Science of the Total Environment 613–614 (2018) 1031–1036

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Comparison between classical Kelvin-Voigt and fractional derivative

Kelvin-Voigt models in prediction of linear viscoelastic behaviour of
waste activated sludge
Ehsan Farno a, Jean-Christophe Baudez b, Nicky Eshtiaghi a,⁎
a
RMIT University, Chemical, and Environmental Engineering, School of Engineering, Melbourne, Australia
b
LIST, Environmental Research and Innovation Department, Belvaux, Luxembourg

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• A method is proposed to fit the fraction-

al derivative Kelvin-Voigt to creep data.
• Fractional derivate Kelvin-Voigt model
predicts both creep and frequency
sweep data.
• Fractional derivate Kelvin-Voigt is more
accurate than Kelvin-Voigt for sludge.

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

Article history: Appropriate sewage sludge rheological models are essential for computational fluid dynamic simulation of
Received 5 June 2017 wastewater treatment processes, in particular aerobic and anaerobic digestions. The liquid-like behaviour of
Received in revised form 18 September 2017 sludge is well documented but the solid-like behaviour remains poorly described despite its importance for
Accepted 19 September 2017
dead-zone formation. In this study, classical Kelvin-Voigt model, commonly used for sludge in literature, were
Available online xxxx
compared with fractional derivative Kelvin-Voigt model regarding their predictive ability for describing the
Editor: Simon Pollard solid-like behaviour.
Results showed that the fractional Kelvin-Voigt model best fitted the experimental data obtained from creep and
Keywords: frequency sweep tests. Whereas, classical Kelvin-Voigt could not fit the frequency sweep data as this model is not
Kelvin-Voigt a function of angular velocity. Also, the Kelvin-Voigt model was unable to predict the creep data at low stresses.
Fractional derivative Kelvin-Voigt © 2017 Elsevier B.V. All rights reserved.
Viscoelastic model
Activated sludge
Rheological modelling

1. Introduction mixing, increase the biogas production, prevent foaming and solid accu-
mulation at the bottom of the digester, and utilize the digester volume
Anaerobic digestion of sludge, where microorganisms consume or- entirely (Bhattacharjee et al., 2017).
ganic matter in the absence of oxygen, can significantly reduce the Municipal sludge is a non-Newtonian material as it exhibits yielding,
sludge volume and produce biogas, a renewable energy source. Intimate shear thinning and viscoelastic behaviour even at low concentrations
contact between microorganisms and feed sludge is crucial for highly (Eshtiaghi et al., 2013; Farno et al., 2016; Ratkovich et al., 2013;
efficient digester operation which can only be achieved by proper Seyssiecq et al., 2003). Process modelling generally focused on the
mixing. Digester operation will fail if bacteria does not obtain nutrient. liquid-like behaviour and the shear thinning properties. The viscoelastic
Uniform distribution of organic matter and temperature via a good behaviour of sludge was often ignored for numerical simulations
(Terashima et al., 2009) due to the additional complexity that arises in
⁎ Corresponding author. the simulation of mixing phenomenon when all rheological behaviour
E-mail address: Nicky.eshtiaghi@rmit.edu.au (N. Eshtiaghi). of sludge (such as shear-thinning, viscoelasticity and yielding

https://doi.org/10.1016/j.scitotenv.2017.09.206
0048-9697/© 2017 Elsevier B.V. All rights reserved.
1032 E. Farno et al. / Science of the Total Environment 613–614 (2018) 1031–1036

