Impact of suspended solids on the activated sludge ...

Chemical Engineering Science 141 (2016) 154–165

Contents lists available at ScienceDirect

Chemical Engineering Science journal homepage: www.elsevier.com/locate/ces

Impact of suspended solids on the activated sludge non-newtonian behaviour and on oxygen transfer in a bubble column C. Durán a, Y. Fayolle a,n, Y. Pechaud a,b, A. Cockx c,d,e, S. Gillot f a

Irstea, UR HBAN, 1 Rue Pierre-Gilles de Gennes, F-92761 Antony, France Université Paris-Est, Laboratoire Géomatériaux et Environnement (EA 4508), UPEMLV, 5, Boulevard Descartes – Champs sur Marne, 77454 Marne-la-Vallée, France c Université de Toulouse, INSA, UPS, INP, LISBP, 135, Avenue de Rangueil, Toulouse, France d INRA, UMR792 Ingénierie des Systèmes Biologiques et des Procédés, Toulouse, France e CNRS, UMR5504, Toulouse, France f Irstea, UR MALY, 5 Rue de la Doua, F-69626 Villeurbanne Cedex, France b

H I G H L I G H T S

Activated sludge rheological behaviour is significantly determined by MLSS content. Gas hold-up and kLa are reduced in activated sludge compared to clean water. kLa is mainly controlled by the sludge apparent viscosity (µapp) and UG. A dynamic representation of µapp in the bubble column is proposed. A model correlating kLa with µapp and superficial gas velocity (UG) was developed.

art ic l e i nf o

a b s t r a c t

Article history: Received 6 February 2014 Received in revised form 4 August 2015 Accepted 4 October 2015 Available online 21 October 2015

This paper presents the experimental study and analysis performed in order to better understand the link between activated sludge properties, rheological behaviour and oxygen transfer. The experimental set-up consists of a bubble column of 0.3 m3 continuously fed with activated sludge and a capillary rheometer, installed in two different wastewater treatment plants: a conventional activated sludge plant and a membrane bioreactor. In the bubble column, equipped with a fine bubble diffuser, the overall gas hold-up (ϵG ) and volumetric oxygen transfer coefficients (kLa) were measured. A fraction of the column outflow was sent to the capillary rheometer where the activated sludge rheological behaviour was investigated. Several properties of the studied activated sludge were characterised (MLSS, MLVSS, soluble COD, surfactants, surface tension, soluble cations) and their impact on rheology and oxygen transfer was examined. The experimental data showed that the parameters K and n, from the Ostwald-de Waele rheological model, were strongly related to the suspended solids concentration (in terms of MLSS or MVLSS). An increase in kLa was observed as the superficial gas velocity (UG) was increased. Compared to clean water, the kLa coefficient was lower in activated sludge and still reduced with an increase of the MLSS concentration. This reduction could be partially attributed to a lower gas holdup (εG) associated with an increase in the sludge apparent viscosity (μapp) leading to a reduction of the specific interfacial area (a). Subsequently, an estimation of the mean shear rate exerted by the bubble swarm in the column allowed to evaluate the sludge apparent viscosity (μapp) of the mixed liquor at a given superficial gas velocity and MLSS concentration. Finally an empirical correlation linking kLa to the superficial gas velocity (UG) and the sludge apparent viscosity was obtained for both types of sludge. The good agreement between the experimental and the fitted data suggests that kLa can be estimated from the superficial gas velocity and the rheological behaviour. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Activated sludge Non-Newtonian fluid Capillary rheometer Hydrodynamics Oxygen transfer Bubble column

n

Corresponding author. Tel.: þ 33 1 40 96 60 32; fax: þ 33 1 40 96 61 99. E-mail address: [email protected] (Y. Fayolle).

http://dx.doi.org/10.1016/j.ces.2015.10.016 0009-2509/& 2015 Elsevier Ltd. All rights reserved.

C. Durán et al. / Chemical Engineering Science 141 (2016) 154–165

1. Introduction In conventional activated sludge wastewater treatment plants, air supply can represent up to 70% in terms of energy expenditure. The process optimization requires a fine understanding of how the different operational parameters impact the aeration efficiency of the system. Studies performed in clean water have shown that the aeration efficiency depends on design parameters such as the reactors geometry and the characteristics of the aeration system (type, layout and depth) as well as on the operating conditions such as superficial gas velocity or liquid circulation velocity (Gillot et al., 2005). Under process conditions, the activated sludge (AS) properties also have an impact on the oxygen transfer and the aeration efficiency is always lower compared to its value in clean water. In particular, several investigations performed on activated sludge (AS) systems highlighted that increasing the MLSS (mixed liquor suspended solids) concentration (between 2 and 30 g L 1) reduces the volumetric mass transfer coefficient (Cornel et al., 2003; Krampe and Krauth, 2003; Jin et al., 2006; Germain et al., 2007; Henkel et al., 2009). Actually, the only presence of solids (biological flocs and particulate material) represents an obstacle for the oxygen transfer at the gas–liquid interface (steric effect, Mena et al., 2005). Concerning the soluble fraction, Germain et al. (2007) have shown in a membrane bioreactor (MBR) that the soluble organic matter (soluble COD) ranging from 54 to 198 mg L 1 has a negative impact on oxygen transfer in AS and attributed this effect to the presence of surfactants. Some measurements carried out in clean water have demonstrated that soluble substances such as surfactants (even at small concentrations i.e. 1 mg L 1), salts and glucose interfere with oxygen transfer by accumulating at the gas–liquid interface and generating different overlapping effects: (i) lowering the surface tension (Wagner and Pöpel, 1996; Gillot et al., 2000), (ii) preventing bubble coalescence (Zlokarnik, 1980; Craig, 2004) (iii) or/and reducing the oxygen diffusivity into the liquid (Rosso et al., 2006; Hebrard et al., 2009; Jamnongwong et al., 2010). For Rosso et al. (2005), Gillot and Héduit (2008) and Henkel (2010), the negative impact of the soluble substances on oxygen transfer is reduced at higher sludge retention time (SRT) in relation to a more advanced removing or sorption of soluble substances such as surfactants. Moreover the coincidence of different characteristics and physicochemical properties of AS determine the fluid viscosity which is a key property governing the bioreactor hydrodynamics and consequently impacting the volumetric oxygen transfer coefficient (kLa). Especially the MLSS concentration has been identified to play a determining role in the rheological behaviour of activated sludge (Rosenberger et al., 2002; Tixier et al., 2003; Mori et al., 2006; Yang et al., 2009). For a given surface tension, viscosity can affect the bubble size at detachment (Kulkarni and Joshi, 2005), their rising velocity and the bubble coalescence phenomena (Mena et al., 2005). As activated sludge is a non-Newtonian fluid (Seyssiecq et al., 2003; Ratkovich et al., 2013) its apparent viscosity depends on the shear rate which can be exerted by the stirring system and by the airflow rate. Although several authors studied the influence of sludge properties on oxygen transfer on one hand, and on rheology on the other hand, Wagner et al. (2002) presented the impact of sludge apparent viscosity at a mean shear rate (40 s 1) on oxygen transfer in two full-scale MBR wastewater treatement plants (WWTP). However, in such processes, the mean shear rate is variable and depends on process operating conditions such as airflow rate or mixing characteristics. Moreover, no studies have until now evaluated, on the same type of activated sludge, the relationship existing between the activated sludge physicochemical properties, its dynamic rheological behaviour, the airflow rate and oxygen transfer.

