Journal of Chemical Engineering of Japan, Vol. 52, No. 11, pp.

811–821, 2019 Research Paper

Orthogonal Design-Based Grey Relational Analysis for

Influence of Factors on Calcination Temperature in
Shaft Calciner
Jindi Huang1, Jing Li 2, Mingzhou Li 3 and Kang Yan3
1
School of Energy and Mechanical Engineering, Jiangxi University of Science and Technology, Nanchang,
330013, P.R.China
2
School of Metallurgy and Environment, Central South University, Changsha, 410083, P.R.China
3
Faculty of Materials Metallurgy and Chemistry, Jiangxi University of Science and Technology, Ganzhou,
341000, P.R.China

Keywords: Shaft Calciner, Petroleum Coke, Mathematical Model, Grey Relational Analysis, Orthogonal Design

Calcination temperature is the key factor affecting the quality of calcined petroleum coke and dominated by calcination
parameters in a shaft calciner. In this study, a method for evaluating the effects of various factors on the temperature dis-
tribution in shaft calciners is proposed. A two-dimensional transient mathematical model was developed to describe the
complex gas–solid coupled mass and heat transfer problems that occur during petroleum coke calcination in shaft cal-
ciners. In this model, a three-parallel-distributed activation energy model (DAEM)-reaction model was used to describe
the pyrolysis kinetics of petroleum coke, and a dichotomy method was used to guarantee constant temperature in the
flue of Layer 2 (T2) to meet the operational control requirements in actual production. Then, a statistical method called
orthogonal design-based grey relational analysis was used to quantize the influence degree of salient factors (discharge
rate per pot [DRPP], moisture content [MC], volatile content [VC], and volatile distribution ratio [VDR] in the flue of Layer
1) on the temperature distribution in the shaft calciner. An analysis of variance of the grey relational degree showed that
VC was the most influential factor, followed by DRPP and MC; VDR was the least influential factor behind calcination
temperature. Therefore, VC is the key factor affecting the calcination temperature and must be strictly controlled during
production.

experimental and numerical studies have been conducted.

Introduction
Edwards (2015) found a close relationship between calci-
Petroleum coke calcination is an important link in the nation temperature and real and vibrated bulk densities of
carbon anode preparation used in the aluminum industry. CPC. Liu (2016) highlighted a remarkable correspondence
The primary functions of this process are removing mois- between CPC quality and the maximum calcination tem-
ture and volatile matter in green petroleum coke (GPC) and perature in the shaft calciner. However, information about
achieving appropriate crystalline structures (Martins et al., the true temperature distribution in shaft calciners remains
2001; Xiao et al., 2016a; Zhong et al., 2018; Li et al., 2018a). unclear due to the sealing characteristics of pot and the limi-
Calcination technology considerably affects the quality of tations of existing measurement techniques. Different mod-
calcined petroleum coke (CPC), specifically in terms of bulk els have been proposed to solve this problem and describe
density and grain size distribution. Therefore, the control of the petroleum coke calcination process using computational
calcination parameters, which significantly influences the fluid dynamics. Zhou and Luo (2014) proposed a simpli-
quality of CPC, should be optimized in petroleum coke cal- fied three-dimensional (3D) model without considering the
cination. endothermic pyrolysis of petroleum coke. In their work, the
Shaft calciners have many merits, such as low carbon loss thermal characteristics, flow field, and temperature field in
rate, and excellent CPC quality, which make them promis- a shaft calciner were simulated. Xiao et al. (2016b) devel-
ing equipment for petroleum coke calcination in the carbon oped a 3D mathematical model to consider petroleum coke
industry. In a shaft calciner, the calcination parameters af- pyrolysis and investigated the distribution characteristics
fecting CPC quality include calcination temperature, dura- of temperature, species, and velocity fields in a shaft cal-
tion, and volatile burning. Calcination temperature exerts ciner. In addition, Xiao et al. (2016a, 2016b) established a
the largest influence on the quality of CPC; thus, many steady-state two-dimensional (2D) numerical model and
studied the effect of calcination technology parameters on
the temperature distribution in the pot using single-factor
Received on May 11, 2018; accepted on August 27, 2019
DOI: 10.1252/jcej.18we128
analysis. Their research emphasized that discharge rate per
Correspondence concerning this article should be addressed to pot (DRPP), excess air coefficient (EAC), volatile content
J. Huang (E-mail address: [email protected]) and J. Li (E-mail address: (VC), moisture content (MC), and volatile distribution ratio
[email protected]). (VDR) in the flue of Layer 1 had remarkable effects on the