behaviour) were considered together (Anderson et al., 2000; Arratia et Blair element) which represent intermediate characteristics lay be-
al., 2005). Sludge is simply considered rigid in the solid-like regime. tween the limits of purely viscous and purely elastic elements
However, it has been highlighted that the coexistence of viscoelasticity, (Jóźwiak et al., 2015).
yielding and shear thinning produce flow instabilities as well as chaotic The experimental observation of creep behaviour of sludge also
flow patterns specifically in the process of mixing (Arratia et al., 2005). showed a power-law relationship with time (Markis et al., 2014).
Thus, taking into account the solid-like behaviour is of crucial impor- In addition, the municipal sludge is composed of biopolymers
tance in process modelling. such as polysaccharides and proteins (Neyens and Baeyens,
An example of flow instabilities with these types of material is shear 2003) which exhibit scale-free features (Park et al., 2005). As a re-
banding which may also occur during sludge flow curve measurements sult of that, the fractional derivative models appear to be a good
(Anderson et al., 2000). Shear banding in a mixing tank such as anaero- candidate for the modelling of viscoelastic behaviour of sludge.
bic digester can create un-sheared regions or dead zones which reduces To our best knowledge, literature is scarce about the viscoelastic
the active volume of the digesters (Baudez et al., 2011). modelling of sludge and its model properties in which the retar-
Only few studies have investigated the mixing of fluids dation times obtained by fitting experimental data to the linear
possessing both viscoelasticity and shear-thinning characteristics creep compliance of generalised Kelvin-Voigt model (Baudez et
(Terashima et al., 2009). Indeed, Bhattacharjee et al. (2017) demon- al., 2011; Baudez et al., 2013) or Burger model (Baudez and
strated that the relaxation time of viscoelastic material impacts the Coussot, 2001). Therefore, it sounds interesting to compare the
mixing performance. The relaxation time can be used to estimate fractional derivative models and the classical models applied to
the significance of elastic forces to viscous forces in a fluid flow waste activated sludge.
(Poole, 2012) and the larger the relaxation time, the longer it takes This paper aims at modelling the linear viscoelastic behaviour of mu-
to reach a steady state regime. nicipal waste activated sludge using generalised Kelvin-Voigt and frac-
Mathematical modelling of the viscoelastic behaviour of mate- tional derivatives Kelvin-Voigt models. It also reports sludge
rials is then crucial to accurately design operational processes retardation time at different solid concentrations (3.5%, 4.5% and
(Eldred et al., 1995). The viscoelastic models can be used in compu- 6.1%), different temperatures (from 20 to 80 °C) and different thermal
tational fluid dynamic simulation as well as in the estimation of ma- treatment time (up to 1 h).
terial characteristics such as relaxation and retardation times. The Creep compliance function as well as storage and loss modulus
relaxation time is calculated via fitting the experimental data obtain- calculated from classical Kelvin-Voigt model and fractional deriva-
ed from the stress relaxation test to the relaxation modulus derived tives Kelvin-Voigt model were fitted to experimental data and their
from the viscoelastic models. Similarly, the retardation time is calcu- error in the prediction of experimental data were compared. A new
lated via fitting the experimental data from the creep test to the procedure is also proposed for finding the retardation time (or relax-
creep compliance function derived from the viscoelastic models. Re- ation time).
laxation spectrum of a material (and the related retardation spec-
trum calculated from creep data) is derived from the generalised 2. Materials and methods
viscoelastic models and may reveal the nature of relaxation process-
es that a material has undergone (Vincent, 1990). 2.1. Experimental
Classical viscoelastic models such as the Kelvin-Voigt model are
widely used to describe the viscoelastic behaviour of sludge. They Waste activated sludge was collected from one of the Melbourne
are usually made of two or more purely viscous element (dashpot) wastewater treatment plants and stored in sealed containers at 4 °C
and purely elastic element (spring) in different combinations (e.g. for one month before performing the tests. This procedure helps sludge
parallel or series). Since sludge are neither purely elastic and nor become more stabilized during the experiments (Curvers et al., 2009).
purely viscous, increase in the number of these elements to a series Sludge was concentrated from initial concentration of 3.5% to higher
(generalised viscoelastic models) results in more accurate prediction concentration (up to 6.1%) using vacuum filtration process. Table 1 pro-
of a material natural behaviour. In fact, an increase in the number of vides sludge compositions.
elements in series spreads the decays of stress over longer period of Experimental data was collected using a stress controlled rheometer
time (Vincent, 1990). (HR3 rheometer from TA Instrument) equipped with a cup and bob ge-
But the practical use of generalised classical viscoelastic models is ometry. Before performing each rheological test, sample was pre-
limited because they lack the simplicity of the standard models and sheared for 10 min following by 1 min rest. Experimental data was ob-
they induce arbitrary decisions for assigning an adequate number of el- tained from two tests:
ements to the model (Nicolle et al., 2010; Tschoegl, 2012). Furthermore,
1. Creep test from a very low stress to the yielding point (where an
the creep compliance and the relaxation modulus calculated from clas-
abrupt increase in strain in log-log graph of strain-stress was ob-
sical viscoelastic models are usually an exponential function of time;
served). Further details of experimental procedure can be found in
however, the experimental observation of relaxation modulus in some
Markis et al. (2016).
complex materials such as colloidal gels (Yin and Solomon, 2008),
2. Frequency sweep performed between 1 and 100 rad/s at constant
bread dough (Ng and McKinley, 2008) and biopolymers (Jóźwiak et
deformation of 0.1% in linear viscoelastic region (LVE). The LVE re-
al., 2015) shows a power-law decays of time. The power-law relaxation
gion, where the storage and loss modulus are independent of strain,
modulus of these materials is considered to be related to scale-free
was determined using an amplitude sweep test (0.1–100%) at con-
(power-law distribution of network) fractal microstructure of them
stant angular frequency of 0.1 rad/s. Further details of experimental
(Sokolov et al., 2002).
procedure can be found in Farno et al. (2016).
A more realistic and accurate viscoelastic model for complex mate-
rials such as sludge can be obtained using fractional derivatives visco-
elastic models (Schiessel et al., 1995). Fractional derivatives
Table 1
viscoelastic models give better results in the linear viscoelastic region Sludge composition.
and in wider frequency range (Papoulia et al., 2010). These models ap-
Total Solid (w/w %)a Volatile solid (w/w %)a Soluble COD [mg/L]
pear to be capable of mimicking the behaviour of more complex mate-
rials among which biopolymers (Jóźwiak et al., 2015) which fit to 3.5% 74% 850
power-law relaxation modulus (Madbouly et al., 2013). The advantage 6.1% 68% 2550