155

In this context, the objective of this work was to evaluate the influence of two key parameters of aerated bioreactors (sludge properties and air superficial gas velocity) on the activated sludge rheological behaviour, hydrodynamics (mean shear rate) and on the volumetric oxygen transfer coefficient. The impact of operating conditions on mean shear rate is considered in order to introduce a dynamic representation of the apparent viscosity and its relationship with oxygen transfer. To this aim, the sludge properties and rheology, together with gas hold-ups and oxygen transfer coefficients were determined using an experimental set-up installed on two wastewater treatment plants (WWTPs): a conventional AS plant and a membrane bioreactor (MBR).

2. Materials and methods The experimental set-up was installed on the WWTPs of the following municipalities: Marolles/Saint Vrain (conventional activated sludge plant, later called CAS) and Briis-sous-Forges (membrane bioreactor, later called MBR). These facilities are designed for 22 000 PE and 17 000 PE respectively, treat mainly domestic effluents and are operated under extended aeration (F/M ratio o0.1 kg BOD5 (kg VSS) 1 d 1). The installation consists of a high cylindrical bubble column (H¼ 4.5 m; Dc ¼0.29 m; Dc/H¼6.4.10 2; Scw/VC ¼ 13.8 m2 m 3) installed near the aeration tank of the WWTPs. Bubble dynamics in such experimental apparatus, with a liquid height similar than those encountered in full scale aeration tanks, can be considered as representative of full scale gas–liquid dynamics as they induce similar contact time between bubble and sludge. As shown in Fig. 1, activated sludge was continuously drawn out either from the aeration tank, the sludge recirculation loop or the membrane reactor, using a helical rotor pump and fed into the column at the bottom. This recirculation induces a continuous feeding and renewal of the activated sludge during the measurements, hence allowing constant physicochemical characteristics of biological flocs and interstitial liquid during each operating conditions and avoiding endogenous respiration of activated sludge. The superficial liquid velocity (UL) in the column was maintained low and constant for a given aeration test and ranged between 2.7 10 3 and 4.5 10 3 m s 1. In the column, the liquid media was aerated by means of two compressors supplying air through a flexible fine perforated EPDM membrane (orifice diameter E0.7 10 3 m; Sp ¼0.024 m2; SP/Sc ¼0.36) installed at the bottom of the column. The injected airflow rate was measured using a volumetric gas metre (dresser). The superficial air velocity (UG) ranged from 1 10 3 to 5 10 3 m s 1 (comparable to the observed range in full-scale aeration tanks). At a height of 4.42 m from the diffuser, the sludge flow left the column by overflow and was then directed towards the capillary rheometer. 2.1. Activated sludge rheology To determine the rheological behaviour of the activated sludge under study, a capillary rheometer was designed and constructed (Fig. 1b) inspired by the tubular rheometer used by Ndoye et al. (2013) for the rheological study of a whey protein suspension. This type of rheometer is mechanically simple and allows the application of a wide range of shear stress (between 10 2 and 107 s 1). The instrument configuration helps to avoid low MLSS concentration samples to settle during the measurements. In such systems, a rheological measurement consists in determining the longitudinal pressure loss associated to the liquid flow rate through a capillary tube of known geometry. The designed device is composed of four tubes of 4, 7, 12 and 14 mm of diameter (D ¼2R). The column sludge outflow was partly pumped through these tubes using a helical rotor pump characterised by a pulseless

156


gas outlet from the column (to O2 analysis) sludge outflow pressure sensors

sludge in excess

manometers

sludge flow measurement scale dissolved O2 sensors

Q

sludge for rheology helical rotor pump

fine bubble diffuser

L1, D1

C

L2, D2 scale L3, D3

air supply activated sludge

air flow meter

L4, D4

Fig. 1. Experimental set-up installed on site: bubble column (1a) and capillary rheometer (1b).

and low shear flow (PCM – Moineau Technology). To determine the flow rate (Q) in the tubes, the sludge flow was weighed at the outlet of the tubes using a scale connected to a PC. Two pressure capture points, separated by a given length (L¼ 0.4, 0.8, 1.0 and 1 .0m respectively), were located on each tube in order to measure the longitudinal loss pressure between the two points (ΔP/L) and using a micro-manometer with a piezoresistive sensor. In a given capillary tube the shear rate and shear stress at the tube’s wall (γ_ w ;τw ) are calculated using Eqs. (1) and (2) respectively, according to the Poiseuille's law:

γw ¼ τw ¼

4Q

π R3 R U ΔP 2L

ð1Þ ð2Þ

For a non-Newtonian fluid circulating in a capillary tube, the velocity profile depends on the non-Newtonian character. Activated sludge is generally considered as a shear-thinning fluid (Seyssiecq et al., 2003; Ratkovich et al., 2013). To consider the impact of the sludge non-Newtonian character on the tubular rheological measure, the Rabinowitsch equation was used to correct the experimental results (Dupuis, 2008). The shear rate is then calculated by applying a correction factor as follows: 4Q 3m þ1 ð3Þ γw ¼ 3 4m πR where m is deduced from the slope of ln (τw ) versus ln(4Q/πR3). The ratio of the shear stress to the shear rate is the apparent viscosity (μapp ):

μapp ¼

τw γ̇ w

ð4Þ

By varying the sludge and flow rate (Q), the applied shear stress ranged from 50 to 400 s 1, which covers the range of shear stress in the bubble column (see Fig. 7 below). The sludge temperature was measured at the tubes outlet using a PT-100 sensor. The experimental results obtained on site at 11 72 °C (CAS) and 227 2 °C (MBR) were converted to 20 °C by using the following correlation (in analogy to the Arrhenius equation):

μappASð201CÞ B ¼ A expðT þ 273:15Þ μappASðT1CÞ

ð5Þ

where A and B are empirical coefficients. Their values (A ¼169.6 and B ¼ 1531.4 K 1) were deduced by minimizing the sum of squared residuals between the experimental rheograms obtained at three different temperatures (10, 15, and 20 °C, not presented)

and the estimated values using Eq. (5). When the temperature rises from 10 to 20 °C, the apparent viscosity at a given shear rate is reduced by an average of 24.5% which is indeed similar to the decrease of water dynamic viscosity (23.3%) within the same range of temperature (Kestin et al., 1978). Based on the error of the different instruments, the measurement precision was theoretically estimated at 77%. Rheological measurements carried out in clean water showed that the instrument mean relative error was 75% by comparison with the dynamic viscosity of the water at the measurement temperature (Kestin et al., 1978). In addition, the rheograms (τ versus γ_ ) obtained with the different tubes overlapped, showing the adequacy of the applied correction (Rabinowitsch, Eq. (3)). The experimental rheological behaviour was modelled using the Ostwald-de Waele equation: ̇ ðn1Þ τ μapp ¼ ̇ ¼ Kγ γ

ð6Þ

This simple model relates the shear stress (τ) to the shear rate (γ ̇) by integrating two rheological parameters in a power law: K, the consistency index and n, the flow index. The use of this model allowed to represent the sludge rheological behaviour in the applied range of shear stress (50 s 1 o γ_ o400 s 1) without adding other parameters unnecessarily (Ratkovich et al., 2013). K and n are calculated by fitting this model to experimental rheograms. The resulting R2 are comprised between 0.982 and 0.999 (CAS and MBR). 2.2. Overall gas hold-up measurements The column overall gas hold-up ðεG Þ was measured using two hydrostatic pressure sensors (Endress Hauser) submerged at the top and the bottom of the aerated volume and connected to a data acquisition system (Fig. 1a). When the column was filled with activated sludge (or with clean water), the pressure difference between the two sensors was measured and recorded with air supply (during the oxygen transfer tests) and without air supply. Under aerated and non-aerated conditions the pressure values were averaged during at least three minutes of steady state hydrodynamic conditions. As ρair {ρwater ρsludge , the global gas hold-up can be estimated using the following equation: ΔPwith\air εG ð%Þ ¼ 1 100 ð7Þ ΔPwithout\air The relative error of this method is estimated to 710%. The measurement of the overall gas hold-up by means of the hydrostatic pressure has been validated by comparison to results obtained by the


