com
Environmental Engineering and Management Section, Department of Civil Engineering, Indian Institute of Technology, Kharagpur 721 302, India
Received 30 January 2007; received in revised form 23 May 2007; accepted 23 May 2007
Available online 20 July 2007
Abstract
Studies have been undertaken to explore the applicability of dierent kinetic models for the performance appraisal of upow anaerobic sludge blanket (UASB) reactors treating wastewater in the range of 3004000 mg COD/l. Three kinetic models namely, Monod,
Grau second-order, and Haldane model are considered for the analysis. Both linear and nonlinear regressions have been performed
to examine the best-t among the kinetic models. In this process, ve error analysis methods have been used to analyze the data. Apart
from optimization of kinetic coecients with minimization of associated errors, prediction of euent COD has also been undertaken to
verify the applicability of kinetic models. In both the cases, Grau second-order model is found to be the best class of t for wide range of
data sets in UASB reactor.
2007 Elsevier Ltd. All rights reserved.
Keywords: Error analysis; Grau model; Haldane model; Monod model; UASB reactor
1. Introduction
The upow anaerobic sludge blanket (UASB) reactor is
a well accepted anaerobic reactor types for treatment of
sewage as well as low and high strength industrial wastewaters. The success of high chemical oxygen demand (COD)
removal has led to its increasing applicability in the treatment of domestic wastewaters, especially under Mediterranean climatic conditions (Castillo et al., 1999). The
removal of organic matters is reasonably good; especially
for biodegradable organic carbon, making supplementary
treatment unnecessary in many instances (Schellinkhout
and Osorio, 1994).
An understanding of the process kinetics is vital in the
design, development and operation of UASB reactors.
Based on the biochemistry and microbiology of anaerobic
*
Corresponding author. Tel.: +91 3222 283440; fax: +91 3222 282254.
E-mail addresses: puspendu@civil.iitkgp.ernet.in, puspendubhunia@
yahoo.co.uk (P. Bhunia), ghangrekar@civil.iitkgp.ernet.in, ghangrekar@
iitkgp.ac.in (M.M. Ghangrekar).
0960-8524/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2007.05.053
process, kinetics provides a judicious basis for process analysis, control, and design. In addition to quantitative
description of the substrate utilization rates, process kinetics also deals with operational and environmental factors
aecting these rates.
Bacterial growth kinetics are based on two fundamental
relationships, i.e., growth rate and substrate utilization
rate. Various kinetic models are reported for anaerobic
processes (Castillo et al., 1999; Isik and Sponza, 2005;
Lokshina et al., 2001; Pavlostathis and Giraldo-Gomez,
1991) predominantly based on Monods equation or its
modications. However, the ample of information obligatory for many of these models, limits their eld application
globally. The kinetic method of designing a biological
treatment system is expected to be superior over empirical
methods, wherein a design is based on few empirical factors. The values of kinetic coecients are to be estimated
by means of regression analysis of experimental data generated from a lab-scale and/or pilot-scale study.
A review of literature reveals that most of the kinetic
models are nonlinear in nature (Ong, 1990). Thus, one
2133
Nomenclature
Q
Vb
Xo
Xe
X
So
Se
hc
l
lm
Kd
Ks
Y
Ks2
a, b
hH
Ki
n
p
y
2. Methods
Two identical UASB reactors (R1 and R2), made of
Plexiglas having eective volume 12.57 l, internal diameter
100 mm, and eective height 1.6 m were used in the study.
Synthetic feed having sucrose as carbon source was used in
this study. Synthetic wastewater used in this study was
mainly consists of 0.89 g sucrose, 1.5 g NaHCO3, 0.318 g
NH4Cl, 0.064 g MgSO4, 0.035 g K2HPO4, and 0.009 g
KH2PO4 per g of COD. Trace metals (Fe, Ni, Mn, Zn,
Co, Cu, and Mo) were added as per the composition suggested by Ghangrekar et al. (2005). The reactors were
seeded with sludge collected from bottom of a septic tank.
