Sei sulla pagina 1di 11

Separation and Purication Technology 16 (1999) 193203

A novel otation column for oily water cleanup


Xuqing Gu 1, Shiao-Hung Chiang *
Chemical and Petroleum Engineering Department, University of Pittsburgh, Pittsburgh, PA 15261, USA Accepted 22 January 1999

Abstract A novel multi-stage loop-ow otation column (MSTLFLOA) was developed for ne particle separations. The unique design feature of the MSTLFLO column is the use of draft-tubes to achieve multi-stage conguration and minimize back-mixing/entrainment in otation. As a potential application, the column was studied for its performance in separating emulsied oil from water. The separation eciency was found to be highly dependent on the unique hydrodynamic conditions provided in the column. The separation kinetic constants were therefore correlated with the hydrodynamic parameters, including gas holdups, Sauter bubble sizes and liquid circulation rates. A classic tank-inseries model was successfully employed in the modeling of the process. 1999 Published by Elsevier Science B.V. All rights reserved.

Nomenclature A AR C D d 32 F J K N u N ORE Q RF T V Area of the riser or downcomer (m2) Area ratio of riser to downcomer Concentration (ppm) Diameter of bubbles Sauter mean diameter Feed rate (m3 s1) Supercial gas velocity (m s1) kinetic constant Circulation number Order of kinetics Oil removal eciency (%) Volumetric ow rate of liquid circulation (m3 s1) Rate of otation Time (s) Volume (m3)

Subscripts 0 Initial 2 Innity b Bubble B Bottom C Circulation d Downcomer D Dispersion i The ith L Liquid r Riser 1. Introduction Oil-polluted wastewater is a very common occurrence. Oil-bearing euents come from a wide range of sources: petroleum reneries, military maintenance, paper de-inking operations, meat processing, chemical processing and manufacturing plants, barrel and drum cleaning, industrial central laundries, iron and steel plants, soap manufacturing, soybean processing, mill waste, aluminum forming and others [1,2]. Direct discharge of

Greek Letters e Gas holdup t Mean residence time (s)


* Corresponding author. 1 Current address: E.I. DuPont de Nerours and Company, Jackson Lab., Deepwater, NJ 08023, USA.

1383-5866/99/$ see front matter 1999 Published by Elsevier Science B.V. All rights reserved. PII: S1 3 8 3- 5 8 66 ( 9 9 ) 0 00 0 4 -0

194

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

industrial wastewater must comply with the limits imposed by the US Environmental Protection Agency ( EPA) on the oil and grease content of the euents. From economic viewpoint and environmental concern, current total oil and grease restriction for discharge of treated water is already down to 40 ppm or lower, and more stringent discharge regulation is expected in the near future. Various separation techniques for oily water cleaning have been proposed in the last couple of decades [3,4]. The most commonly used techniques include: 1. otation (mostly dissolved air otation); 2. gravitational methods; 3. chemical treatment; 4. membrane methods ( Ultra-ultration); 5. biological treatment; and 6. combinations of the above. However, the applications of these techniques have been limited owing to both technical and economic reasons. Technically, some of the separation methods are restricted by the critical diameter of oil droplets (particles) being treated. For instance, gravitational separation is suitable for particles larger than 60 mm; coalescence/ltration operation requires particle sizes to be 1020 mm, while oil particles of diameter smaller than 10 mm can only be adequately treated by chemical treatment, adsorption ltration [5] and biological treatment . From an economic point of view, most of the alternatives to air otation are not cost eective when they are applied to treat large volumes of wastewater. For these reasons the use of otation, especially dispersed air otation for removing emulsied oil from water, has attracted increasing attention because of its greater separation eciency, lower capital investment and operational costs [613]. Entrainment and backmixing are major problems which place limits on the eectiveness of the conventional otation machines. The slow recovery in conventional otation columns is due to the poor contact conditions between gas bubbles and particles [14]. Therefore, the successful application of a otation technique in large-scale wastewater treatment lies in the development of a proper column design. To meet this challenge, a novel multi-stage loop-ow otation column (MSTLFLO) was developed [15]. The novel design