157

Table 1 Characteristics of CAS and MBR sludge in the rheology study. AS characteristics

CAS

MLSS (g L 1) K (10 3 Pa s)n nn MVLSS (g L 1) MVLSS/MLSS (-) SVI (mL g 1) CODsol (mg L 1) Analysed cations (mg L 1) Na þ (mg L 1) K þ (mg L 1) Mg2 þ (mg L 1) Ca2 þ (mg L 1) Surfactants (mg L 1) Anionic surf. (mg L 1) Non-ion. surf. (mg L 1) Cationic surf. (mg L 1) Surface tension(mN m 1)n ρML (kg m 3) Date (dd/mm/yy) Sludge samplennn

2.8 3.6 0.83 1.8 0.64 109 19.9 276.3 59.4 20.2 19.7 177 0.8 0.6 0.2 o 0.2 72.1 997 26/06/13 AR

MBR 4.5 8.8 0.73 3.1 0.69 101 25.4 222.4 47.2 13.8 17.3 144.1 1.2 0.9 o0.2 0.2 72.2 1000 13/02/13 AR

4.6 11.6 0.67 3.2 0.70 115 27 271.4 74.6 24.2 18 154.6 0.8 0.3 o0.2 0.5 72.5 999 12/07/13 RL

5.0 8.8 0.73 3.3 0.66 107 29.3 139.4 34.8 10.5 10.7 83.3 0.9 0.9 o 0.2 o 0.2 72.5 1003 06/02/13 AR

8.5 45.7 0.54 5.7 0.67 104 26 265.9 68 22.2 23.6 152 1.1 0.8 o0.2 0.3 72.4 1001 08/02/13 RL

8.6 66.0 0.48 5.9 0.66 117 22.4 158.6 37.4 11.9 12.3 97 1.2 0.9 o 0.2 0.3 72.2 1002 15/02/13 RL

4.0 11.6 0.68 2.6 0.65 162nn 25.8 182.5 54.1 17.4 7.6 103.4 1.3 1.0 o 0.2 0.2 71.8 999 14/08/13 AR

6.1 30.9 0.59 4.1 0.67 165nn 35.3 47.1 12.8 4.5 1.7 28.1 0.7 0.5 0.3 o 0.2 72.5 991 19/07/13 AR

6.4 33.6 0.58 4.4 0.69 133nn 32.3 215.2 60.2 22 8.7 124.3 1.8 1.3 0.4 o 0.2 72.4 999 18/07/13 AR

7.9 74.1 0.49 5.2 0.66 177nn 29.3 34.7 10.3 3.9 1.5 19 2.4 0.9 1.2 0.4 72.5 1000 16/08/13 MR

10.2 127.8 0.43 7.1 0.70 145nn nd 251 70 33.6 12.3 135.1 3.1 1.7 1.1 0.3 68.8 1000 09/07/13 MR

n

Results at 20 °C. Floating aggregates observed. nnn AR ¼aerated reactor; RL¼ recycling loop; MR ¼ membrane reactor (MR). nd: not determined. nn

level difference method performed with a float level sensor (Endress Hauser) which provides an absolute measurement uncertainty of 70.04%. The overall gas-hold up results obtained at the temperature T (εGT ) for CAS and MBR sludge are reported at 20 °C (εG20 ) using Eq. (8).

ϵG20 ¼ ϵGT θ'ð20TÞ with θ' ¼ 1:015

ð8Þ

The temperature correction coefficient (θ`) has been computed from the results of aeration test performed in the bubble column with clean water at a mean temperature of 22, 16 and 10 °C (not presented). Assuming that the effect of temperature on gas holdup in clean water is similar to that on activated sludge, this temperature correction was applied to the studied sludge. 2.3. Oxygen transfer measurements Oxygen transfer tests in clean water were performed according to the non-steady state method (NF EN 12255-15, 2004). In activated sludge, the tests were carried out using the off-gas method (ASCE, 1997). The oxygen transfer efficiency was estimated through the measurement of the oxygen content in ambient air and off-gas stream exiting the liquid surface of the column by using an electrochemical oxygen sensor (Teledyne Analytical Instruments). The dissolved oxygen concentration in the liquid media was measured using three stirred electrochemical sensors (YSI) submerged at three different heights in the aerated volume (Fig. 1a). The temperature and conductivity of the liquid media were measured using a sensor (WTW) submerged at middle height in the column. The gas-phase oxygen fractions, the dissolved oxygen concentrations, the liquid temperature and conductivity were recorded using a data acquisition system (Yokogawa). The volumetric oxygen transfer coefficients measured at the temperature T (kL aT ), are reported at 20 °C (kL a20 ) using the following temperature correction (ASCE, 2007): kL a20 ¼ kL aT U θ

ð20TÞ

with θ ¼ 1:024

ð9Þ

2.4. Physicochemical characterization of activated sludge For each measurement series, a sludge sample was taken and the following parameters were determined: MLSS, MLVSS (mixed liquor volatile suspended solids) (according to NF T90-105-2,

1997), density and sludge volume index (SVI). On the soluble fraction, the following analyses were also performed after a 0.45 mm filtration: static surface tension (ring tensiometer LAUDA), sodium, potassium, magnesium and calcium ions were measured by ion chromatography. The following chemical analyses were performed using Hach Lange cuvette tests: soluble chemical oxygen demand (COD) (LCK 314), anionic surfactants (LCK 332), nonionic surfactants (LCK 333), and cationic surfactants (LCK 332). 2.5. Statistical analysis In order to examine the link between AS physicochemical properties, AS rheological behaviour and oxygen transfer coefficient (kLa), the data results were studied by following a stepwise multiple regression analysis (Gardener, 2012). In this approach the starting point of the multiple regression is a “blank model” that includes only the intercept. Subsequently the next best independent variables (the AS characteristics) are added to the model (one at a time) by calculating the Akaike Information Criteria (AIC) which is proportional to the residual sum of squares (RSS) worked out from a line (or plane) of best fit. The multiple regression is retained as long as the overall regression and the included individual independent variables are statistically significant (p-valueo0.001). From the calculation of the Pearson's correlation coefficient (r), the analysis results allow to determine the beta coefficients (standardised regression coefficients) that indicate how strongly the standard deviation of a given independent variable impacts the standard deviation of the response variable. The statistical computing was performed using the software environment R (http://www.r-project.org/).

3. Results and discussion 3.1. Sludge properties The activated sludge characteristics are presented in Table 1. These data were measured in order to highlight the impact of physicochemical parameters on rheological properties of the sludge and on hydrodynamics and oxygen transfer characteristics. While the soluble COD concentration and sludge density remained almost constant for the different sludge concentrations, the main

158


CAS

14

MLSS (g/L) 2.8 3.0 4.5 4.6 5.0 8.5 8.6

10 8

6 4

MLSS (g/L)

12

µ app (10-3 Pa.s)

µapp (10-3 Pa.s)

12

MBR

14

4.0 6.1 6.3 7.9 10.2

10 8

6 4 2

2

0

0

0

100

200 300 400 -1 shear rate (s )

500

0

100

200 300 400 -1 shear rate (s )

500

_ for the two types of AS (CAS and MBR). MLSS range from 2.8 g L 1 to 10.2 g L 1. Results at 20 °C. Fig. 2. Apparent viscosity (μapp) versus shear rate (γ)

3.2. Activated sludge rheology The variation of the sludge apparent viscosity as a function of the applied shear rate obtained for both sludge origins (CAS (a) and MBR (b)) and for MLSS concentrations ranging from 2.8 to 10.2 g L 1 is shown in Fig. 2. The Ostwald-de Waele rheological parameters (K and n) versus sludge MLSS concentration are presented in Fig. 3 for the sludge samples from CAS and MBR. The results modelled by Rosenberger