of fractional viscoelastic models is due to fractional elements (Scott- a

w/w: weight per weight.
 E. Farno et al. / Science of the Total Environment 613–614 (2018) 1031–1036 1033

Average coefficients of variation (CV) were 10% for three repeats of permanent structures including cell walls and the second term stands
the creep tests and 0.5% for three repeats of the frequency sweep for the gel-like behaviour of cells' internal filamentous cytoskeleton.
tests. Coefficient of variation is defined as follows: Eq. (6) converts to Kelvin-Voigt model if α = 1.
The fractional-order derivative is defined as follows:
Standard deviation
CV ¼  100 ð1Þ Z t
mean 1 d xðτÞ
Dα ½xðt Þ ¼ dτ 0bαb1 ð7Þ
Γ ð1−αÞ dt 0 ðt−τ Þα

2.2. Rheological modelling where Γ(x) is complete Gamma function. The Eq. (6) shows memory
features due to the (t-τ)− α term. The variations of storage and loss
Generalised Kelvin-Voigt model is composed of a series of parallel modulus of this system with frequency (ω) are given as follows:
spring and dashpots (see Fig. 1a) (Eldred et al., 1995). In a parallel con-
figuration, total shear stress is the sum of shear stress acting on each el- 0
π 
G ðωÞ ¼ E1 ωα cos α þ E2 ð8Þ
ement while total deformation is equal to the deformation of each 2
element:
π 
G″ ðωÞ ¼ E1 ωα sin α ð9Þ
σt ¼ σ V þ σ E ð2Þ 2

γt ¼ γV ¼ γE ð3Þ when ω tends to zero, G′ goes to a static load (E2) while G″ goes to zero
which is expected and similar to the Kelvin-Voigt model. When ω tends
where σV [Pa] and σE [Pa] are shear stresses acting on dashpot and to infinity; however, the modulus follow power-law functionality of fre-
spring, respectively. γV and γE are the deformation of dashpot and quency (ωα) with α typically between 0 and 1. This rate of modulus
spring respectively. growth with frequency is slower than the linear growth for Kelvin-
In this equation, dashpot is an ideal viscous element obeying the Voigt model (Eldred et al., 1995). The strain response of this model to
Newton's law and spring is an ideal elastic element obeying the Hook's a step stress is given (Schiessel et al., 1995) as follows:
law.