concept is the use a series of concentric draft-tubes to establish a multi-stage operation within a single column. In each stage, the gas holdup dierence between the inside of the draft-tube (riser) and the outside (downcomer) induces a uid circulation, which greatly enhances the collision and attachment probability between bubbles and oil particles. The draft-tubes (stages) are separated by coneshaped baes which prevent inter-stage mixing and only allow a net laminar ow through a small gap (~1 mm) between the rim of the cone and the column inner wall. The ow pattern in the MSTLFLO column and an isometric view of the cone baes are shown in Figs. 1 and 2, respectively.

Fig. 1. Flow patterns in MSTLFLO column.

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

195

Fig. 2. Cone bae.

An initial study of this column was reported earlier [16 ]. This paper presents an update of the column modications and experimental results. In addition, based on results from hydrodynamics studies [17], a relationship between separation performance and hydrodynamic conditions is established. The process model developed in this work can be used as a basis for scaling-up the process.

2. Experimental 2.1. Equipment and operating conditions The MSTLFLO column process owsheet is illustrated in Fig. 3. Compressed house air is introduced into the column, passing through an air lter, a regulator, two ow meters and a porous metal sparger with mean pore size of 10 mm. The Sauter mean size of the gas bubbles ranges from 1.0 to 1.3 mm. A simulated oily wastewater prepared by adding a pre-determined amount of model oil in a 50 gallon feed tank and emulsied by pumping and recycling through a static mixer for 90 min. The oilwater emulsion prepared in this manner is stable and no coalescence of oil droplets is observed within 3 h. The emulsion is then pumped into the column from the top, regulated by a valve and a liquid ow meter, and passes through a liquid distributor. Liquid level in the column is controlled by a continuously adjustable swing hydraulic arm. The treated water leaves the bottom of the column via the clean water discharge line, while the oil-laden foam overows into the foam discharge tank. The key operating

Fig. 3. MSTLFLO process owsheet.

conditions employed in this study are listed in Table 1. Draft tubes are mounted inside the column by three ne thread screws at both top and bottom. The space between the bottom rim of the cone bae and the top of the draft tube is 0.025 m and
Table 1 Operating conditions Parameters Superficial gas velocity (102m s1) Liquid feed rate ( l min1) Frother dosage (ppm) Initial oil concentration (ppm) Conditions 04 05 020 500

196

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

the distance from the top rim of the bae to the bottom of the upper draft tube is 0.041 m. The geometric parameters of the draft-tubes used in this study are presented in Table 2. A previous study [18] has shown that ow pattern becomes undesirable when draft-tubes diameters are greater than 0.076 m or smaller than 0.051 m. When draft tubes of a larger diameter are used, the bubbles in the downcomer are large with a broad size distribution and the gas holdup is signicantly reduced. When the draft-tube diameter is too small, some of rising bubbles from the lower stage tend to enter the annular region of the upper stage (a so-called bypass phenomenon occurs). Based on these considerations, the attention was focused on a restricted range of the drafttube diameters, with the ratio of cross-sectional area of inner tube to annular region varying from 0.4 to unity (i.e. AR=0.41.0). 2.2. Materials and methods Light mineral oil (LMO) supplied by Fisher Scientic Company was used as the model oil with a density of 874 kg m3 at 25C. 2-Ethyl-1-haxanol (2-EH ), 99+%, supplied by Aldrich chemical company was selected as the frother for this study. Oil is generally a mixture of many dierent hydrocarbon compounds. A non-dispersive infrared (NDIR) technique-based oil content analyzer, Horiba OCMA-220, was chosen for this study. The minimum detectable oil concentration of the analyzer is 0.1 ppm.