10

1.0

0.8

1

K - CAS

0.6 0.1 0.4 0.01 0.2

0.001

n, Flow Index (-)

K, Consistency Index (Pa.s )

difference between both types of sludge is the sludge volume index (SVI), with average values of 107 and 156 mg L 1 for CAS and MBR respectively. The surfactant concentration is comprised between 0.4 and 3.1 mg L 1. The anionic surfactant is the most concentrated one (from 0.4 to 1.7 mg L 1) and the non-ionic surfactants are detected only in MBR activated sludge (with concentrations between 0 to 1.2 mg L 1). These low values could be related to the characteristics of studied wastewater treatment plants (extended aeration, low F/M ratio, SRT 415d) which enhance the biodegradation and adsorption of surfactants. Wagner and Pöpel (1996), Gillot et al. (2000) and Capela et al. (2002) measured the impact of different surfactant types and concentrations (from 1 to 7.5 mg L 1) on clean water oxygen transfer characteristic parameters. These studies highlight that for such concentrations, oxygen transfer is reduced by non-ionic surfactant addition (detected in our study in only one sample) and is slightly or not influenced by anionic surfactant addition (which is the most concentrated one in our samples). The impact of anionic surfactant addition on oxygen transfer has been observed by some authors (Painmanakul et al., 2005; Rosso et al., 2006) but for significantly higher concentration (from 50 mg L 1). The resulting static surface tension is comprised between 68.8 and 72.6 mN m 1 (at 20 °C). Excluding the value obtained for MBR highest concentrated activated sludge (MLSS¼10.2 g L 1/ surfactant concentration ¼3.1 g L 1), the mean measured surface tension is closed to tap water value (72.5 mN m 1 at 20 °C). Jimenez (2013) has measured the oxygen diffusion coefficient in the interstitial liquid of CAS sludge samples in a bubble wake in a glass column using the planar laser-induced fluorescence (PLIF, Dietrich et al. (2015)). The obtained results highlight that the oxygen diffusion coefficient in AS interstitial liquid is equivalent to the one measured in clean water (1.85–2.02 10 9 m2 s 1 in AS interstitial liquid and 1.95–2.00 10 9 m2 s 1 in clean water).

K - MBR K Model (Rosenberger et al., 2002) n - CAS n - MBR n Model (Rosenberger et al., 2002)

0.0 0

5

10

15

20

[MLSS] (g.L )

Fig. 3. Ostwald-de Waele rheological parameters (K and n) versus MLSS concentration for CAS, MBR activated sludge and results from Rosenberger et al. (2002).

et al. (2002) for nine MBR sludge samples in a higher MLSS concentration range (2.7–35 g L 1), are also presented on this figure. The rheological results presented in Fig. 2 show that the sludge apparent viscosity decreases as the applied shear stress increases, which confirms the shear-thinning behaviour observed in the literature (Rosenberger et al., 2002; Seyssiecq et al., 2003; Mori et al., 2006; Yang et al., 2009; Ratkovich et al., 2013). In Fig. 2, it can also be observed that for CAS and MBR the curves obtained at similar MLSS concentrations seem to overlap, highlighting that for AS from these two different origins and for a fixed applied shear rate, the viscosity is mainly controlled by the MLSS concentration. The sludge apparent viscosity (μapp) and the consistency index (K) increase simultaneously with the sludge MLSS concentration (Fig. 3). This is usually explained by a higher number of interactions between particles that consequently move less freely and exert more resistance to flow. On the contrary, it is observed that the flux index (n) decreases when the MLSS concentration increases, thereby accentuating the shear-thinning character of the activated sludge (i.e. accentuating the decrease in viscosity with an increase in shear rate). For both origins of sludge samples, the evolution of these rheological parameters (K and n) with the MLSS concentration follows a similar trend and is comparable to the results obtained by Rosenberger et al. (2002), despite the different rheometer geometries (tubular rheometer in our case and concentric cylinders for Rosenberger et al., (2002)) and experimental protocols. Moreover, despite a low data scattering observed between the rheological parameters and the MLSS concentration, confirming that MLSS concentration is the main parameter


159

Table 2 Empirical models and constants for Ostwald-de Waele rheological parameters K and n for both origins of sludge (CAS and MBR). WWTP

CAS MBR

Empirical constants

Equations

A

B

C

D

0.46

1.01

0.07

0.88

0.61

0.13

1.17

K ¼ 103 eAMLSS Eq:\ð10Þ B

D

0.63

n ¼ 1þ CMLSS Eq:\ð11Þ

Table 3 Results of the multiple regression analysis on the rheology experimental data. The response variables are the Ostwald-de Waele rheological parameters K and n.

K, consistency index n, flow index

Beta MLSS

R2

p-Value

0.93 0.96

0.87 0.92

o 0.001 o 0.001

controlling the rheology of activated sludge, slight differences could be observed between the tendencies obtained for the two sludge origins (CAS and MBR). This could be due to differences in physicochemical properties of biological flocs such as size, structure and strength. To represent the evolution of the Ostwald-de Waele parameters as a function of the MLSS concentration measured in this work (in the range 2.8–10.2 g L 1), the following two equations were obtained (Table 2): The models for K and n are similar to the ones proposed by Rosenberger et al. (2002) and accurately reproduce the tendencies of Fig. 3. The values of the coefficients of Eqs. (10) and (11) are not the same for both origins of sludge samples which confirms the slight differences in the tendencies observed on Fig. 3. The use of these two equations together with Eq. (6) allows calculating the apparent viscosity of activated sludge within a range of MLSS concentration between 2.8 and 10.2 mg L 1 for both AS origins. When the MLSS concentration is zero, the calculated value corresponds to the dynamic viscosity of water at 20 °C (10 3 Pa s). Results from the multiple regression analysis (Table 3) indicated that among the characterised sludge properties, for the two types of sludge (CAS and MBR), only the MVLSS concentration which is closely related to the MLSS concentration (r ¼0.99) showed to be significantly correlated to the Ostwald-de Waele rheological parameters (K and n) and therefore to the apparent viscosity, as illustrated in Fig. 3, confirming previous observations (Rosenberger et al., 2002; Tixier et al., 2003; Mori et al., 2006; Yang et al., 2009). 3.3. Overall gas hold-up The overall gas hold-up versus superficial gas velocity for both WWTPs is shown in Fig. 4 for clean water and a MLSS concentration ranging from 3.0 to 10.0 g/L. Oxygen transfer tests were not performed systematically at the same time as rheology measurements. For this reason the MLSS concentrations presented in the next figures do not always coincide with the values presented in Table 1. The experimental results obtained at a temperature T are estimated at 20 °C by using Eq. (8). The overall gas hold-up increases with the superficial gas velocity in both clean water and activated sludge. For a given superficial gas velocity, the gas hold-up is most generally lower in activated sludge than in clean water regardless of the sludge MLSS concentration (the order of magnitude of this decrease is 20% in

Fig. 4. Overall gas hold-up (εG20 ) versus superficial gas velocity (UG) for clean water, CAS and MBR sludge at 20 °C.

Table 4 Estimated mean bubble rise velocity (UB) in the bubble column for clean water and the two types of activated sludge at 20 °C (UB ¼ UεGG ).