Vol. 52  No.©11 

Copyright 2019 
2019
The Society of Chemical Engineers, Japan 811
temperature distribution in the pot. However, no appropri- Table 1 Pyrolysis kinetic parameters of Fushun petroleum coke
ate evaluation method has been proposed to investigate the Ci ki E0i σi
influence degrees of calcination parameters on the tempera- Petroleum coke
[−] [s−1] [kJ/mol] [kJ/mol]
ture in shaft calciners.
Pseudo component #1 0.23 7.62×103 84.16 10.23
Orthogonal design-based grey relational analysis (GRA)
Pseudo component #2 0.58 7.62×103 111.15 18.58
(Kadier et al., 2015; Zuo et al., 2016) is an efficient evalua- Pseudo component #3 0.19 7.62×103 192.89 18.42
tion and optimization method that is widely used in many
fields, such as engineering technology (Kuo et al., 2008;
Wang et al., 2015), and buildings (Lee and Lin, 2011). Zuo
et al. (2016) studied the influence of factors on the emis-
sion efficiency of micro combustors using orthogonal de-
sign and fuzzy grey correlation analysis and found H2/air
equivalence ratio to be the most important factor. Shinde
and Pawar (2017) presented a multi-objective optimization
method that uses orthogonal design-based GRA to improve
journal bearing performance and found an optimal set of
surface texturing parameters. Kadier et al. (2015) proposed
a new method of evaluating the performance of cathodes
in a microbial electrolysis cell through GRA and discovered
through a microbial electrolysis cell test that stainless steel
A286 had the best efficiency.
This work aims to introduce a method of evaluating the
effect of calcination parameter factors on the temperature
distribution in shaft calciners. A 2D transient mathematical
model was developed to describe the mass and heat transfer
problems occurring during the calcination of petroleum
coke in a shaft calciner. In this model, a three-parallel-dis-
tributed activation energy model (DAEM)-reaction model
was used to describe the pyrolysis kinetics of petroleum
coke, and a dichotomy method was applied to restrict the
flue wall temperature of Layer 2 (T2), thereby meeting the
operational control requirements in actual production. In
addition, a statistical method called orthogonal design-
based GRA was used to quantize the influence degree of
four factors, namely, DRPP, VC, MC, and VDR, on the
temperature distribution in the pot of a shaft calciner and
determine the primary and secondary relationships of the
Fig. 1 Model prediction of TG and dα(T)/dT curves compared with
influence degrees of the four factors.
experimental data

1. Experiment and Methods

It is assumed that the petroleum coke contains three pseudo
1.1 Thermogravimetric analysis experiment components that react independently of each other. There-
Fushun petroleum coke from Fushun, China was selected fore, the overall pyrolysis behavior can be considered as the
as the experimental sample. The pyrolysis analysis of Fushun weighting of the individual pyrolysis reactions of the three
petroleum coke was carried out in a thermogravimetric ana- pseudo components. The conversion rate can be expressed
lyzer (STA-449F3 Jupiter, Netzch, Germany). In the experi- as Eq.(1).
ment, about 25 mg of sample was placed on the crucible of ∞ t E
the thermal analyzer. Non-isothermal runs were carried out α(T) = � − � exp �−k i � exp �− � dt� f (E) dE (1)
� � RT
from 25 to 1400°C at heating rate of 10°C/min. The high-
purity N2 (99.999%) with a flow rate of 100 mL/min was The three-parallel-DAEM-reaction model can be de-
used as the carrier gas to provide an inert atmosphere. scribed by Eqs. (2)–(4).

dα(T)
1.2 Kinetic analysis using three-parallel-DAEM-reaction =
model dT
� ∞ ki E ki
In this work, the three-parallel-DAEM-reaction model
� Ci � exp �− − ψ (E, T)� f i (E) dE (2)
was employed for kinetics analysis of the complex reactions i=� � η RT η
in the pyrolysis process of petroleum coke (Cai et al., 2013).

812 Journal of Chemical Engineering of Japan

 the coke bed and is introduced into the flues (Layers 1 and
2) through the volatile channels. Then, the volatile matter
is mixed with the air and burnt directly in the flue, and the
heat from the combustion is subsequently administered to
the petroleum coke indirectly from the flue walls through
convection and radiation. During calcination, thecombus-
tion of volatile matter in the flues is the main source of heat
in the shaft calciner. Each pot is symmetrically heated by
heating the flue walls on both sides, and the temperature
distribution in each flue of the same layer (z-direction)
can be considered homogeneous. Thus, the heat loss in the
z-direction can be ignored, and the shaft calciner can be
considered as a 2D model. Figure 2(c) shows the model do-
main. In this model, a multiple-block-structured tetrahedral
mesh is used for meshing the furnace wall and the porous
medium zone in the pot. The elemental composition and
Fig. 2 Detailed structure of shaft calciner (a) 3D diagram, (b) sketch physical parameters of the petroleum coke and bricks used
map of material movement, and (c) 2D diagram of model do- in the model are discussed in the references (Xiao et al.,
main 2016a; Li et al., 2018a).
T E
ψ (E, T) = � exp �− � dT (3) 2.2 Mathematical Model
� RT 2.2.1 Model assumptions The 2D transient mathemat-
ical model is based on the following assumptions.
−(E − E �,i )
�
�
f i (E) = √ exp � � (4) 1. The petroleum coke bed is considered an isotropic po-
σ i �π �σ i � rous medium, thereby satisfying the required continu-
ous medium condition (Hu et al., 2001).
Based on the thermogravimetric (TG) analysis data, the 2. The discharging process is regarded as a plug flow,
Lemberg–Marquardt algorithms nonlinear least squares fit- and the force of the gas to the solid phase is neglected
ting method is used to calculate the unknown kinetic pa- (Niven, 2002; Lin et al., 2009).
rameters (E0,i, ki and σi) in the three-parallel-DAEM-reac- 3. The pyrolysis reaction is described by the three-paral-
tion model. The calculation equation is defined as follows. lel-DAEM-reaction model, and the surface and inter-
nd nal temperatures of the particles are homogeneous (Hu
F = � [α(T)exp,i − α(T)cal,i ]
�
(5) et al., 2001; Xiao et al., 2016a).
i=�
4. A local temperature difference exists between the solid
The pyrolysis kinetic parameters of Fushun petroleum particle and the gas flow in the porous medium region
coke are shown in Table 1. Figure 1 compares the predicted of the coke bed in the pot (Lin et al., 2009).
TG and reaction rate (dα(T)/dT) curves by the three-paral- 5. The flue wall temperature of each layer is considered ho-
lel-DAEM-reaction model with experimental data. It can be mogeneous (Hu et al., 2001; Xiao et al., 2016a).
seen that the predicted TG and dα(T)/dT curves agree well 2.2.2 Governing equations In this model, the calcula-
with the experimental data, which indicates that the three- tion domain is divided into two parts, namely, the porous
parallel-DAEM-reaction model can describe the pyrolysis region in the pot and the heterogeneous solid phase region
behavior of Fushun petroleum coke well. of the furnace wall. The furnace wall comprises refractory
bricks of silica and refractory clay. In the porous region, the
discharge behavior is described by a plug flow and the coke
2. Numerical Methods
pyrolysis is described by the mass, species, and energy con-
2.1 Structure and operational principle of shaft calciner servation equation of the solid petroleum coke. The prob-
Figure 2(a) shows that a typical shaft calciner comprises lems of porous medium seepage about the heat transfer
24 vertical refractory pots; each group, which has four pots, process of the volatile matter in the coke bed are solved by
is surrounded by eight layers of horizontally oriented heat- the energy conservation equations of the gas phase. In the
ing flues, thus forming a sandwich structure. Each pot is a solid phase region, the heat transfer process is described by
simple block with 1.66 m length, 0.36 m width, and 6.627 m the energy conservation equation.
height. The heating flue has a zigzag structure with 3.996 m The governing equations used in this model are described
length, 0.215 m width, and 0.479 m height and a flue dia- by the following general formula. Table 2 shows the specific
phragm with 0.065 m thickness. In Figure 2(b), GPC is fed parameters for each equation.
continuously into the hoppers, moves down through the
∂ (αρϕ k ) ∂ ∂ϕ k
pots, and is discharged through the cooling water jacket. The + �αρu i ϕ k − Γk � = Sϕ (6)
∂t ∂x i ∂x i
volatile gas released from the green coke travels up through