3. Results and discussion 3.1. Process kinetics The macro kinetics of the oilwater separation in MSTLFLO was investigated using a simulated oilwater emulsion. It was reported that under mechanical agitation conditions, 300500 ppm, depending on types of oil, is the upper limit for preparing a stable emulsion without spontaneous phase separation [9]. Generally, the most dicult oilwater separation concerns an oil concentration starting from 500 ppm to a specic direct discharge level, typically 550 ppm [2] . Therefore, the initial oil concentration of this work was set at 500 ppm, comparable to literature data [9,13]. There is a wide divergence of opinion as to the kinetic order of an oil otation process. Angelidou [7] reported rst order kinetics for oil concentration between 50100 ppm. Takahashi et al. [5] pointed out that the actual otation in his experiments indicates a deviation from the rst order process and suggested that a distribution of the rate constant k exists: C C 2 =

2 0

j(k) exp(kt) dk

(1)

Table 2 Draft-tube geometry Parameters Nominal draft-tube sizes 0.051 m Column inside diameter, D (m) c Draft-tube inside diameter, D in (102 m) Draft-tube outside diameter, D out ( 102 m) Riser area, A (104 m2) r Downcomer area, A (104 m2) d Area ratio, AR=A /A r d 0.102 5.24 6.03 21.56 52.49 0.41 0.076 m 0.102 6.99 7.62 38.32 35.47 1.08

where j(k) is the distribution function of the rate constant. In a similar research, Sato [9] showed that for gas velocity in the range of 0.5515 cm s1, the kinetic order ranges from 2 3.5, while Pal and Masliyah [19]indicated that at high oil concentrations (>0.25 wt%), the overall kinetic order was 0.6. These inconsistent results imply that the oil particle otation diers from solid particle otation in that oil particleoil particle attachment may play as an important role as bubbleparticle attachment at high concentrations. In typical rst order particle otation, the collection rate (dC/dt) is considered to be proportional to particle concentration. Since the probability of collision among a group of particles is exponentially related to the number of particles, the particle particle interaction is not important in rst order kinetics. However, the interaction among oil particles seems to play a signicant role in their

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

197

collection. It can be easily imagined that this interaction increases with the increase of gas velocity and oil concentration. Therefore, if an oil otation process is expressed in the form in Eq. (2), the kinetic order n will decrease with the oil concentration, as can be summarized from literature data. dC dt =kCn (2)

ments were carried out and each water sample was analyzed two or three times. Average values were reported. Experimental results are plotted in Figs. 4 and 5 for two dierent draft-tube diameters. In Fig. 4, an oil concentration change curve in a conventional column is also shown as a comparison. It is seen that the oil removal in the MSTLFLO column

Lai [20] advocates that natural phenomena follow principle of continuity and do not obey a discrete classication of kinetic orders. He claims that the right-hand side of Eq. (2) is not only a function of concentration, but also a function of time. Following this line of reasoning, a simple kinetic expression was suggested for the oil otation processes: dC dt =K CC 2 t (3)

where C is the asymptote of oil concentration, 2 which depends on hydrodynamic conditions. Integration of Eq. (3) yields: =K log t+log a (4) CC 2 where a is an integration constant. To verify the kinetic expression of Eq. (3), a series of oil removal tests were performed. Oil removal eciency, g, is used as the performance criterion and dened as log C C f 100% (5) g= 0 C 0 where C and C are the initial and nal concen0 f trations, respectively. Previous studies have shown that the hydrodynamic behaviors are very similar in dierent stages. Therefore, the kinetic study was carried out in the bottom stage under batch operation conditions. The column was lled with oily wastewater at the beginning of each test, and water samples were collected at predetermined time intervals (1, 2, 4, 8, 16 and 32 min). For each condition, two experi1
Fig. 4. Oil concentration change in operation (AR=l.08).

Fig. 5. Oil concentration change in operation (AR=0.41).