Clean water CAS

MBR

MLSS (g L 1)

UB(m s 1)

– 3.0 4.7 8.6 6.1 10.0

0.29 0.38 0.34 0.38 0.30 0.40

average). The mean bubble rise velocity in clean water and activated sludge can be estimated from the previously presented gas hold-up results, by calculating the inverse of the slope of the curve εG versus UG showed in Fig. 4. In general, the estimated values, shown in Table 4, suggest that the bubbles rise faster in activated sludge than in clean water which explains the overall gas hold-up reduction. The reduction of the overall gas fraction in non-Newtonian fluids compared to clean water has also been observed by some authors but in different operating conditions. Mineta et al. (2011) observed the reduction of gas holdup in batch oxygenation tests performed in a bubble column filled with AS at MLSS concentrations ranging from 2 g L 1 to 8 g L 1 and gas hold-up between 0.5% and 2%. Additionally this author observed a decline of the gas fraction with the increase of the MLSS concentration more pronounced than in our study. Fransolet et al. (2005) also observed a decrease in gas holdup with an increase of xanthan concentration in non-Newtonian shear-thinning xanthan aqueous solutions (for concentrations from 1 to 3 g L 1). Moreover, at higher xanthan concentrations (4–5 g L 1), the impact on gas hold-up was less pronounced. Three overlapping viscosity effects help to explain the observed evolutions: (i) as the solid fraction rises, the liquid becomes more viscous and consequently the bubble coalescence is favoured

160


40

40 CAS

35

35

y = 11.11x 0.76

25

Clean water

y = 7.64x 0.97

20

y=

15

MLSS 3.0 g/L

5.77x 0.98

MLSS 4.7 g/L

y = 5.77x0.70

10

MBR

30

MLSS 8.6 g/L

5

k La 20 (h -1)

k La 20 (h-1)

30

25

20 15 10

y = 11.11x 0.76

Clean water

y = 3.45x 1.12

MLSS 6.1 g/L

y = 4.04x 0.58

MLSS 10.0 g/L

5

0

0 0

1

2

3

4

5

-3

UG (10 m/s)

0

1

2

3

4

5

-3

UG (10 m/s)

Fig. 5. Oxygen transfer coefficient (kLa20) versus superficial gas velocity (UG) for CAS and MBR sludge.

which leads to an increase of bubble size (Mena et al., 2005) (ii) at the bubble formation stage in a non-Newtonian fluid, the bubble growth time is extended with the increasing solids concentration due to higher viscoelastic stresses exerted on the bubble and consequently bubble size is increased (Kulkarni and Joshi, 2005) (iii) the velocity of the rising bubbles is reduced in viscous liquids due to a higher bubbles drag coefficient (Mena et al., 2005). As depicted on Fig. 4, independently of the MLSS concentration of the AS introduced continuously in the column, the overall gas hold-up as a function of superficial gas velocity follows a similar trend, highlighting that a further increase of the sludge MLSS concentration (from 3.0 to 10.0 g L 1) does not necessarily lead to further gas hold-up reduction. The slight variations of εG with the increment in the MLSS concentration and the associated viscosity, could be explained by the counterbalancing effects of viscosity on bubble drag coefficient and on the increase of mean bubble size (confirmed by visual observations) due to increase in bubble growth time and coalescence. In other words, as the viscosity increases, larger bubbles can be generated and may not necessarily raise faster because an increment in the bubble drag coefficient, associated to the increase in liquid viscosity, would actually extend the bubbles residence time in the liquid. 3.4. Volumetric oxygen transfer The evolution of oxygen transfer coefficient with the superficial gas velocity for both WWTPs is shown in Fig. 5 for clean water and a MLSS concentration ranging from 3.0 to 10.0 g L 1. Similarly to the gas hold-up results, the oxygen transfer results presented in Fig. 5 show that kLa increases with the superficial gas velocity which can be mainly explained by an increment in the gashold up and therefore a larger specific interfacial area ða ¼ 6εG =db Þ. The results also confirmed other works in the literature that describe a reduction of the kLa coefficient with the sludge MLSS concentration (Krampe and Krauth, 2003; Jin et al., 2006; Germain et al., 2007; Henkel et al., 2009; Mineta et al., 2011). The MLSS concentration effect on the kLa coefficient seems to be amplified in comparison to the effect on gas hold-up (see Fig. 4). As diffusivity and gas hold-up are not affected by sludge physicochemical characteristics (as previously mentioned), the MLSS effect could be attributed to bubble size variations due to increase in bubble growth time and coalescence, inducing a decrease in interfacial area (a) or by a decrease in kL due to transport limitation, associated to the increase in liquid viscosity. It is also observed that the slopes characterizing the evolution of the kLa coefficient with the superficial gas velocity UG, decreases with an increase in the sludge MLSS concentration. This suggests that the impact of superficial gas velocity on bubble size is more

Table 5 Results of the multiple regression analysis on the aeration tests data. The response variable is the coefficient kLa20. Beta-MLSS Beta-UG R2 p-Value

0.77 0.56 0.87 o 0.001

pronounced by increasing the MLSS concentration (and the associated viscosity). Besides, the results of the multiple regression analysis showed that the kLa coefficient is mostly impacted by the MLSS concentration and obviously by the superficial gas velocity (UG). The Table 5 summarizes the best fitting parameters resulting from the multiple regression analysis of the aeration tests results. On the other hand, the multiple regression analysis did not show a statistically significant impact (p-value 40.001) of the soluble organic matter (CODsol), surfactants (cationic, anionic or non-ionic) or analysed cations (Na þ , K þ , Ca2 þ , Mg2 þ ) on the oxygen transfer coefficient (kLa). However, it is known that the dissolved substances can also participate to the depletion of the oxygen transfer coefficient (kLa) by accumulating at the gas–liquid interface, (i) reducing the bubbles terminal rise velocity (Alves et al., 2005; Sardeing et al., 2006) and/or (ii) hindering the oxygen diffusivity into the liquid (Rosso et al., 2006; Hebrard et al., 2009; Jamnongwong et al., 2010) and consequently reducing the liquidside oxygen transfer coefficient (kL). Nevertheless, the effect of these substances has mainly been observed in clean water with added substances. In activated sludge very few results are reported and are not concluding. While Germain et al. (2007) observe a negative impact of the soluble COD on oxygen transfer in nine MBR AS samples and attributes this to surfactants, Henkel (2010) reports a non-significant impact of anionic surfactants on oxygen transfer in four MBR treating greywater.

4. Modelling the oxygen transfer coefficient in AS considering the non-Newtonian behaviour 4.1. Model development In order to improve the estimation of the oxygen transfer coefficient, empirical correlations based on key parameters and properties have been proposed. For the activated sludge process, these correlations consider for instance the coupled effect of solid retention time (SRT) and surface flow rate (Rosso et al., 2005), SRT and the bubbles contact time in the liquid phase (Gillot and


Héduit, 2008), SRT and the MLVSS concentration (Henkel, 2010) or viscosity for a fixed applied shear rate (Wagner et al., 2002; Pittoors et al., 2014). Although these correlations clarify the influence of these parameters on oxygen transfer, they do not take into account the shear-thinning behaviour of mixed liquor. From the rheology and oxygen transfer experimental results obtained in this study, an empirical model that relates hydrodynamics, rheology (by considering apparent viscosity and its dependence to shear rate) and oxygen transfer in the bubble column was thus developed based on the dimensional analysis. The chosen parameters are the transfer number (NT), previously defined as a scale-up dimensionless number characteristic of the oxygen transfer (Capela et al., 2001; Gillot et al., 2005) and the column Reynolds number (ReC, ratio of inertial forces to viscous forces). The transfer number compares mass transfer through the gas/liquid interface to inertial forces, but also considers viscous forces, which are the main variability factors of our system. These dimensionless numbers are defined in Eqs. (12) and (13). ! 2 kL a20 μapp;ML 1 3 NT ¼ ð12Þ UG ρ2ML :g ReC ¼

U G Dc ρML

ð13Þ

μapp; ML

The transfer number is expressed as a function of the column Reynolds number as follow: NT ¼ e1 ReeC2

ð14Þ

where e1 and e2 are numerical constants. The oxygen transfer coefficient can be written using Eqs. (12)– (14), which correspond to the formalism proposed by GarciaOchoa and Gomez (2009). e 2