Vol. 52  No. 11  2019 813

 ⎧
⎪
⎪
Table 2 Governing equations Nc
⎪
⎪ � m i − Cvol mdry_gpc = �
⎪
⎪
⎪
Conservation α ϕ Γ Sϕ ⎪
⎪
⎪
i=�
⎪
⎪
⎪
Ne
⎪ � mi Ai j − B j = �
s.t. ⎨
Solid phase (porous region)
⎪
n
− ∑ Ri (13)
⎪
⎪
Mass 1−εs 1 0 i=�
⎪
⎪
⎪
Nc
⎪ � m i ∆H i − ∆H vol C vol mdry_gpc = �
i=1
Species 1−εs Cs,i 0 −Ri ⎪
⎪
⎪
⎪
⎪
⎪
⎪
i=�
mi ≥ �
Energy 1−εs cpTs λs,eff/cp Sh−qrgs−qcgs
⎩
Gas phase n
Mass εs 1 0 ∑ Ri The Chaniwalah–Parikh correlation (Jayakumar, 2008) is
i=1
Energy εs cp Tg λg,eff/cp qrgs+qcgs used to calculate the high heating value of the GPC.
Where qcgs =hcgs Sv (Ts−Tg), qrgs =εsg σSv (T s4−Tg4) HHV = (��.��C + ���.��H − ��.��O
(14)
For the solid phase (brick wall zone) −�.��N + ��.��S − �.��Ash) × ���
Energy 1 cp Tw λw/cp 0
Herein, C, H, O, N, S, and Ash are the mass percentages of
the elements and ash in the GPC, respectively.
The descending motion of petroleum coke is assumed as a The heating value of the combustion of volatile matter is
plug flow and described by Eq. (7). the difference between the high heating value of the GPC
n and CPC.
∂ ∂
((1 − ε s ) ρ s ) + ((1 − ε s ) ρ s u i ) = − � R i (7)
∂t ∂x i HHVdry_gpc − HHVcpc × (� − Cvol )
∆H vol =
i=1
(15)
The Dirchlet boundary condition is used for the velocity- Cvol
inlet: The gas evolution amount of the petroleum coke pyrolysis
qm process is described by the above three-parallel-DAEM-re-
u x � = �, u y� =− (8)
(�−ε s ) ρ s A c × ���� action model, and the mass source term for the gas products
per unit volume is expressed by Eq. (16).
2.2.3 Petroleum coke pyrolysis model In accordance n
dα(T)i
with previous research (Merrick, 1983; Lin et al., 2009; Ri = ρ � Ci i = �, �, � (16)
dt
Wang et al., 2010), the composition of the pyrolysis volatile i=�

matter in the petroleum coke is assumed as CH4, C2H6, H2, The heat absorption source term caused by petroleum
CO, CO2, H2O, O2, NH3, H2S, SO2 and N2. coke pyrolysis is given by Eq. (17).
In order to accurately determine the content of volatile
matter, seven conservation equations for volatile matter py- Sh = − � hi Ri (17)
i
rolysis have been established according to the mass, element
and energy conservation. 2.2.4 Heat exchange model The gas products of the
Mass conservation: volatiles combustion reaction are calculated by using the
Nc equilibrium constant method (Li et al., 2018b). The main
� m i − Cvol mdry_gpc = � (9) reaction equations are expressed as Eqs. (18) to (24).
i=�

Element conservation: CH � + �O � = CO � + �H � O − ∆H C H �
[CO� ][H� O]
�
(18)
K� =
Ne
� mi Ai j − B j = � j = C, H, O, N , S (10) [CH� ][O� ]
�
i=�

Energy conservation: C � H � + �.�O � = �CO � + �H � O − ∆H C � H �

[CO� ] [H� O]
Nc � �
(19)
� m i ∆H i − ∆H vol C vol mdry_gpc = � (11) K� =
[C� H� ][O� ]
�.�
i=�

However, the seven given conservation equations are im-

possible to solve with eleven unknown variables. To solve H � + �.�O � = H � O − ∆H H � O
[H� O] (20)
K� =
this problem, a simplex method of linear programming is
[H� ][O� ]
�.�
introduced to establish an optimal model. The objective
function and constraints are expressed as Eq. (12).
Nc CO + �.�O � = CO � − ∆H C O
min z = � m i (12) [CO� ] (21)
K� =
[CO][O� ]
i=� �.�

subject to:

814 Journal of Chemical Engineering of Japan

 �N H � + �.�O � = �H � O + N � − ∆H N H � tween the flue gas and furnace wall and other heat losses
caused by air leakage in the flue of each layer.
[N� ][H� O]
�
K� = (22) The heat produced by the volatile matter combustion and
[NH� ] [O� ]
� �.�
the flue wall temperature of each layer are evaluated by the
mass and energy conservation equation, as shown in Eq.
H � S + �.�O � = H � O + SO � − ∆H H � S
(29). The flue wall temperature of each layer is iteratively
[SO� ][H� O] (23)
K� = calculatedby a dichotomy method.
[H� S][O� ]
�.�
nr nr Tk , i n

∆G � m r ∆H ���,r + � � m r c p dT =
K j = exp �− � (24) i i ���
RT np np Tk ,out
� m p ∆H ���, p + � � m p c p dT + Q l oss ∆t
For a multicomponent gas phase system, the independent j j ���
reaction can be represented by a matrix as follows.
k = �...� (29)
(Vj,i )(I i ,k ) = (D j,k ) (25)
Q l oss = Q r g w + Q c g w + Q e w + Qother (30)
The equations of the equilibrium constant method are ex-
pressed by Eqs. (26)–(28), respectively. The Newton–Raph- Herein, T0,in is the inlet temperature of the flue of Layer 1.
son method is used for solving simultaneous equations. Other heat losses (Qother) include nameless heat loss caused
by air leakage through observation holes and other outlets.
X i V ji
Yj = Z j K j � � � j = �, �, ⋯, N c − N e (26) The following equations are used to describe the radiative
i Zi and convective heat transfers between the gas and flue wall
(Meisingset and Balchen, 1995; Eskelinen et al., 2015; Xiao
Pk = � I i ,k X i + � D j,k Y j k = �, �, ⋯, N e (27)
i j
et al., 2016b).

Z = � X i + � Yj (ε w + �)
(28) Qrgw = σ A f c ε g �Tg� − Tw� � (31)
i j �
Therefore, the chemical equilibrium conservation equa- Qcgw = hcgw A f c (Tg − Tw ) (32)
tions are used to calculate the gas composition in the flue
of each layer. Meanwhile, the heat flux boundary condition The convective heat transfer coefficient between the gas
(qw) is applied to describe the heat transfer process between and the flue wall is calculated by the Dittus–Boelter equa-
the flue wall and gas. In this model, the heat loss is mainly tion (Moradikazerouni et al., 2019) as follows.
caused by the heat radiation and thermal convection be-

Fig. 3 Computational flow chart for simulations

Vol. 52  No. 11  2019 815

 �.�
λ d f eug ρg cp µ n discretization for the computation domain in the Cartesian
hcgw = �.��� � � � � (33)
coordinate system. The quadratic upstream interpolation
df e µ λ
for the convective kinematics (QUICK) scheme of the finite
The convective and radiation heat transfer equation for volume method is used to construct the discrete equations.
both exteriors of the flue wall is as follows. Furthermore, the first-order hybrid scheme is introduced
Qew = hew Aew (Tew − T f ) + εew σ Aew �T∞ − Tew �
� � to accomplish discrete processing of boundary nodes. The
(34)
governing equations are solved using an efficient sparse ma-
The boundary condition for heat flux in the flue wall is trix library, namely, UMFPACK (Davis, 2004). The compu-
defined as follows. tational flow chart for the simulations is shown in Figure 3.
∂T Qloss
−λ � = qw = (35) 2.3 Evaluation method
∂x i w Afw
2.3.1 Orthogonal design A previous study suggested
2.2.5 Restriction algorithm for T2 In accordance with that the temperature distribution in shaft calciners is con-
actual operating conditions, T2 is set to be approximately siderably dependent on calcination parameters (Xiao et al.,
1350°C and the flue wall temperature of the other layers is 2016a), such as DRPP, EAC, VC, MC, and VDR. In this new
allowed to develop naturally. In this model, the air excess 2D transient model, the flue wall temperature of Layer 2
coefficient (β) is dynamically adjusted to induce T2 to reach (T2) is constrained in accordance with the actual production
the target temperature. A dichotomy method is used to per- requirements; consequently, the EAC dynamically changes
form coupling iteration calculations by updating β, which is
determined by Eq. (36).
Table 3 Orthogonal design of factors and levels
f (x) = Tx − Tc (36)
Levels
2.2.6 Coupling model calculation flow The 2D tran- Factors
1 2 3 4 5
sient model is implemented using a C# language program
and a group of partial differential equations for describing DRPP [kg/h] 75 85 95 105 115
the mass and heat transfer process in terms of such pro- MC [wt%] 6 8 10 12 14
VC [wt%] 7 9 11 13 15
cesses as discharge movement and petroleum coke pyrolysis.
VDR [wt%] 30 40 50 60 70
The cell-centered scheme is extended to implement region