198

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

is not only more ecient but also faster than that in a conventional column. To verify the postulated kinetics in Eq. (3), 1/(CC ) is plotted versus otation time t in a 2 loglog plot for dierent operating conditions. The results in Figs. 6 and 7 show that for both drafttube diameters, experimental data under each operating conditions appear as straight lines, indicating

the validity of Eq. (4). The slopes of these lines correspond to the kinetic constant K, as dened in Eq. (3), while C, is an adjustable parameter which can be used to maximize the linearity for each case. At the maximal correlation, the slope of each line was noted as K. The data are summarized in Table 3. As dened in Eq. (3), a larger K corresponds to a faster cleaning process. From Table 3, it is seen that for the AR=0.41 drafttube, the increase of supercial gas velocity and frother dosage improves the separation eciency with an increase in K. However, for the AR=1.08 case, the K reaches a maximum value at J =3 cm s1 and C =15 ppm. Further increase g F of gas supercial velocity and frother dosage result in a decrease in K. This is because for the AR= 1.08 draft-tube, at every high gas velocity and frother concentration, the increase of gas holdup cannot trade o the decrease in circulation velocity and increase of bubble size. This shows the separation performance is strongly aected by the hydrodynamic behavior of the conditions in question. Also, it is noted that the smaller (AR=0.41) drafttube has a slightly larger K, or better separation
Table 3 Summary of kinetic constants AR 1.08 1.08 1.08 1.08 1.08 1.08 1.08 1.08 1.08 1.08 1.08 1.08 0.41 0.41 0.41 0.41 0.41 0.41 0.41 0.41 0.41 SGV J (cm s1) g 1.0 1.0 2.0 2.0 3.0 3.0 3.0 3.0 4.0 4.0 4.0 4.0 2.0 2.0 3.0 3.0 3.0 4.0 4.0 4.0 4.0 Frother C (ppm) F 15 20 15 20 5 10 15 20 5 10 15 20 15 20 10 15 20 5 10 15 20 K 0.8138 1.3941 1.8235 1.8757 1.2985 1.8809 2.2131 2.2267 1.4271 1.9964 2.2009 2.1702 1.8249 1.9676 1.9966 2.2001 2.2524 1.5183 2.1057 2.3804 2.4482

Fig. 6. Regression of kinetic constant K (AR=1.08).

Fig. 7. Regression of kinetic constant K (AR=0.41).

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

199

performance, which is not entirely expected. It indicated that the liquid circulation is a very inuential parameter for oil removal. A more detailed discussion of the kinetic model and kinetic constants has been reported elsewhere [21]. 3.2. Correlation of oil removal with hydrodynamic parameters As shown above, the oil removal eciency strongly depends on hydrodynamic conditions. Therefore, an attempt was made to quantitatively correlate the kinetic constants K with hydrodynamic parameters. Based on experimental data, it was found that the volumetric ow rate is the best parameter to characterize the liquid circulation feature. Together with gas holdup and bubble Sauter diameter, a correlation was developed in an explicit form: aebQc L (6) dd 32 where a, b, c and d are empirical constants. Within the experimental range, the empirical constants were determined by the least square method [22] and the results are listed in Table 4. Using hydrodynamic data and empirical constants in Table 4, the kinetic constants were calculated. A comparison of calculated and experimental K is shown in Fig. 8. The standard errors are within the 5% zone. It is noteworthy that the separation performance of two dierent drafttube diameters can be described using the same combination of hydrodynamic parameters. The empirical constants in Table 4 show that an increase of gas holdup and liquid circulation, and a decrease of the Sauter mean diameter of bubble will increase K and enhance oil separation. Among the three parameters, bubble size is the most inuential (d>b,c). This again proves that generation of small bubbles has crucial eect on K=
Table 4 Regressed constants a 1.0034 b 0.6022 c 0.5189 d 1.1504

Fig. 8. Comparison of kinetic constants between calculated values and experimental data.