þ e2 Þ 2 3 kL a20 ¼ e'1 Uð1 μapp;ML G

ð15Þ

However, in the present work, due to the model construction deduced from dimensional analysis, the constant exponents

161

associated to variables UG and μapp are both interrelated through the e2 empirical constant value. In this expression, the constant e’1 regroups all aspects related to the reactor configuration (geometry, diffuser type and distribution, etc) as well as the characteristics of the sludge properties impacting kLa20 that were not measured in these experiments. This correlation expresses that for a given system, kLa20 is only controlled by the gas superficial velocity and the rheological properties of the AS. The derivation of the e1 and e2 empirical coefficients is summarized in Fig. 6 and described in the next paragraphs. To estimate the mixed liquor apparent viscosity (μapp,ML) in the bubble column it is necessary to know the mean shear rate ðγ_ Þ prevailing in the aerated column. This can be determined according to the theoretical equation proposed by Sanchez Pérez et al. (2006) (Eq. (16)), which is based on the specific energy dissipation rate under the assumption that the loss of kinetic energy due to the gas leaving the column and the gas friction at the column walls can be neglected. The mean shear rate is calculated as follows:

γ̇ ¼

1 g ρU G K

1=ðn þ 1Þ

ð16Þ

The rheological measurements performed in this study showed the close correlation of K and n with the MLSS concentration for both types of AS (Eqs. (10) and (11) in Table 2). By using these two equations, the mean shear rate in the bubble column is estimated in the range of the applied superficial gas velocity and MLSS concentration. Fig. 7a shows the evolution of the estimated shear stress with the superficial velocity for the different MLSS concentrations. The highest mean shear stress in the bubble column is applied at the lowest values of MLSS concentration and viscosity (clean water) and at the highest superficial gas velocity. The estimated value ranges between 23 and 221 s 1.

Fig. 6. Derivation of the empirical coefficients e1 and e2 in Eq. (12).

162


45

250

40

150

Clean water

35

MLSS 3.0 g/L CAS

30

MLSS 4.7 g/L CAS

MLSS 6.1 g/L MBR

100

MLSS 8.6 g/L CAS 50

MLSS 10.0 g/L MBR

kLa20(h-1)

mean shear rate (s-1)

200

+UG

25 20

15 10

5 0

0 0

1

2

3

4

5

6

0

UG (10-3 m/s)

5

10 app (10

-3

15 Pa.s)

20

25

Fig. 7. (a) Calculated mean shear rate as a function of the superficial gas velocity and (b) kLa coefficient versus calculated apparent viscosity for the different MLSS concentrations and the two types of activated sludge.

Clean Water 3.E-04

CAS

MBR

40

+10%

+10%

35

NT -Model (-)

-10%

2.E-04 2.E-04 1.E-04

k La 20 - Model (h-1)

3.E-04

-10%

30 25

20 15 10

5.E-05 0.E+00 0.E+00

5 0 1.E-04 2.E-04 N T - Experimental data (-)

3.E-04

0

10 20 30 kLa 20 - Experimental data (h-1)

40

Fig. 8. Agreement between experimental and modelled transfer number, NT (left) and between experimental and modelled mass transfer coefficient (right).

Knowing the mean shear stress exerted by the air bubbles allows the calculation of the mixed liquor apparent viscosity in the bubble column associated to the measured superficial gas velocity (UG) and MLSS concentrations by using the Ostwald-de Waele equation (Eq. (6)) as well as the rheology experimental results (Eqs. (10) and (11)). Fig. 7b shows the evolution of the kLa coefficient with the apparent viscosity for the different MLSS concentrations and the two types of sludge in the bubble column. It can be observed that for the highest MLSS concentration and the lower superficial gas velocity, the sludge apparent viscosity reach values up to 20 times higher than clean water dynamic viscosity. Subsequently the e1 and e2 empirical constants are computed by minimizing the weighted mean absolute error between the modelled transfer number and the one obtained using experimental data at a given superficial gas velocity (UG) and MLSS concentration. The resulting models are: NT ¼ 5:25x104 Re0:23 C

ð17Þ

0:44 kL a20 ¼ 3:11x102 U0:77 G μapp;ML

ð18Þ

A good agreement is observed between experimental and modelled data for transfer number and oxygen transfer coefficient (models allow to predict the values with a precision of 7 10% in most cases) as presented in Fig. 8. The transfer number (NT) decreases with an increase in the column Reynolds number (Eq. (17) and Fig. 9). The experimental transfer number values are in the range of 9.9–26.3 10 5 and seem to increase with an increase in the MLSS concentration. The minimal values are obtained for clean water results (9.9– 12.3 10 5) and are in the same order of magnitude than previous results for full-scale aeration tanks with total floor coverage (Capela et al., 2001; Gillot et al., 2005). By integrating inertial forces and physicochemical parameters of the liquid phase including viscous forces, the transfer number (NT) is a reliable scale-up factor and a tool for oxygen transfer modelling in aerated tanks. However, for various MLSS concentration, the physical significance of this dimensionless number must not be considered as equivalent to transfer efficiency as for clean water conditions. Nevertheless, the use of mass transfer number (NT) in the next paragraph will contribute to further analysis and interpretation of the viscous forces effect on oxygen transfer efficiency.


163

model the evolution of gas–liquid flow characteristics with MLSS concentration increase. This model considers indirectly the influence of MLSS concentration on key parameters affecting mass transfer rate such as bubble size or mean contact time. No extra parameter is needed to represent the evolution of mass transfer coefficient despite two origins of activated sludge (CAS and MBR, with a low F/M ratio and SRT 415 d) and thus differences in physicochemical characteristics of the sludge. The transfer number (NT) can also be expressed as a combination of mass transfer Stanton number (StM, ratio of total mass transfer coefficient to inertia forces which could be considered as equivalent to oxygen transfer efficiency) and Galileo number (Ga, ratio of gravity forces to viscous forces). The physical significance of the mass transfer Stanton number allows to dissociate mass transfer efficiency from viscous effect and physicochemical characteristics of the liquid phase and is preferred to transfer number for interpretation of AS rheology effect on gas–liquid mass transfer as previously discussed. The Froude number (Fr, ratio of inertial forces to gravitational forces) is also introduced for this interpretation.

4.2. Interpretation of the impact of AS rheology on gas–liquid mass transfer The resulting modelled kLa evolutions (Eq. (18)) with the superficial gas velocity for the different MLSS concentrations were drawn and compared to the experimental data. Fig. 10 shows the agreement between the experimental and the fitted data for CAS and MBR respectively. The good agreement between experimental and modelled results, highlights that the impact of MLSS concentration on oxygen transfer is mainly controlled by its effect on mixed liquor apparent viscosity, considering the sludge shear-thinning behaviour and its evolution with aeration intensity (in relation to the exerted mean shear rate). It allows to integrate in a mass transfer

NT ¼ St M Ga1=3 St M ¼

Ga ¼

Fr ¼

Fig. 9. Transfer number (NT) versus column Reynolds number (Rec) for clean water, CAS and MBR.

45

45

40

40

35

35

30

ð19Þ

kL a20 Dc UG

ð20Þ

g D3C ρ2ML

ð21Þ

μ2app;ML

U 2G g DC

ð22Þ

30 Clean Water MLSS 3.0 g/L MLSS 4.7 g/L MLSS 8.6 g/L

25 20 15

25

Clean Water

20

MLSS 6.1 g/L

15

MLSS 10.0 g/L

10

10

5

5

0

0 0.0

1.0

2.0

3.0

4.0

5.0

0.0

1.0

2.0

3.0

4.0

5.0

6.0

Fig. 10. Measured and calculated oxygen transfer coefficient (kLa20) as a function of the superficial gas velocity (UG) for CAS (left) and MBR (right) at different MLSS concentrations. Calculated data are in dotted lines.