α 
σ t
dγðt Þ γðt Þ ¼ 1−Eα − ð10Þ
σ t ðt Þ ¼ η þ Eγðt Þ ð4Þ E τ
dt
where Eα is generalised Mittag-Leffler function of order α, E = E1(τ1 / τ)α,
where η [Pa.s] and E [Pa] are dashpot and spring constants, respectively.
τ = (E1τα1 / E2)
1/α
, τ1 = η1 / E1, with η1 and E1 are spring and dashpot con-
Strain response of generalised Kelvin-Voigt model (composed of a
stants of fractional element. Mittag-Leffler function of order α and μ is
series of n parallel spring-dashpot with equal time constant) to an
defined:
input step stress (σ) is a stretched exponential function of time
(Mainardi and Spada, 2011) as follows: X
∞
xn



  Eα;μ ðxÞ ¼ ð11Þ
σ t m n¼0
Γ ðαn þ μ Þ
γ ðt Þ ¼ 1− exp − ð5Þ
E τ
It should be noted that μ is assumed to be equal to one in Eq. (10).
where γ [−] is strain, σ [Pa] stress, E [Pa] spring constant (elastic mod- Generalised Kelvin-Voigt and fractional derivative Kelvin-Voigt
ulus). τ [s] =η/E is retardation time in which η [Pa.s] is dashpot constant models presented in Eqs. (5), (8), (9) and (10) were fitted to experi-
(viscous modulus). The three fitting parameters of this model are E, η mental data using non-linear least square function in MATLAB. The
and m. MATLAB function for evaluation of the Mittag-Leffler function were ex-
In the dynamic mode, according to Kelvin-Voigt model, storage tracted from the work of Garrappa (2015). It is worth noting that, Eqs.
modulus is not a function of frequency while loss modulus linearly in- (5) and (10) have three fitting parameters. In Eq. (5), the m value can
creases with increasing angular frequency. However, this is not the be an arbitrary chosen or can be found as a fitting parameter. Whereas,
case for many materials. Furthermore, the Kelvin-Voigt model does in Eq. (10), alpha is a fitting parameter which can be also found as a
not keep any time history or memory features (Eldred et al., 1995). fitting parameters in Eqs. (8) and (9). This can be a great advantage
Simplest fractional derivative Kelvin-Voigt model composing three for Eq. (10) over Eq. (5). One can find alpha by fitting Eqs.(8) and (9)
fitting parameter is formed by a Scott-Blair element (springpot) in par- to experimental frequency sweep data which is a more reliable test
allel to an elastic element (spring) (See Fig. 1b) (Mainardi and Spada, (check reported CV in Section 2.1.) and then substitute it in Eq. (10) to
2011). find other fitting parameters. To fit the model, weighted non-linear
α
least square function should be used once alpha is fixed. In the result
d γ ðt Þ section, the uncertainty over the m value in Eq. (5) will be highlighted.
σ t ðt Þ ¼ E1 τα1 α þ E2 γ ðt Þ 0bαb1 ð6Þ
dt Root-mean-square-error (RMSE) (as shown Equation11) was used to
evaluate the goodness of fit of the models.
where E1 [Pa·sα] and α are the constants of fractional element and τ1 is
a characteristics time of fractional element. E2 [Pa] is the spring con- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
stant. As it was described for biological soft tissues (Nicolle et al., ∑ðmeasured value−Predicted valueÞ
RMSE ¼ ð11Þ
2010), the first term in Eq. (6) represents the elastic response of the number of measurement points

