J. C , 1972, Turbulent fluid and particle interaction, Progress in Heat and Mass Transfer, 3: 433^52.

18. Tchen, Chan-Mou, 1947, Mean value and correlation problems connected with the motion of small particles suspended in a turbulent fluid {PhD Thesis, Delft University of Technology).

19. Davies, J. T., 1986, Particle suspension and mass transfer rates in agitated vessels, Chem Eng and Processing, 20: 175-181.

20. Hetsroni, G., 1982, Handbook of Multiphase Systems (Hemisphere Publishing Company, Washington).

21. Squires, K . D . and Eaton, J. K . , 1990, Particle response and turbulence modification in isotropic turbulence, Pliys Fluids A, 3: 1190-1203.

22. Fung, J. C. H . and Perkins, R. J., 1989, Particle trajectories in turbulent flow generated by true varying random Fourier Modes, in Advances in Turbulence 2, Fernholz, H . H . and Fiedler, H . E. (eds) (Springer Verlag Berfin).

23. Bird, R. B., Stewart, W. E. and Lightfoot, E. N . , 1960, Transport phenomena (John Wiley and Sons, Inc, New York ) .

24. Batchelor, G. K . , 1953, Tlie Tlieory of Homogeneous Turbulence (Cambridge University Press, Cambritlge).

25. Hinze, J. O., 1959, Turbulence (McGraw-Hif l , New Y o r k ) . 26. Joshi, J. B., 1980, Axial mixing in multiphase contactors: a unified

correlation. Trans Inst Cliem Eng, 58: 155-165. 27. Pandit, A . B. and Joshi, J. B., 1984, Three phase sparged reactors,

some design aspects. Rev Cliem Eng, 2: 1. 28. Dow, W. M . and Shreveport, L . A. , 1950, The uniform distribution

of a fluid flowing through a perforated pipe, J AppI Mech, 17: 431¬438.

29. Coleman, M . C , 1956, Variable area flow meters, Trans Inst Cliem Engrs, 34: 339-350.

30. Luyben, K . Ch. A. M . and Kleijntjens, R. H . , 1988, Werkwijze en inrichting voor hel scheiden van vaste stoffen, Dutch Patent, no: 8802728.

A C K N O W L E D G E M E N T

The authors are grateful to Professor F. T. M . Nieuwstadt for his theoretical support, and to Michiel Bosse, Mark de Groot, Hilda van de Laar, Rene Langevoort and Yvonne van Voorst for their cooperation in the project. The authors also are grateful to G. van der Tooien and P. Vetter for their indispensable technical assistance. Financial grants were given by Delft University of Technology, The Netherlands Integrated Soü Research Programme, Ministry of Education and Science.

A D D R E S S

Correspondence concerning this paper should be addressed to Dr Ir . R. G. J. M . van der Lans, Kluyver Laboratory for Biotechnology, Faculty of Chemical Technology and Materials Science, Delft Univer­sity of Technology, Julianalaan 67, 2628 BC Delft , The Netherlands.

Tlie manuscript was received 4 June 1991 and accepted for publication after revision 18 February 1992.

Trans IChemE, Vol 70, Part B , May 1992