Analysis of parameters such as pH, COD, alkalinity, suspended solids (SS), and volatile suspended solids (VSS)
was accomplished as per APHA standard methods
(APHA, 1998). For euents from the reactors, COD was
determined for settled samples. Amount of sludge present
in the reactor was determined by measuring VSS proles
over height of the reactor at various stages under each
loading condition. Together with the data pertaining to
washout of sludge through euent and manual sludge
wastage, an integral picture was composed to nd out solid
residence time (SRT) in the reactor. The COD of synthetic
wastewater used for these experiments was between
300 mg l1 and 600 mg l1. Hydraulic retention time
(HRT) ranging from 8 to 4 h was used under dierent loading rates. All experiments were performed at ambient room
temperature, between 22 and 28 C. In each experimental
run, performance of the reactor was evaluated under constant organic loading rate (OLR).
3. Model descriptions
In all the models, equations are expressed considering
UASB reactor as two continuous ow completely mixed
reactors in series, namely sludge bed zone and sludge blanket zone. The substrate removal in sludge blanket zone is
neglected, i.e., substrate concentration at sludge blanket
2134
dX
Q
Q
Xo
X e l X Kd X
dt
Vb
Vb
dS
Q
Q
lX
So
Se
dt V b
Vb
Y
1
2
The association between specic growth rate and rate limiting substrate concentration can be expressed by Monod
Eq. (4) as
l
VbX
Q Xe
lm S e
K s Se
The constant lm indicates maximum growth rate of microorganisms when the substrate is being used at its maximum
rate, and Ks indicates the level of substrate concentration at
one-half the maximum specic substrate utilization rate. If
it is presumed that biomass concentration of inuent
wastewater, Xo is negligible and at steady-state conditions
(when dX
0, and dS
0), then the equations derived from
dt
dt
Eqs. (1)(4) are as
Table 1
Details of dierent performance parameters under steady-state conditions
Sl. no.
Inuent COD
(mg l1)
Euent COD
(mg l1)
Sludge bed
volume (l)
X (g l1)
Inuent ow
rate (l d1)
SRT
(d)
HRT
(h)
Reference
1
2
3
4
5
6
7
8
9
10
11
12
13
14
602.08
603.74
622.22
312.16
318.31
574.83
463.55
313.13
308
448.52
448.52
580.15
482.63
778.25
49.42
48.19
79.43
57.71
105.42
83.33
76.7
57.69
75.14
56.18
58.35
34.33
28.8
62.18
4.32
7.85
5.49
2.36
3.53
4
4
4
4.16
5.1
5.1
4.16
7.85
4.32
35.68
31.12
32.33
17.3
13.2
37.98
32.07
26.23
32.21
23.4
31.4
38.3
29.35
39.95
37.7
75.4
75.4
37.7
37.7
37.7
50.9
75.4
75.4
50.9
50.9
62.83
94.25
56.55
213.17
130.2
89.13
261.6
228.04
106.7
113.73
49.77
89.84
273.58
297.93
94.98
76.15
150.86
8
4
4
8
8
8
6
4
4
6
6
4.8
3.2
5.33
Present
experiment
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
913
1012
948
932
1023
3130
3214
3034
3024
3009
4054
3435
988
1019
3068
985
965
3253
1944
2068
58
60
55
29
82
210
256
282
241
235
938
439
59
43
294
35
64
164
94
111
2.5
2.5
2.5
2.5
2.5
2.3
2.6
2.7
2.83
3
2.7
2.9
2.7
4.5
4.5
8.5
10.5
9.5
7
6
26.63
26.02
24.55
25.95
24.48
30.28
29.78
40.49
42.69
40.45
38.82
45.78
28.91
32.275
22.68
35.82
35.975
44.315
43.26
36.96
6
8
10.322
12.533
14.953
7.345
10.03
14.392
18.113
19.592
19.917
19.672
12
12
12
36
36
36
24
24
215
159.5
115.5
104.5
79.5
91.05
51.12
47.95
40.68
40.49
22.28
33.26
85.59
145.82
44.76
86.3
143.73
76.93
143.5
135.88
16
12.12
9.3
7.7
6.4
13.07
9.57
6.67
5.305
4.9
4.82
4.88
8
16
16
8
16
16
12
12
Ghangrekar
(1998)