Fig. 1. Schematic representation of (a) generalised Kelvin-Voigt model and (b) Fractional derivative Kelvin-Voigt model.
1034 E. Farno et al. / Science of the Total Environment 613–614 (2018) 1031–1036

3. Results and discussion

In LVE region, fractional derivative Kelvin-Voigt model success-

fully predicts the storage and loss modulus which experimentally
obtained from frequency sweep tests (Fig. 2 a–c). Fractional deriv-
ative Kelvin-Voigt model predicts well these experimentally ob-
tained modulus (G′, G′′) at different solids concentration (3.5%
and 6.1% WAS at 20 °C, Fig. 2.a), different temperature ranges (20,
50 and 80 °C, Fig. 2.b), and different thermal treatment time (1,
30 and 60 min thermal treatment at 60 °C, Fig. 2.c). In Fig. 2, dash
line represents fractional Kelvin-Voigt model fitted to experimen-
tal data. But because the storage modulus in Kelvin-Voigt model is
not a function of angular velocity, this model failed at representing
the experimental data so it was not shown in Fig. 2. The existence of
biopolymers in sludge may be the reason for the variation of stor-
age and loss modulus with weak power-law functions of frequency.
Also, the variation of storage and loss modulus with frequency in
Fig. 2 suggests that the relaxation time cannot be estimated simply
from the graph as there is no evident crossover point.
Table 2 presents the fitting parameters of fractional derivative
Kelvin-Voigt model as well as the root-mean-square error of fitting re-
sults for 6.1% WAS. Both elastic modulus (E1) and fractional modulus
(E2) (refer to Table 2) decreased with an increase in temperature and
treatment time. α slightly increased with an increase in temperature
and time which means sludge exhibited higher viscous characteristics
at higher temperature and longer thermal treatment. This result is con-
sistent with Farno's experimental observation (Farno et al., 2014, 2016).
In Fig. 3, the strain prediction of both Kelvin-Voigt and fractional
Kelvin-Voigt models are compared with strain data obtained experi-
mentally from the creep test in LVE region. Fig. 3.a presents the variation
of strain with creep time (in LVE) for two solid concentrations (3.5% and
4.5% WAS).
Fig. 3.b and c present the impact of treatment temperature (for 6.1%
WAS at 20, 60 and 80 °C) and treatment time (for 6.1% WAS at 60 °C
after 1, 30 and 60 min) on the variation of strain with creep time,
respectively.
Table 3 presents the fitting parameters of creep responses (corre-
sponding to equal stresses) for Kelvin-Voigt and fractional derivative
Kelvin-Voigt models (Eq. (5) and Eq. (10)) for the above mentioned
solids concentration, temperatures and treatment times. The retarda-
tion time (τ) calculated as a fitting constant of Eq. (10) is constant at
two concentrations of 3.5 and 4.5%. Whereas, the retardation time cal-
culated from Eq. (5) slightly decreases with increase in concentration.
Fig. 2. Variation of storage and loss modulus with angular velocity (a) for 3.5% and 6.1% Also, the retardation time from Eq. (9) remains constant with treatment
WAS at 20 °C (b) for 6.1% WAS at different temperatures (50, 60 and 80 °C) and (b) for
temperature and time; while, the retardation time from Eq. (5) slightly
6.1% WAS at 50 °C after different thermal treatment time (1, 30 and 60 min); dash lines
represent the fractional Kelvin-Voigt predictions. fluctuates with treatment time. In Table 3, the spring constant de-
creased with increase in both time and temperature which means
sludge exhibited lower elastic characteristics at higher temperatures
as well as longer thermal treatment. On the contrary, the spring

Table 2
Fitting parameters of fractional Kelvin-Voigt model (Eqs. (6) and (7)) for 6.1% WAS at different temperature and different treatment time.