0.8

0.6

Clean water

StM(-)

CAS - 3.0 g/L CAS - 4.7 g/L

0.4

CAS - 8.6 g/L MBR - 6.1 g/L MBR - 10.0 g/L

0.2

0 1.E-06

1.E-05 Fr (-)

Fig. 11. Mass transfer Stanton number (StM) versus Galileo number, Ga (left) and Froude number, Fr (right) for clean water, CAS and MBR.

164


To interpret the impact of AS sludge rheology on gas–liquid mass transfer, the mass transfer Stanton number is presented as a function of the Galileo number on Fig. 11.The mass transfer Stanton number is comprised between 0.17 and 0.76 with maximum values for clean water operating conditions (from 0.61 to 0.76) and minimum values for maximal MLSS concentration (from 0.17 to 0.23 for a MLSS concentration of 10 g L 1). It appears that the mass transfer Stanton number is strongly correlated to the Galileo number highlighting that the oxygen transfer efficiency is mainly driven by the AS viscous effects on gas–liquid dynamics in the bubble column, by controlling bubble size (at bubble formation but also its evolution during bubble rise, due to coalescence and break-up) and rising velocity (in relation to drag forces). It also highlights that the apparent viscosity could be considered as the key parameter for interpretation of MLSS impact on oxygen transfer efficiency, by integrating the sludge shear-thinning behaviour. On Fig. 11, the mass transfer Stanton number is also presented as a function of the Froude number. This representation allows the comparison between the impact of the inertial forces (related to gas superficial velocity) and the impact of MLSS concentration (related to viscous forces evolution) on the mass transfer efficiency, in the range of operating conditions of our study. As previously discussed, the mass transfer efficiency is mainly controlled by viscous effects (correlated to MLSS concentration and shear rate). The mass transfer Stanton number also appears to slightly decrease with an increase in Froude number due to an increase in superficial gas velocity, which is more pronounced for clean water operating conditions than for AS ones. For clean water, this impact is mainly due to an increase in bubble size and related rising velocity. For AS operating conditions, the shearthinning behaviour induces a decrease of the mixed liquor apparent viscosity with an increase in gas superficial velocity, which could counterbalance the effect of superficial air velocity on bubble size in the bubble column filled with AS sludge. It could also explain the slighter impact (or the absence of impact in some cases) of superficial air velocity on oxygen transfer efficiency in AS operating conditions.

5. Conclusions The objective of this study was to evaluate the link of activated sludge physicochemical properties and airflow rate on the rheological behaviour and oxygen transfer in aerated bioreactors. First, the rheological measures showed that for both types of sludge (CAS and MBR), the mixed liquor suspended solid (MLSS) concentration plays a similar and significant role on the shearthinning behaviour of AS. Under a given shear rate, an increase in MLSS concentration causes an increment in apparent viscosity and accentuates the shear thinning behaviour (rise of the consistency index K and drop of the flow index n). The oxygen transfer tests in the bubble column (with a representative liquid depth of full-scale aeration tanks and fed continuously) showed that for a given superficial gas velocity, the gas hold-up (εG) and oxygen transfer coefficient (kLa) is reduced in activated sludge compared to clean water. A further depletion of the kLa coefficient was observed with an increase in the MLSS concentration. The magnitude of this impact was higher for the more concentrated sludge and can be attributed to the rise in apparent viscosity which may lead to the production of larger bubbles at the formation stage that lead to a reduction of the specific interfacial area (a). The experimental validation of this hypothesis for AS sludge operating conditions would be of great interest. From the experimental results, a powerful model was developed using dimensional analysis, correlating the kLa coefficient

with the superficial gas velocity (UG) and the apparent viscosity, integrating the shear-thinning rheological behaviour of AS. The originality of this correlation lies in the consideration of the impact of the physico-chemical properties, rheology and superficial gas velocity on the mean shear stress in the bubble column. This allowed to represent more accurately the hydrodynamics in the bubble column and hence to reproduce adequately the experimental oxygen transfer results for activated sludge. Finally, the apparent viscosity, integrating the shear-thinning behaviour of activated sludge rheology, is the key parameter to interpret the impact of operating conditions (as MLSS concentration and superficial gas velocity) on oxygen transfer efficiency in the bubble column.

Nomenclature COD D Dc F/M Fr g Ga H K kLa L MLSS MLVSS n NT ΔP=L Q R ReC Sc Scw StM Sp SRT SVI T UB UL UG VC

carbon oxygen demand (mg L 1) capillary tube diameter (m) column diameter (m) food to microorganism ratio (kg BOD5 (kgVSS) 1 d 1) Froude number (dimensionless) gravity constant (m s 2) Galileo number (dimensionless) column height (m) consistency index (Pa s) volumetric oxygen transfer coefficient (s 1) Length of capillary tubes (m) mixed liquor suspended solid concentration (g L 1) mixed liquor volatile suspended solid concentration (g L 1) flow index (dimensionless) transfer number (dimensionless) longitudinal pressure loss (Pa m 1) flow rate in the capillary tube (m3 s 1) capillary tube radius (m) column Reynolds number (dimensionless) column area (m2) column wall area (m2) mass transfer Stanton number (dimensionless) perforated surface of diffuser (m2) sludge retention time (d) sludge volume index (mL g 1) temperature (°C) bubble rise velocity (m s 1) superficial liquid velocity (m s 1) superficial gas velocity (m s 1) column volume (m3)

Greek letters

εG γ̇ μapp θ ρ ρML τ

overall gas hold-up (%) shear rate (s 1) apparent viscosity (Pa s) kLa temperature correction factor (dimensionless) density (kg m 3) density of the mixed liquor (kg m 3) shear stress (Pa)

Acknowledgements The authors wish to thank Degremont SA for supporting financially the PhD of Camilo Durán. They are also grateful to the


company SEE and employees who supervise and operate the WWTPs of Marolles en Hurepoix and Briis-sous-Forges, and to the SIA from Marolles, for authorizing access and completion of the experimental work on site. The participation of Pierre Mauricrace, Sylvain Pageot and Florian Maillard from Irstea in the implementation of the measures is gratefully acknowledged.