improving otation recovery of ne particles (oil droplets in this study). The eect of gas holdup and volumetric circulation rate are almost equally important in MSTLFLO operation. If an increase of gas holdup is accompanied by a decrease in circulation velocity, the separation eciency remains unchanged, as shown in the AR=1.08 draft-tube case beyond J = 3 cm s1 and g C =15 ppm. F 3.3. Modeling of the MSTLFLO process 3.3.1. Physical model for multi-stage operation The MSTLFLO process, as a multi-stage operation, can be simulated using the classic tank-inseries model [23] if each stage can be considered as a well mixed CSTR. In each stage, hydrodynamic behaviors and kinetic features have been well investigated in the previous sections. One characteristic of the MSTLFLO process is that the oil is transported from the lower stages to upper ones, which is in the opposite direction of the net liquid ow. A sketch of the physical model of the MSTLFLO process is shown in Fig. 9. In Fig. 9, oily water with concentration C is 0 fed into the column from the top. Oil-laden foam overows from the top of column and clean water

200

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

where t and C are the mean residence time and i i oil concentration of stage i(i=1,2,3); RF is the oil removed from each stage, which is a function of the concentration of that stage (C ). i If the oil clean-up kinetics is of rst order, Eqs. (7a)(7c) will become a set of linear dierential equations which can be solved analytically. However, oil separation test results have shown that the kinetic expression is in the form of Eq. (3). The kinetic constant K can be calculated using Eq. (6) with the empirical constants given in Table 4. Hence, Eqs. (7a)(7c) become: F C C 2 1= (C C )K 1 0 1 1 dt V (1e ) ts 1 1 1 C C 2 (top stage) +K 2 2 ts 2 F C C dC 2 1= (C C )K 2 2 2 dt V (1e ) 1 ts 2 2 2 C C 2 (middle stage) +K 3 3 ts 3 dC F C C 3= 2 (C C )K 3 2 3 3 dt V (1e ) ts 3 3 3 (bottom stage) dC

Fig. 9. A sketch of the physical model for MSTLFLO process.

(8a)

is discharged from the bottom. Based on mass conservation, the accumulation of oil concentration in each stage is equal to the amount of the oil input subtracting that of the output. For each stage, the input streams are: (1) net uid ow bringing in the oil from the upper stage; and (2) rising bubbles transporting the oil from the lower stages. Similarly, the output streams are: (1) net uid ow carrying the oil to the lower stages; and (2) rising bubbles transporting oil of the current stage to the upper ones. Such a model is based on the assumption that the oil concentration in each stage is uniform and back mixing is negligible. 3.3.2. Mathematical model and calculation The mathematical expressions for the above discussion can be written as follows: dC C C 1= 0 1 RF(C )+RF(C ) (top stage) 1 2 dt t 1 (7a) dC C C 2= 1 2 RF(C )+RF(C ) (middle stage) 2 3 dt t 2 (7b) dC C C 3= 2 3 RF(C ) (bottom stage) 3 dt t 3 (7c)

(8b)

(8c)

where V , K and e are volume, kinetic constant i i i and average gas holdup of stage i(i=1,2,3); C , 2 is the asymptote of oil concentration and ts is the corresponding (shifted) time for a oil concentration. Eqs. (8a)(8c) are a set of non-linear dierential equations which can only be solved numerically. For batch operation, the rst terms on the righthand side of Eqs. (7) and (8) are omitted. If the column is lled with wastewater of oil concentration C , the equations can be solved with the 0 following initial and boundary conditions: i.c.: t=0 C =C =C =C 1 2 3 0 dC i <e, i=1,2,3 b.c.:t=2, dt (9) (10)

If the column is operated in a continuous mode, the initial concentration can be set either as C or 0

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

201

0, from which a certain period is required before a steady state is reached. In order to avoid the computation problem caused by the singular point at t=0 in Eq. (3), the otation time in numerical propagation was replaced by a denite shifted time which was back-calculated at a given concentration using: 1 t =k s a(CC ) 2