Table 4 Orthogonal design (L25 (54)), normalized values, and GRGs

DRPP MC VC VDR T8 Tcp Normalized GRC

Case GRG Rank
[kg/h] [wt%] [wt%] [wt%] [K] [K] T8 Tcp T8 Tcp

1 75 6 7 30 1274 1258 0.000 0.373 0.333 0.444 0.388 23

2 75 8 9 40 1335 1319 0.445 0.689 0.474 0.617 0.545 14
3 75 10 11 50 1364 1344 0.657 0.823 0.593 0.738 0.666 9
4 75 12 13 60 1386 1363 0.818 0.923 0.733 0.866 0.799 4
5 75 14 15 70 1404 1378 0.949 1.000 0.907 1.000 0.954 1
6 85 6 9 50 1325 1296 0.372 0.571 0.443 0.538 0.491 17
7 85 8 11 60 1358 1324 0.613 0.718 0.564 0.639 0.602 11
8 85 10 13 70 1380 1341 0.774 0.809 0.688 0.723 0.706 7
9 85 12 15 30 1407 1352 0.971 0.861 0.945 0.783 0.864 2
10 85 14 7 40 1308 1261 0.248 0.391 0.399 0.451 0.425 20
11 95 6 11 70 1354 1305 0.584 0.616 0.546 0.566 0.556 13
12 95 8 13 30 1386 1315 0.818 0.668 0.733 0.601 0.667 8
13 95 10 15 40 1407 1328 0.971 0.739 0.945 0.657 0.801 3
14 95 12 7 50 1305 1241 0.226 0.285 0.393 0.411 0.402 22
15 95 14 9 60 1336 1266 0.453 0.413 0.477 0.460 0.469 18
16 105 6 13 40 1387 1294 0.825 0.559 0.741 0.531 0.636 10
17 105 8 15 50 1408 1307 0.978 0.628 0.958 0.573 0.766 5
18 105 10 7 60 1299 1214 0.182 0.146 0.380 0.369 0.374 24
19 105 12 9 70 1332 1240 0.423 0.277 0.464 0.409 0.437 19
20 105 14 11 30 1371 1248 0.708 0.320 0.631 0.424 0.528 15
21 115 6 15 60 1411 1288 1.000 0.529 1.000 0.515 0.758 6
22 115 8 7 70 1292 1186 0.131 0.000 0.365 0.333 0.349 25
23 115 10 9 30 1341 1200 0.489 0.072 0.495 0.350 0.422 21
24 115 12 11 40 1369 1218 0.693 0.164 0.620 0.374 0.497 16
25 115 14 13 50 1391 1230 0.854 0.228 0.774 0.393 0.584 12

816 Journal of Chemical Engineering of Japan

accordingly. Therefore, the investigated factors are DRPP, better” principle is used to evaluate the levels in the GRA by
MC, VC, and VDR, and all factors has five levels, as given the following equation (Shinde and Pawar, 2017).
in Table 3. In this work, an appropriate orthogonal design
x i� (k) − min x i� (k)
table L25 (54) is used as shown in Table 4. x i (k) = , i = �-m, k = �-m (37)
max x i� (k) − min x i� (k)
2.3.2 GRA This work aims to enhance the calcination
temperature in the shaft calciner while maximizing the qual- Step 2: Calculation of the grey relational coefficient
ity of CPC. GRA is an effective approach and uses the grey (GRC) (Kadier et al., 2015).
relational grade (GRG) to measure the correlation degree
between the reference and comparison matrixes (Zuo et al., ξ (x � (k), x i (k)) =
2016). This method is used to evaluate the influence of the min min �x � (k) − x j (k)� + ζ max max �x � (k) − x j (k)�
j k j k
abovementioned calcination parameters on the temperature ,
�x � (k) − x i (k)� + ζ max max �x � (k) − x j (k)�
distribution in the shaft calciner. The calculation steps for j k
the GRA are as follows (Taşkesen and Kütükde, 2014). i = �-m, j = �-m, k = �-n (38)
Step 1: Normalization of the original data series.
Factors and variables typically have different physical Herein, x0(k) is a comparison sequence, which generally has
meanings and data dimensions (Zuo et al., 2016). To re- a value of 1.0, and ζ is the distinguishing coefficient. The
duce the identification analysis error, the original data must value of ζ ranges from 0 to 1. In this work, the value is set
be nondimensionalized before calculation. The “larger-the- as 0.5 in accordance with the literature (Kadier et al., 2015).

Fig. 4 Grid independence study (a) 20 mm, (b) 15 mm, and (c) 5 mm Fig. 5 Comparison of simulation and experimental results

Fig. 6 Simulation results of temperature contours in shaft calciner for 25 cases

Vol. 52  No. 11  2019 817

 Step 3: Calculation of the GRG by the following equation. ceeding test cases.
n 3.1.2 Model verification Figure 5 compares the pres-
�
γ �i = � ξ (x � (k), x i (k)) (39) ent simulation and measurement results of temperatures of
n k=�
the flue wall and the center of the pot, where the flue wall
temperature is measured using infrared radiation thermom-
eters (Raytek 3I1ML3U), and the temperature of center pot
3. Results and Discussion
is from the existing literature data (Liu, 2016). The operation
3.1 Grid independence and model verification conditions are that the DRPP, MC, VC, and VDR are kept
3.1.1 Grid independence Three various grids measur- at 85 kg/h and 10, 12.5, and 70 wt%, respectively. As can be
ing 20, 15, and 5 mm, which have 9,715, 16,644 and 130,806 seen in Figure 5, the simulated results of flue wall tempera-
cells, respectively, are used for grid independence verifica- ture are very consistent with the measurement results, and
tion, as presented in Figure 4. The temperatures in the the error is less than 71 K. Meanwhile, the simulated results
center of the pot (x=0.0 m) are compared while DRPP, MC, of the temperature of center pot are basically the same as
VC, and VDR are kept at 85 kg/h and 10, 12.5, and 70 wt%, those measured by Liu (2016). The simulated temperature of
respectively. The temperatures differ by approximately 16 K center pot increases slowly in the L1–L2 regions firstly, then
between the 15 and 5 mm grids and by approximately 74 K rises rapidly from 400 to 1200 K in the L3–L6 regions, and
between the 15 and 20 mm grids. Therefore, the 15 mm grid, increases gently in the L7–L8 regions lastly. A relatively large
whichhas 16,644 cells, is applicable and effective in the suc- difference (about 210 K) exists, but only near the L5 region.
whereas, the differences in the other regions is less than
61 K, and the relative error is less than 13%.
The main sources of error are measurement error and the
model assumptions. However, the error is acceptable, and
the model is an effective tool for investigating the effects
of the factors on the temperature distribution in the shaft
calciner.