Q Y hc S o S e
V b 1 K d hc
K s 1 K d hc
Se
hc lm K d 1
5
6
Eqs. (5) and (6) are nonlinear in nature; hence, it is indispensable to transform them to equivalent linearized forms
so that linear regression can be used for estimating the values of kinetic constants. Two dierent linearized equations
can be framed to obtain values Y and Kd, which are as
Q S o S e 1 1
1
Kd
VbX
Y hc Y
and
1
Q So S e
Y
Kd
hc
V b X hc
The other forms of linearized equation reported in the literature for estimation of lm and Ks are as follows:
X Vb
1 Ks 1
1
Q S o S e Y lm S e lm
Q S o S e Y
Q S o S e Y
lm K s
X Vb
X V b Se
15
S o 1 haH
16
10
V b :S e X
S e
Se K s
Q S o S e K i k k
k
11
hH
a b hH
E
Q S o S e lm S e
lm
2135
13
If the second term of the right part of Eq. (13) is acknowledged as a constant, Eq. (14) will be obtained (Grau et al.,
1975; Buyukkamaci and Filibeli, 2002)
18
19
i1
20
i1
21
S o hH
b hH a
So Se
14
slope of the straight line using Eq. (7) are 0.083 mg VSS mg
COD1 and 0.006 d1, respectively as illustrated in
Fig. 1(a). The values of lm and Ks were determined from
Fig. 1(b) (using Eq. (9)) as 0.058 d1 and 226.1 mg l1,
respectively. The value of sludge yield Y is higher than that
reported for acetoclastic methanogens (0.010.05 mg
VSS mg COD1) and lower than acidogens (0.140.17 mg
VSS mg COD1) (Pavlostathis and Giraldo-Gomez,
1991). This is due to determination of overall yield for
the mixed culture. The magnitude of endogenous decay
coecient observed is well within the range reported for
acetoclastic methanogens (0.0040.037 d1) (Pavlostathis
and Giraldo-Gomez, 1991). The obtained lm value is at
lower side of the wide range (0.0410.912 d1) reported
for mixed and pure cultures of acetoclastic methanogens
at temperature 3537 C (Lokshina et al., 2001; Pavlostathis and Giraldo-Gomez, 1991; Fukuzaki and Nagai,
1990). This may be due to the lower temperature (22
28 C) and mixed culture used in the kinetic coecient
determination.
To examine the deviation of kinetic coecients due to
linearization, the kinetic data sets were analyzed with linear
0.7
y = 11.999x + 0.0725
2
R = 0.7646
0.6
Q(So-Se)/(VbX)
2136
0.5
0.4
0.3
0.2
0.1
0
0
0.01
0.02
0.03
0.04
0.05
1/SRT
b 1800
Monod Model
1600
Haldane Model
1400
(VbSeX)/(Q(So-Se))
1200
1000
800
600
R = 0.7747
400
200
0
0
200
400
600
800
1000
Se
Vb
Y Ks S e
Substituting
Then,
X lm S e
y1
Y K s S e
QS o S e
Vb
24
y 1 and
1
l Se
m
K d obtained from Eq: 7 and Eq: 8
hc K s S e
25
l Se
Substituting m
K d y 2 in the above equation
K s Se
26
Then, h1c y 2 .
Eq. (24) (nonlinear form of Monod equation) is used to
nd out Y, lm, and Ks and Eq. (26) is used for determination of Kd. In Eqs. (24) and (26), the rst y1 (or y2) denotes
the theoretical value and second y1 (or y2) denotes the
experimental value.
The values of kinetic coecient obtained from Fig. 1 for
Monod model are presented in the linear transform
(LTFM) columns and the kinetic coecients determined
by nonlinear regression using ve dierent error functions
are presented in Tables 2 and 3, respectively. From the
SNE values of Table 2, it is seen that signicant errors
are associated with the determination of kinetic coecients
Y, lm, and Ks by linearized model. The SNE values indicate
that the nonlinear regression with SAE parameter set produces the best overall t for the calculation of Y, lm, and
Ks. The optimum values of Y, lm, and Ks thus obtained
2137
1
1 hH =a
27
where y 3 SSoe .