(C #20 ppm) and the smaller draft-tube (AR= out 0.41) is 97% (C #17 ppm). out The concentration prole of all three stages are simulated and plotted in Fig. 11. As shown in the gure, for batch operation, there is basically no oil input to the bottom stage if back-mixing from middle stage is neglected. Therefore, the bottom stage is cleaned the fastest, and the top stage the slowest. Since an asymptote of oil concentration exists, the oil concentrations in three stages will converge after a period of time. Continuous operation. In continuous operation, the net liquid ow plays an important role in oil removal. It aects the degree of back-mixing as well as the mass balance in each stage. At the beginning of a continuous test, the column can be lled with oily water of initial concentration or clean tap water. Two dierent ow rates (1 l min1 and 3.8 l min1) were tested. Both cases were simulated and results of the concentration prole are plotted in Figs. 1214. The solid and broken lines represent the oil concentration proles for dierent start-up conditions. Circular symbols represent experimental data. These gures show that the oil concentration proles converge to the same line which represents the predicated oil concentration of steady state.

3.3.3. Model verication of prediction Batch operation. When the column is operated in batch mode, the feed rate (or net ow) is zero. The oil concentration change in each stage is simply determined by the dierence between oil removed in the current and lower stage. A comparison of experimental data and model calculated values are shown in Fig. 10. This gure demonstrates good agreement between the model calculation and experimental data for all dierent condition. It is also seen from the gure that the maximum oil removal eciency for AR=1.08 draft-tube is about 96%

Fig. 10. Comparison of model calculation and experimental data for batch operation.

Fig. 11. Oil concentrations in dierent stages.

202

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

Fig. 12. Model predictions of continuous operation (AR=1,08, F =1 l min1). L

Fig. 14. Model predictions of continuous operation (AR=0.41, F =3.8 l min1). L

Fig. 13. Model predictions of continuous operation (AR=1.08, F =3.8 l min1). L

at the high feed rate: (1) the complete mixing assumption as a CSTR in each stage is not guaranteed; and (2) back-mixing between stages may become signicant. An error analysis between the model predications and experimental data is shown in Table 5. It is seen that at the low feed rate (1 l min1), the circulation number is over 20, which can be considered as a perfect CSTR situation according to Blenke [24]. Under this situation, the model prediction is only 3% from the actual data. In comparison, at the high feed rate (4 l min1), the circulation numbers are below 10, leaving the well mixing assumption in the process model questionable. Consequently, the relative error of the model calculation is increased to about 20%. Nevertheless, the oil removal eciencies under the high feed rate conditions are still higher than 90%.

For the low-feed rate case (F =1 l min1), the L model-predicated values are in very good agreement with the experimental data. Moreover, the oil concentration of bottom-stage discharge approach the batch operation value, achieving a 95% removal eciency. For the high feed-rate case (F =3.8 l min1), the model predication and L experimental data show a small discrepancy. The model predicts higher oil removal eciency than actual data. The error is caused by the fact that

Table 5 Error analysis of the model calculation AR SGV (cm s1) 3.0 3.0 4.0 C F (ppm) 15 15 20 F L ( l min1) 1.0 3.8 3.8 n u 21 6 7 g (%) 95 91 92 Relative error (%) 3.1 19.5 19.6

1.08 1.08 0.41

X. Gu, S.-H. Chiang / Separation and Purication Technology 16 (1999) 193203

203

4. Conclusions In an eort to apply the otation technique to oil removal from wastewater, a novel multi-stage loop-ow otation column (MSTLFLO) was developed. The unique liquid circulation greatly enhances the contacts between oil particles and gas bubbles, resulting in signicantly better separation eciencies than conventional otation columns. For batch operation, removal eciency can be as high as 9697% within 5 min of otation. For continuous operation, the oil removal eciencies range from 9093% for feed rates of 13.8 l min1. Higher oil removal was achieved using smaller draft-tubes. The process was successfully modeled as a CSTR-in-series process. The kinetics of the separation was simulated and the kinetic constants were correlated with the hydrodynamic parameters.