3.2 Results of orthogonal design

The simulation results of the temperature field obtained
for 25 cases are shown in Figure 6, where L1–L8 indicate the
corresponding regions of the different flue layers. This fig-
ure shows that DRPP, MC, VC, and VDR directly influence
the temperature field in the shaft calciner, and all cases can
reach maximum calcination temperature at the neighboring

Fig. 7 Maximum temperature of Tcp and T8 for 25 cases Fig. 8 Effects of each factor on GRG

Table 5 Main effects of factors on GRG

Levels
Factors ΔMax-Min Rank
1 2 3 4 5

DRPP 0.671 0.617 0.579 0.548 0.522 0.149 2

MC 0.566 0.586 0.594 0.600 0.592 0.034 3
VC 0.388 0.473 0.570 0.678 0.828 0.440 1
VDR 0.574 0.581 0.582 0.600 0.600 0.027 4

818 Journal of Chemical Engineering of Japan

 Table 6 ANOVA results for GRG 12, 15, and 60 wt%, respectively. Further, the temperature
Factors SeqSS DF AdjMS F-value Contribution [%]
of Tcp has been increased by about 2 K compared to that of
case 5. The difference between the maximum and minimum
DRPP 0.068 4 0.017 16.735 10.092 GRG values of each factor is used to determine the influence
MC 0.003 4 0.001 0.838 0.505
degrees of the factors on the temperature distribution in the
VC 0.599 4 0.150 146.518 88.359
shaft calciner. Table 5 shows thatVC has the most noticeable
VDR 0.003 4 0.001 0.731 0.441
effect on calcination temperature, followed by DRPP, MC,
Error 0.008 8 0.001 — 0.603
Total 0.682 24 — — 100.0 and VDR.
To identify the factor that most considerably affects the
GRG, analysis of variance (ANOVA) is adopted (Taşkesen
and Kütükde, 2014). Table 6 and Figure 9 present the
ANOVA results. VC has a contribution rate of 88.359%,
And thus has the largest influence on the GRG; the other
important factors are DRPP and MC, which have contribu-
tion rates of 10.092% and 0.505%, respectively. VDR has the
smallest effect on the GRG, having a contribution rate of
merely 0.441%. Therefore, VC is the most influential factor,
followed by DRPP and MC; VDR is the least influential fac-
tor behind calcination temperature.

Conclusions
In this work, a 2D transient mathematical model was
developed to describe the mass and heat transfer problems
Fig. 9 Effects of factors on GRG according to ANOVA results that occur during the calcination process of petroleum coke
in shaft calciners. In this model, a three-parallel-DAEM-
region of L8. reaction model was used to describe the pyrolysis kinetics
Through various investigations, the maximum tempera- of petroleum coke, and a dichotomy method was applied to
ture in the center of the pot (Tcp) and the flue wall tempera- realize the restriction of the flue wall temperature of Layer 2.
ture of Layer 8 (T8) are determined as shown in Figure 7. In addition, a statistical method called orthogonal design-
The maximum calcination temperature in the center of based GRA was used to quantize the influence degrees of
the pot for Case 5 is more than 1373 K and those for the four factors, namely, DRPP, VC, MC, and VDR, on the
other cases are at least approximately 1173 K. Moreover, T8 temperature distribution of the pot in a shaft calciner and
is greater than Tcp for all cases, and their temperature dif- determine the primary and secondary relationships of the
ference ranges from 16 to 161 K. Therefore, the degrees of influence degrees of the four factors.
influence of these four factors on the temperature distribu- The main conclusions are as follows:
tion must be quantized and the optimal combination of cal- 1. The pyrolysis kinetics of Fushun petroleum coke
cination parameters for achieving the maximum calcination can be successfully described by the three-parallel-
temperature must be determined. This problem is solved by DAEM-reaction model. The ranges of E0i and σi were
using the GRA method. 84.16–192.89 kJ/mol and 10.23–18.42 kJ/mol, respective-
ly. The ki was 7.62×103 s−1.
3.3 Results of GRA 2. The ANOVA of GRG showed that VC was the most
The Tcp and T8 under various conditions are regarded as influential factor, followed by DRPP and MC; VDR was
evaluating indicators. First, the original data series is non- the least influential factor behind calcination tempera-
dimensionalized by Eq. (37). Then, GRC and GRG are cal- ture. Therefore, in the condition of fixing the flue wall
culated by Eqs. (38) and (39) respectively. Table 4 tabulates temperature of Layer 2, VC is the key factor affecting the
the values of the orthogonal design, normalized values, and calcination temperature and must be strictly controlled
GRG. Case 5 has the maximum GRG value, which means during production.
that it has the optimal combination of parameters among 25
cases. Acknowledgement
The main effects of the factors on GRG are investigated by
This work was funded by the Science and Technology Program
range analysis (Taşkesen and Kütükde, 2014). Table 5 and of Education Department of Jiangxi Province in China (GJJ170527,
Figure 8 provide the maximum GRG and effects of each GJJ180449), Natural Science Foundation of Jiangxi Province of China
factor, respectively. By maximizing the values of GRG, the (20192BAB206019).
maximum values of Tcp and T8 can be obtained at DRPP1,
MC4, VC5, and VDR4 optimal conditions; in the optimal
combination, DRPP, MC, VC, and VDR are at 75 kg/h and