References Alves, S.S., Orvalho, S.P., Vasconcelos, J.M.T., 2005. Effect of bubble contamination on rise velocity and mass transfer. Chem. Eng. Sci. 60 (1), 1–9. ASCE, 2007. Measurement of Oxygen Transfer in Clean Water, ASCE/EWRI 2-06, pp. 42. ASCE, 1997. Standard Guidelines for In-Process Oxygen Transfer Testing, ASCE 18–96, pp. 60. Cornel, P., Wagner, M., Krause, S., 2003. Investigation of oxygen transfer rates in full scale membrane bioreactors. Water Sci. Technol. 47 (11), 313–319. Craig, V.S.J., 2004. Bubble coalescence and specific-ion effects. Curr. Opin. Colloid Interface Sci. 9 (1–2), 178–184. Capela, S., Héduit, A., Roustan, M. (2002) Influence of the water depth on fine bubble aeration efficiency in the presence of surfactants. In: Proceedings of IWA 3rd World Water Congress, Melbourne, AUS, 7–12 April 2002. pp. 8. Capela, S., Roustan, M., Héduit, A., 2001. Transfer number in fine bubble diffused aeration systems. Water Sci. Technol. 43 (11), 145–152. Dietrich, N., Francois, J., Jimenez, M., Cockx, A., Guiraud, P., Hébrard, G., 2015. Fast measurements of the gas–liquid diffusion coefficient in the gaussian wake of a spherical bubble. Chem. Eng. Technol. 38 (5), 941–946. Dupuis, D., 2008. Mesure de la viscosité-Viscosimètres et rhéomètres. Techniques de l'ingénieur. Métrologie relative aux fluides-Vitesses et débits Base documentaire: TIB402DUO, Ref. article: r2351. Fransolet, E., Crine, M., Marchot, P., Toye, D., 2005. Analysis of gas holdup in bubble columns with non-Newtonian fluid using electrical resistance tomography and dynamic gas disengagement technique. Chem. Eng. Sci. 60, 6118–6123. Garcia-Ochoa, F., Gomez, E., 2009. Bioreactor scale-up and oxygen transfer rate in microbial processes: an overview. Biotechnol. Adv. 27 (2), 153–176. Gardener, M., 2012. Statistics for Ecologist Using R and Excel. Pelagic publishing, Exeter. Germain, E., Nelles, F., Drews, A., Pearce, P., Kraume, M., Reid, E., Judd, S.J., Stephenson, T., 2007. Biomass effects on oxygen transfer in membrane bioreactors. Water Res. 41 (5), 1038–1044. Gillot, S., Capela-Marsal, S., Roustan, M., Héduit, A., 2005. Predicting oxygen transfer of fine bubble diffused aeration systems—model issued from dimensional analysis. Water Res. 39 (7), 1379–1387. Gillot, S., Capela, S., Heduit, A., 2000. Effect of horizontal flow on oxygen transfer in clean water and in clean water with surfactants. Water Res. 34 (2), 678–683. Gillot, S., Héduit, A., 2008. Prediction of alpha factor values for fine pore aeration systems. Water Sci. Technol. 57 (8), 1265–1269. Hebrard, G., Zeng, J., Loubiere, K., 2009. Effect of surfactants on liquid side mass transfer coefficients: a new insight. Chem. Eng. J. 148 (1), 132–138. Henkel, J., 2010. Oxygen Transfer Phenomena in Activated Sludge (PhD thesis). Technische Universität Darmstadt, Germany, pp. 178. Henkel, J., Lemac, M., Wagner, M., Cornel, P., 2009. Oxygen transfer in membrane bioreactors treating synthetic greywater. Water Res. 43 (6), 1711–1719. Jamnongwong, M., Loubiere, K., Dietrich, N., Hebrard, G., 2010. Experimental study of oxygen diffusion coefficients in clean water containing salt, glucose or surfactant: consequences on the liquid-side mass transfer coefficients. Chem. Eng. J. 165 (3), 758–768. Jimenez, M., 2013. Etude du transfert de matière gaz/liquide en milieux complexes: quantification du transfert d'oxygène par techniques optiques (PhD thesis). Institut National des Sciences Appliquées de Toulouse, France, pp. 210. Jin, B., Yin, P., Lant, P., 2006. Hydrodynamics and mass transfer coefficient in threephase air-lift reactors containing activated sludge. Chem. Eng. Process. 45 (7), 608–617.

165

Kestin, J., Sokolov, M., Wakeham, W.A., 1978. Viscosity of liquid water in range 8 °C to 150 °C. J. Phys. Chem. Ref. Data 7 (3), 941–948. Krampe, J., Krauth, K., 2003. Oxygen transfer into activated sludge with high MLSS concentrations. Water Sci. Technol. 47 (11), 297–303. Kulkarni, A.A., Joshi, J.B., 2005. Bubble formation and bubble rise velocity in gas liquid systems: a review. Ind. Eng. Chem. Res. 44 (16), 5873–5931. Mena, P.C., Ruzicka, M.C., Rocha, F.A., Teixeira, J.A., Drahos, J., 2005. Effect of solids on homogeneous-heterogeneous flow regime transition in bubble columns. Chem. Eng. Sci. 60 (22), 6013–6026. Mineta, R., Salehi, Z., Yoshikawa, H., Kawase, Y., 2011. Oxygen transfer during aerobic biodegradation of pollutants in a dense activated sludge slurry bubble column: actual volumetric oxygen transfer coefficient and oxygen uptake rate in p-nitrophenol degradation by acclimated waste activated sludge. Biochem. Eng. J. 53 (3), 266–274. Mori, M., Seyssiecq, I., Roche, N., 2006. Rheological measurements of sewage sludge for various solids concentrations and geometry. Process Biochem. 41 (7), 1656–1662. Ndoye, F.T., Erabit, N., Flick, D., Alvarez, G., 2013. In-line characterization of a whey protein aggregation process: aggregates size and rheological measurements. J. Food Eng. 115 (1), 73–82. NF EN 12255-15, 2004. Wastewater Treatment Plants. Part 15. Measurements of the Oxygen Transfer in Clean Water in Aeration Tanks of Activated Sludge Plants, pp. 22. NF T90-105-2, 1997. Qualité de l’eau – Dosage des matières en suspension – Méthode par centrifugation. p. 5. Painmanakul, P., Loubière, K., Hébrard, G., Mietton-Peuchot, M., Roustan, M., 2005. Effect of surfactants on liquid-side mass transfer coefficients. Chem. Eng. Sci. 60 (22), 6480–6491. Pittoors, E., Guo, Y., Van Hulle, S.W.H., 2014. Oxygen transfer model development based on activated sludge and clean water in diffused aerated cylindrical tanks. Chem. Eng. J. 243, 51–59. Ratkovich, N., Horn, W., Helmus, F.P., Rosenberger, S., Naessens, W., Nopens, I., Bentzen, T.R., 2013. Activated sludge rheology: a critical review on data collection and modelling. Water Res. 47 (2), 463–482. Rosenberger, S., Kubin, K., Kraume, M., 2002. Rheology of activated sludge in membrane bioreactors. Eng. Life Sci. 2 (9), 269–275. Rosso, D., Huo, D.L., Stenstrom, M.K., 2006. Effects of interfacial surfactant contamination on bubble gas transfer. Chem. Eng. Sci. 61 (16), 5500–5514. Rosso, D., Iranpour, R., Stenstrom, M.K., 2005. Fifteen years of offgas transfer efficiency measurements on fine-pore aerators: key role of sludge age and normalized air flux. Water Environ. Res. 7 (3), 266–272. Sanchez Pérez, J.A., Rodriguez Porcel, E.M., Casas López, J.L., Fernández Sevilla, J.M., Chisti, Y., 2006. Shear rate in stirred tank and bubble column bioreactors. Chem. Eng. J. 124 (1–3), 1–5. Sardeing, R., Painmanakul, P., Hébrard, G., 2006. Effect of surfactants on liquid-side mass transfer coefficients in gas–liquid systems: a first step to modeling. Chem. Eng. Sci. 61 (19), 6249–6260. Seyssiecq, I., Ferrasse, J.-H., Roche, N., 2003. State-of-the-art: rheological characterisation of wastewater treatment sludge. Biochem. Eng. J. 16 (1), 41–56. Tixier, N., Guibaud, G., Baudu, M., 2003. Determination of some rheological parameters for the characterization of activated sludge. Bioresour. Technol. 90 (2), 215–220. Wagner, M., Pöpel, H.J., 1996. Surface active agents and their influence on oxygen transfer. Water Sci. Technol. 34 (3–4), 249–256. Wagner, M., Cornel, P. Krause, S., 2002. Efficiency of different aeration systems in full scale membrane bioreactors. In: Proceedings of thr 75th WEFTEC Conference, Chicago, IL, pp. 10. Yang, F., Bick, A., Shandalov, S., Brenner, A., Oron, G., 2009. Yield stress and rheological characteristics of activated sludge in an airlift membrane bioreactor. J. Membr. Sci. 334 (1–2), 83–90. Zlokarnik, M., 1980. Koaleszenzphanomene im system gas forminglflussig und deren EinfluB auf den 02-Eintrag bei der biologischen Abwassereinigung. Korresp. Abwasser 27 (11), 728–734.