1 min 15 min 30 min 45 min 60 min 1 min 15 min 30 min 45 min 60 min

20 °C 50 °C 70 °C
E1[Pa] 416 ± 1 341 ± 5 336 ± 4 314 ± 3 300 ± 3 287 ± 3 200 ± 3 185 ± 3 167 ± 3 154 ± 3 158 ± 2
E2[Pa] 539 ± 1 460 ± 3 427 ± 3 390 ± 2 356 ± 2 350 ± 2 360 ± 2 305 ± 2 285 ± 2 271 ± 2 258 ± 1
α 0.15 0.16 0.16 0.16 0.17 0.17 0.16 0.16 0.16 0.16 0.17
±4 × 10−4 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±10−3
RMSE 2.35 8.94 7.76 6.80 6.41 6.00 5.25 5.82 5.00 4.75 4.49

60 °C 80 °C
E1[Pa] 220 ± 5 187 ± 3 171 ± 3 167 ± 3 155 ± 2 100 ± 6 91 ± 2 95 ± 2 79 ± 1 69 ± 1
E2[Pa] 424 ± 3 218 ± 2 202 ± 2 191 ± 1 182 ± 1 258 ± 4 137 ± 1 110 ± 1 103 ± 1 99 ± 1
α 0.16 0.17 0.17 0.18 0.18 0.17 0.17 0.18 0.18 0.18
±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±4 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3 ±2 × 10−3
RMSE 6.52 4.68 4.45 4.64 3.61 7.23 2.83 2.97 2.02 2.07
 E. Farno et al. / Science of the Total Environment 613–614 (2018) 1031–1036 1035

Fig. 3. Variation of strain with creep time (a) for 3.5 and 4.5% WAS at 20 °C (b) for 6.1% WAS at different temperatures (20, 60 and 80 °C) and (b) for 6.1% WAS at 60 °C after different
thermal treatment time (1, 15 and 60 min); KV and FKV represent Kelvin-Voigt and fractional Kelvin-Voigt models, respectively.

constant increased with concentration, therefore, sludge with higher creep test (refer to Fig. 3). Whereas, Kelvin-Voigt model (Eq. (5)) accu-
concentration exhibits stronger elastic responses. rately predicts the experimental data at the higher creep stresses. Frac-
By finding the alpha value (presented in Table 2) through frequency tional derivative Kelvin-Voigt (Eq. (10)) is a better model due to the
test and substituting in Eq. (10), fractional derivative Kelvin-Voigt bet- following reasons: First: in the creep test, the measured stain in re-
ter fits to the experimental data at lower creep stresses collected from sponse to higher stresses naturally contains deviation as they are close
to the limit of linear viscoelastic region. Second: in the creep test and
at higher stresses, the corresponding strain response increases substan-
Table 3
Fitting parameters of classical and fractional Kelvin Voigt models (Eq. (4) and Eq. (9)). tially so that the least-square function in MATLAB finds the least sum of
squared error when the fitted curve best predicts the higher strains.
Model E [Pa] 1/τ [s−1] α or m
Third: the m value shows no meaningful pattern. In contrast, alpha
3.5% FKV 27 ± 2 0.0001 0.2 shows a clear meaningful pattern; increase in treatment temperature
KV 0.300 2E-06 0.64 ± 0.01
and time slightly increases alpha. Alpha closer to one is due to an in-
4.5% FKV 44 ± 6 0.0001 0.5
KV 0.236 9E-05 1.73 ± 0.2
crease in viscous characteristics of Scott-Blair element (springpot)
which means sludge shows more viscous characteristics at higher tem-
Model E [Pa] 1/τ [s−1] α or m peratures and longer treatment time. This is consistent with literature
6.1% 20 °C FKV 730 ± 9 0.0001 0.16 (Farno et al., 2016).
KV 2.60 ± 0.9 4E−05 1.31 ± 0.03 As a result of that, the best way to find the retardation time (or relax-
50 °C FKV 360 ± 6 0.0001 0.16 ation time) is to fit the frequency sweep data to Eqs. (7)–(8) to find the
KV 0.77 ± 0.1 8E−06 1.03 ± 0.01
alpha and then fit the creep responses to Eq. (10) using calculated alpha
80 °C FKV 55 ± 5 0.0001 0.17
KV 0.08 ± 0.06 4E−06 0.91 ± 0.02 (from Eqs. 7–8) as a constant. This procedure will help to reduce the
fitting error.
Model E [Pa] 1/τ [s−1] α or m