Vol. 52  No. 11  2019 819

 U = velocity vector [m/s]
Nomenclature
Vj,i = stoichiometric coefficients [—]
Ac = horizontal cross section area of pot [m2]
Aew = area of of external wall [m2] α = void fraction [—]
Afc = cross sectional area of each layer of flue wall [m2] α(T) = conversion rate [—]
Afw = area of each layer of flue wall [m2] β = excess air ratio [—]
Aij = mass fraction of species i in element j [—] dα(T)/dT = reaction rate [1/s]
Bj = mass of element j [kg] η = Heating rate [K/s]
cp = specific heat [J/(kg·K)] Γ = diffusion coefficient [kg/(m·s)]
Ci = mass fraction of species i [—] λ = thermal conductivity [W/(m·K)]
dfe = equivalent diameter of flue [m] μ = viscosity coefficient [kg/(m·s)]
Dj,k = molecular formula matrix for dependent species [—] ϕ = generalized variable in mass equation, [—]
e = number of element [—] or generalized variable in energy equation [J/kg]
E = activation energy [kJ/mol] ρ = density [kg/m3]
f (E) = distribution of activation energy [mol/kJ] σ = Stefan–Boltzmann constant [W/(m2 ·K4)]
F = residual sum of squares [—] σi = standard deviation [kJ/mol]
ΔG = Standard Gibbs free energy changes of the species [kJ/mol] ψ(E,T) = integral of the Boltzmann factor [K]
ΔHi = chemical reaction heat of component i [J/kg] εs = bed porosity [—]
ΔHvol = total heating value of combustion of volatiles [J/kg] εg = emissivity of gas [—]
HHV = high heating value [J/kg] εsg = emissivity of grain [—]
hew = heat transfer coefficient of external wall [W/(m2 ·K)] εew = emissivity of exteriors of the flue wall [—]
hcgw = convection heat transfer coefficient between gas εw = emissivity of flue wall [—]
and flue wall [W/(m2 ·K)] ζ = distinguishing coefficient [—]
hcgs = convection heat transfer coefficient between gas γ = grey relational grade [—]
and solid [W/(m2 ·K)] ξ = grey relational coefficient [—]
hi = standard enthalpy of pyrolysis of component i [J/kg]
Ii,k = molecular formula matrix for independent species [—] ‹Subscripts›
ki = pre-exponential factor [1/s] cpc = calcined petroleum coke
Kj = equilibrium constant [—] cgs = convective, gas–solid
m = mass [kg] cgw = convective, gas-wall
n = number of characteristic parameters [—] cal = calculated data
nd = number of data points [—] dry-gpc = green petroleum coke(dry basis)
Nc = number of species [—] exp = experimental data
Ne = number of element [—] ew = external wall
Pk = molar amount of the kth element in the system [mol] eff = effective
qm = mass flow of green petroleum coke [kg/h] f = flue
qw = wall heat flux density [W/m2] fw = flue wall
q = energy source term [W/m3] g = gas phase
Q = heat flow [W] gpc = green petroleum coke
R = gas constant [kJ/(mol·K)] i, j, k = index of species
Ri = chemical reaction mass source term of in = inlet
component i [kg/(m3 ·s)] mois = moisture
Sϕ = energy source term [W/m3], or mass source term n, m = gas/solid species count
[kg/(m3 ·s)] nA = number of reactants
Sh = pyrolysis energy source term of solid phase volatile [W/m3] nB = number of products
Sv = specific surface area [1/m] out = outlet
Tc = target temperature of the flue wall of Layer 2 [K] p = product
T∞ = temperature of environment [K] r = raw material
Tk = temperature of the flue of layer i [K] rgs = radiative, gas–solid
Tx = current flue wall temperature of Layer 2 [K] rgw = radiative, gas-wall
T = temperature [K] s = solid phase
t = time [s] sg = solid–gas
xi0(k) = original data sequence [—] vol = volatiles
x0(k) = comparison sequence [—] w = brick wall zone
xi(k) = normalization data sequence [—] x, y, z = direction
Xi = mole number of the ith independent species [mol]
Yi = mole number of the jth dependent species [mol]
Literature Cited
z = objective function [kg]
Z = total molar amount of the gas phase [mol] Cai, J. M., W. X. Wu and R. H. Liu; “Sensitivity Analysis of Three-Par-
ui = velocity in i-direction(x, y) [m/s] allel-DAEM-Reaction Model for Describing Rice Straw Pyrolysis,”
ux0 = initial velocity of x-direction inlet [m/s] Bioresour. Technol., 132, 423–426 (2013)
uy0 = initial velocity of y-direction inlet [m/s] Davis, T. A.; “Algorithm 832: UMFPACK V4.3—An Unsymmetric-Pat-