References
[1] C. Webb, Separating oil from water, The Chemical Engineer 11 (April 1991) 1922. [2] G.F. Bennett, The removal of oil from wastewater by air otation, CRC Critical Reviews in Environmental Control 18 (1988) 189253. [3] B.R. Hansen, S.R. Davies, Trans. IChemE 72 (March 1994). [4] L.A. Spielman, S.L. Goren, Industrial Engineering and Chemistry, Fundamentals 11 (73) (1972). [5] T. Takahashi, T. Miyahara, Y. Nishizaki, Journal of Chemical Engineering of Japan 12 (5) (1979). [6 ] C.E. Adams Jr., D.L. Ford, W.W. Eckenfelder Jr., Development of Design and Operational Criteria for Wastewater Treatment Processes, Enviro Press, Nashville, 1988. [7] C. Angelidou, The removal of emulsied oil particles from water by otation, Industrial Engineering and Chemical Process Design Development 16 (4) (1977) 436441. [8] G.L. Collins, G.J. Jameson, Experiments on the otation of ne particles the inuence of particle size and charge, Chemical Engineering Science 31 (1976) 985991. [9] Y. Sato, Y. Murakami, T. Hirose, Removal of emulsied

oil particles by dispersed air otation, Journal of Chemical Engineering of Japan 13 (5) (1980) 385389. [10] B.W. Bradley, Flotation oers another water/oil separation alternative, Oil & Gas Journal (Dec. 1985) 4246. [11] Y.Y. Zheng, C.C. Zhao, A study of kinetics on induced air otation for oilwater separation, Separation Science and Technology 28 (5) (1993) 12231240. [12] K.B. Medraycka, The removal of emulsied oil particles: verication of the otation model based on interception, Separation Science and Technology 28 (7) (1993) 13791394. [13] J.E. Gebhardt, M.J. Mankosa, G.L. Iiubred, Removal of oil from produced water by microcel column otation, Internal Report of Control Inter., Inc., Salt Lake City, Utah, 1994. [14] G. Evans, B. Atkinson, G. Jameson, The Jameson cell, in: K.A. Matis (Ed.), Flotation Science and Engineering, Marcel Dekker, New York, 1995, pp. 331336. [15] F.X. Ding, S.H. Chiang, Removal of oil from waste water using multi-stage loop otation column, Fluid/Particle Separation Journal 7 (4) (1994) 149. [16 ] D.X. He, A multiple-loop otation column for wastewater treatment, Separation Technology 5 (1995) 133138. [17] X. Gu, S.H. Chiang, Hydrodynamic behaviors of a multistage otation column, The Canadian Journal of Chemical Engineering, submitted. [18] S.A. Guelcher, A hydrodynamic study of a novel multistage loop-ow otation column (unpublished M.Sc. thesis, Department of Chemical and Petroleum Engineering, School of Engineering, University of Pittsburgh, April 1996) pp. 8991. [19] R. Pal, J. Masliyah, Oil recovery from oil in water emulsions using a otation column, The Canadian Journal of Chemical Engineering 68 (1990) 959967. [20] R.W. Lai, The Two Associated Variables, Toshi, Pittsburgh, 1997. [21] X. Gu, A novel multi-stage otation column for oil removal from wastewater (Ph D. dissertation, Department of Chemical and Petroleum Engineering, School of Engineering, University of Pittsburgh, March 1998) pp. 96100. [22] Numerical Recipe in Fortran the Art of Scientic Computing, 2nd ed., on-line book in Adobe PDF format, httpnr.harvard.edu/nr/bookf.html [23] O. Levenspiel, Chemical Reaction Engineering An Introduction to the Design of Chemical Reactors, Wiley, New York, 1962. [24] H. Blenke, Loop reactors, in: T.K. Ghose, A. Fiechter, X. Blakebrough (Eds.), Advances in Biochemical Engineering, vol. 13, Springer, Berlin, 1979, pp. 121214.

Potrebbero piacerti anche