6.1% at 60 °C 1 min FKV 260 ± 8 0.0001 0.16 4. Conclusion

KV 1.06 ± 0.01 3E−06 0.81 ± 0.01
15 min FKV 130 ± 10 0.0001 0.17 The current work is concerned with modelling of the linear visco-
KV 2.65 ± 0.01 1E−06 0.54 ± 0.01
elastic behaviour of different concentrations of WAS at 20–80 °C up to
30 min FKV 95 ± 7 0.0001 0.17
KV 3.58 ± 0.01 1E−06 0.46 ± 0.01 one hour thermal treatment time. The results obtained from the fre-
45 min FKV 83 ± 6 0.0001 0.18 quency sweep tests clearly show that storage modulus is a function of
KV 5.46 ± 0.01 6E−07 0.38 ± 0.01 frequency therefore a classical Kelvin-Voigt model cannot fit the exper-
60 min FKV 68 ± 6 0.0001 0.18 imental data. Fractional derivate model successfully predicted the ex-
KV 6.26 ± 0.01 2E−06 0.40 ± 0.01
perimental data collected from creep and frequency sweep tests in
1036 E. Farno et al. / Science of the Total Environment 613–614 (2018) 1031–1036

LVE region. It was also shown that using the proposed method, the frac- Madbouly, S.A., Xia, Y., Kessler, M.R., 2013. Rheological behavior of environmentally
friendly castor oil-based waterborne polyurethane dispersions. Macromolecules 46,
tional derivative Kelvin-Voigt model can predict the creep response 4606–4616.
more accurately compared to classical Kelvin-Voigt model. In conclu- Mainardi, F., Spada, G., 2011. Creep, relaxation and viscosity properties for basic fractional
sion, the optimal model for describing the experimental data obtained models in rheology. Eur. Phys. J. Spec. Top. 193, 133–160.
Markis, F., Baudez, J.-C., Parthasarathy, R., Slatter, P., Eshtiaghi, N., 2014. Rheological char-
from the creep and frequency sweep tests of sludge is fractional deriva- acterisation of primary and secondary sludge: Impact of solids concentration. Chem.
tive Kelvin-Voigt model. Eng. J. 253, 526–537.
Markis, F., Baudez, J.-C., Parthasarathy, R., Slatter, P., Eshtiaghi, N., 2016. The apparent vis-
cosity and yield stress of mixtures of primary and secondary sludge: impact of vol-
References ume fraction of secondary sludge and total solids concentration. Chem. Eng. J. 288,
577–587.
Anderson, P.D., Galaktionov, O.S., Peters, G.W., van de Vosse, F.N., Meijer, H.E., 2000.
Neyens, E., Baeyens, J., 2003. A review of thermal sludge pre-treatment processes to im-
Mixing of non-Newtonian fluids in time-periodic cavity flows. J. Non-Newtonian
prove dewaterability. J. Hazard. Mater. 98, 51–67.
Fluid Mech. 93, 265–286.
Ng, T.S., McKinley, G.H., 2008. Power law gels at finite strains: the nonlinear rheology of
Arratia, P.E., Voth, G.A., Gollub, J.P., 2005. Stretching and mixing of non-Newtonian fluids
gluten gels. J. Rheol. 52, 417–449 (1978–present).
in time-periodic flows. Phys. Fluids 17, 053102 (1994–present).
Nicolle, S., Vezin, P., Palierne, J.F., 2010. A strain-hardening bi-power law for the nonlinear
Baudez, J.-C., Coussot, P., 2001. Rheology of aging, concentrated, polymeric suspensions:
behaviour of biological soft tissues. J. Biomech. 43, 927–932.
application to pasty sewage sludges. J. Rheol. 45, 1123–1139 (1978–present).
Papoulia, K.D., Panoskaltsis, V.P., Kurup, N.V., Korovajchuk, I., 2010. Rheological represen-
Baudez, J.C., Markis, F., Eshtiaghi, N., Slatter, P., 2011. The rheological behaviour of anaer-