820 Journal of Chemical Engineering of Japan

 tern Multifrontal Method,” ACM Trans. Math. Softw., 30, 196–199 Coal: 1. The Evolution of Volatile Matter,” Fuel, 62, 534–539 (1983)
(2004) Moradikazerouni, A., M. Afrand, J. Alsarraf, O. Mahian, S. Wongwises
Edwards, L.; “The History and Future Challenges of Calcined Petro- and M.-D. Tran; “Comparison of the Effect of Five Different En-
leum Coke Production and Use in Aluminum Smelting,” JOM, 67, trance Channel Shapes of a Micro-Channel Heat Sink in Forced
308–321 (2015) Convection With Application to Cooling a Supercomputer Cir-
Eskelinen, A., A. Zakharov, S.-L. Jämsä-Jounela and J. Hearle; “Dynam- cuitBoard,” Appl. Therm. Eng., 150, 1078–1089 (2019)
ic Modeling of a Multiple Hearth Furnace for Kaolin Calcination,” Niven, R. K.; “Physical Insight into the Ergun and Wen & Yu Equations
AIChE J., 61, 3683–3698 (2015) for Fluid Flow in Packed and Fluidised Beds,” Chem. Eng. Sci., 57,
Hu, G. X., W. Xu and H. J. Fan; “Gas Flow and Heat Transfer in Moving 527–534 (2002)
Particulate Bed,” J Chem. Ind. Eng. (China), 52, 401–405 (2001) Shinde, A. B. and P. M. Pawar; “Multi-Objective Optimization of Sur-
Jayakumar, R.; Analysis of Power Generation Processes Using Petcoke, face Textured Journal Bearing by Taguchi Based Grey Relational
Ph.D. Thesis, Texas A&M University, 82–83 (2008) Analysis,” Tribol. Int., 114, 349–357 (2017)
Kadier, A., P. Abdeshahian, Y. Simayi, M. Ismail, A. A. Hamid and M. Taşkesen, A. and K. Kütükde; “Experimental Investigation and Multi-
S. Kalil; “Grey Relational Analysis for Comparative Assessment of Objective Analysis on Drilling of Boron Carbide Reinforced Metal
Different Cathode Materials in Microbial Electrolysis Cells,” En- Matrix Composites Using Grey Relational Analysis,” Measurement,
ergy, 90, 1556–1562 (2015) 47, 321–330 (2014)
Kuo, Y., T. Yang and G. W. Huang; “The Use of Grey Relational Analysis Wang, C. H., W. Z. Chen, F. R. Jia, Y. H. Tian and Y. Liu; “Numerical
in Solving Multiple Attribute Decision-Making Problems,” Com- Simulation of Heat Transfer and Combustion in Carbon Rotary
put. Ind. Eng., 55, 80–93 (2008) Kiln,” CIESC J., 61, 1379–1384 (2010)
Lee, W. S. and Y. C. Lin; “Evaluating and Ranking Energy Performance Wang, Z. W., T. Z. Lei, X. Chang, X. G. Shi, J. Xiao, Z. F. Li, X. F. He, J.
of Office Buildings Using Grey Relational Analysis,” Energy, 36, L. Zhu and S. H. Yang; “Optimization of a Biomass Briquette Fuel
2551–2556 (2011) System Based on Grey Relational Analysis and Analytic Hierarchy
Li, J., J. D. Huang, J. Xiao and M. Z. Li; “Simulation of Migration of Process: A Study Using Cornstalks in China,” Appl. Energy, 157,
High Temperature Zone of Petroleum Coke Layer in Vertical Shaft 523–532 (2015)
Calciner,” Chin. J. Nonferrous Met., 28, 1216–1224 (2018a) Xiao, J., J. D. Huang, Q. F. Zhong, F. C. Li, H. L. Zhang and J. Li; “A
Li, M. Z., J. M. Zhou, C. R. Tong, W. H. Zhang, Z. Chen and J. L. Wang; Real-Time Mathematical Model for the Two-Dimensional Tem-
“Thermodynamic Modeling and Optimization of the Copper Flash perature Field of Petroleum Coke Calcination in Vertical Shaft
Converting Process Using the Equilibrium Constant Method,” Calciner,” JOM, 68, 2149–2159 (2016a)
Metall. Mater. Trans., B, Process Metall. Mater. Proc. Sci., 49, Xiao, J., J. D. Huang, Q. F. Zhong, H. L. Zhang and J. Li; “Modeling and
1794–1807 (2018b) Simulation of Petroleum Coke Calcination in Pot Calciner Using
Lin, P. Y., J. J. Ji, Y. H. Luo and Y. Wang; “A Non-Isothermal Integrated Two-Fluid Model,” JOM, 68, 643–655 (2016b)
Model of Coal-Fired Traveling Grate Boilers,” Appl. Therm. Eng., Zhong, Q. F., Q. Y. Mao, L. Y. Zhang, J. H. Xiang, J. Xiao and J. P.
29, 3224–3234 (2009) Mathews; “Structural Features of Qingdao Petroleum Coke from
Liu, Q. Y.; “Discussion on Temperature Distribution of Cocurrent Pot- HRTEM Lattice Fringes: Distributions of Length, Orientation,
Type Calciner and Correlative Problem,” Carbon Tech. (China), 35, Stacking, Curvature, and a Large-Scale Image-Guided 3D Atomis-
64–65 (2016) tic Representation,” Carbon, 129, 790–802 (2018)
Martins, M. A., L. S. Oliveira and A. S. Franca; “Modeling and Simula- Zhou, S. H. and L. J. Luo; “Numerical Simulation and Optimization
tion of Petroleum Coke Calcination in Rotary Kilns,” Fuel, 80, Research of the Shaft Kiln,” Carbon Tech. (China), 2, 52–54 (2014)
1611–1622 (2001) Zuo, W., J. E, X. Liu, Q. Peng, Y. Deng and H. Zhu; “Orthogonal Ex-
Meisingset, H. C. and J. G. Balchen; “Mathematical Modeling of a Ro- perimental Design and Fuzzy Grey Relational Analysis for Emitter
tary Hearth Coke Calciner,” Model. Identif. Control, 16, 193–212 Efficiency of the Micro-Cylindrical Combustor with a Step,” Appl.
(1995) Therm. Eng., 103, 945–951 (2016)
Merrick, D.; “Mathematical Models of the Thermal Decomposition of

Vol. 52  No. 11  2019 821