The values of kinetic coecient associated with dierent
error functions are presented in Table 4. It is seen from
Table 2
Linear and nonlinear kinetic coecients (Ks, Y, lm) of Monod model
Ks
Y
lm
SSE
SAE
ARE
HYBRID
MPSD
SNE
LTFM
SSE
SAE
ARE
HYBRID
MPSD
226.1
0.083
0.058
1.15 10+8
5.09 10+4
2.66 10+1
5.90 10+4
3.56 10+1
4.71
225.77
0.08
0.057
1.14 10+8
5.13 10+4
2.72 10+1
6.06 10+4
3.65 10+1
4.78
226.1
0.07
0.048
1.17 10+8
5.03 10+4
2.62 10+1
5.83 10+4
3.49 10+1
3.84
226.1
0.097
0.062
1.41 10+8
5.24 10+4
2.52 10+1
6.01 10+4
3.34 10+1
4.83
252.69
0.085
0.061
1.2 10+8
5.01 10+4
2.56 10+1
5.77 10+4
3.39 10+1
4.63
275.69
0.059
0.042
1.39 10+8
5.05 10+4
2.47 10+1
6.09 10+4
3.31 10+1
4.77
Table 3
Linear and nonlinear kinetic coecient (Kd) of Monod model
Kd
SSE
SAE
ARE
HYBRID
MPSD
SNE
LTFM
SSE
SAE
ARE
HYBRID
MPSD
0.006
1.36 103
1.62 101
4.28 10+1
3.34 101
5.64 10+1
4.77
0.002
8.63 104
1.37 101
4.64 10+1
2.92 101
6.65 10+1
4.35
0.003
8.85 104
1.34 101
4.20 10+1
2.63 101
5.90 10+1
4.06
0.004
9.75 104
1.37 101
3.92 10+1
2.61 101
5.44 10+1
4.00
0.004
9.75 104
1.37 101
3.93 10+1
2.59 101
5.44 10+1
4.00
0.005
1.13 104
1.47 101
4.02 10+1
2.85 101
5.34 10+1
4.27
HRT/E
2138
20
18
16
14
12
10
8
6
4
2
0
y = 1.0429x + 0.5582
2
R = 0.9765
10
HRT
15
20
Table 4 that determination of kinetic constants using nonlinear regression with ARE, and HYBRID produces lowest
sum of normalized errors. It is also observed that determination of kinetic constants using Eq. (15) does not involve
much error as the SNE value of LTFM is very close to the
minimum values obtained from nonlinear regression.
4.3. Haldane model
In order to acquire the kinetic coecients of Haldane
model, Eq. (18) is plotted (Fig. 1(b)). The values of kinetic
coecients obtained from Fig. 1(b) are k = 0.55 d1,
Ks = 167.211 mg l1, and Ki = 3636.36 mg l1. On account
of the large value of Ki, indicating no inhibition eect of in
reactor COD values on the growth rate of microorganisms,
it is observed that Haldane model basically directs to
Table 4
Linear and nonlinear kinetic coecient (a) of Grau second-order model
a
SSE
SAE
ARE
HYBRID
MPSD
SNE
LTFM
SSE
SAE
ARE
HYBRID
MPSD
0.558
0.160
1.537
41.463
3.022
54.809
4.24
0.74
0.142
1.479
46.645
3.658
71.518
4.64
0.705
0.143
1.463
44.621
3.437
67.582
4.47
0.573
0.157
1.520
41.434
3.0246
55.723
4.24
0.562
0.159
1.533
41.456
3.022
55.033
4.24
0.459
0.186
1.691
43.177
3.258
51.783
4.54
Table 5
Kinetic coecients of Haldane model
Ks
k
Ki
SSE
SAE
ARE
HYBRID
MPSD
SNE
Eq. (17)
SSE
SAE
ARE
HYBRID
MPSD
167.212
0.55
3636.36
1500261
4960.408
33.508
7661.201
49.235
5
167.987
0.596
3636.359
1499391
4563.842
29.934
6166.84
43.218
4.5
167.228
0.594
3636.359
1180940
4930.915
29.948
6184.903
43.21302
4.36
167.212
0.594
3636.36
1180938
4572.722
33.270
6184.697
43.211
4.39
167.219
0.594
3636.359
1180941
4572.77
29.948
7653.526
43.213
4.48
167.212
0.594
3636.36
1180938
4572.722
29.947
6184.697
49.213
4.41
2139
Table 6
Details of experimental data sets used for prediction and validation by dierent kinetic models
Sl. no.
Inuent COD
(mg l1)
Euent COD
(mg l1)
Sludge bed
volume (l)
X
(g l1)
Inuent ow
rate (l d1)
SRT (d)
HRT (h)
Reference
1
2
3
4
5
6
3534
3007
3007
3278
3327
4005
259
213
224
302
342
590
6.08
10.21
12.56
10.95
17.73
13.48
47.15
35.09
30.75
40.68
36.1
28.95
17.71
23.88
31.65
46.83
69.57
71.82
107.98
98.75
76.25
58.37
58.61
35.51
16
12
9
6
4
4
Ghangrekar
et al. (2002)
2140