tation of fractional order viscoelastic material models. Rheol. Acta 49, 381–400.
obic digested sludge. Water Res. 45, 5675–5680.
Park, K., Lai, Y.-C., Ye, N., 2005. Self-organized scale-free networks. Phys. Rev. E 72,
Baudez, J.C., Slatter, P., Eshtiaghi, N., 2013. The impact of temperature on the rheological
026131.
behaviour of anaerobic digested sludge. Chem. Eng. J. 215–216, 182–187.
Poole, R., 2012. The Deborah and Weissenberg numbers. British Soc. Rheol. Rheol. Bull. 53,
Bhattacharjee, P.K., Kennedy, S.P., Xu, Q., Eshtiaghi, N., Parthasarathy, R., 2017. The influ-
32–39.
ence of injection velocity and relaxation time on the spreading of tracers in viscoelas-
Ratkovich, N., Horn, W., Helmus, F.P., Rosenberger, S., Naessens, W., Nopens, I., Bentzen,
tic liquids agitated by submerged, recirculating jets with low reynolds numbers.
T.R., 2013. Activated sludge rheology: a critical review on data collection and model-
AICHE J. 63, 3132–3140.
ling. Water Res. 47, 463–482.
Curvers, D., Saveyn, H., Scales, P.J., Van der Meeren, P., 2009. A centrifugation method for
Schiessel, H., Metzler, R., Blumen, A., Nonnenmacher, T., 1995. Generalized viscoelastic
the assessment of low pressure compressibility of particulate suspensions. Chem.
models: their fractional equations with solutions. J. Phys. A Math. Gen. 28, 6567.
Eng. J. 148, 405–413.
Seyssiecq, I., Ferrasse, J.H., Roche, N., 2003. State-of-the-art: rheological characterisation of
Eldred, L.B., Baker, W.P., Palazotto, A.N., 1995. Kelvin-Voigt versus fractional derivative
wastewater treatment sludge. Biochem. Eng. J. 16, 41–56.
model as constitutive relations for viscoelastic materials. AIAA J. 33, 547–550.
Sokolov, I.M., Klafter, J., Blumen, A., 2002. Fractional kinetics. Phys. Today 55, 48–54.
Eshtiaghi, N., Markis, F., Yap, S.D., Baudez, J.-C., Slatter, P., 2013. Rheological characterisa-
Terashima, M., Goel, R., Komatsu, K., Yasui, H., Takahashi, H., Li, Y.Y., Noike, T., 2009. CFD
tion of municipal sludge: a review. Water Res. 47, 5493–5510.
simulation of mixing in anaerobic digesters. Bioresour. Technol. 100, 2228–2233.
Farno, E., Baudez, J.C., Parthasarathy, R., Eshtiaghi, N., 2014. Rheological characterisation of
Tschoegl, N.W., 2012. The Phenomenological Theory of Linear Viscoelastic Behavior: An
thermally-treated anaerobic digested sludge: impact of temperature and thermal his-
Introduction. Springer Science & Business Media.
tory. Water Res. 56, 156–161.
Vincent, J.F.V., 1990. Structural Biomaterials. Princeton University Press.
Farno, E., Baudez, J.C., Parthasarathy, R., Eshtiaghi, N., 2016. The viscoelastic characterisa-
Yin, G., Solomon, M.J., 2008. Soft glassy rheology model applied to stress relaxation of a
tion of thermally-treated waste activated sludge. Chem. Eng. J. 304, 362–368.
thermoreversible colloidal gel. J. Rheol. 52, 785–800 (1978–present).
Garrappa, R., 2015. Numerical evaluation of two and three parameter Mittag-Leffler func-
tions. SIAM J. Numer. Anal. 53, 1350–1369.
Jóźwiak, B., Orczykowska, M., Dziubiński, M., 2015. Fractional generalizations of Maxwell
and Kelvin-Voigt models for biopolymer characterization. Pub. Lib. Sci. One 10,
e0143090.