Professional Documents
Culture Documents
G R A P H I C A L A B S T R A C T
A R T I C L E I N F O A B S T R A C T
Keywords: Traditionally, the shell and tube heat exchanger (STHE) design minimizes the total annual cost
Shell and tube heat exchanger subject to allowable pressure drop. However, it is often insufficient when considering equipment
Flexibility index limitations and uncertain disturbance factors. Design inefficiency or improper operation of
Genetic algorithm overall process systems could potentially occur. A rigorous design strategy of STHE based on the
Simultaneous optimization flexibility index analysis method is proposed to address the above issues. The genetic algorithm
with rigorous constraints is considered for the STHE design optimization. Simultaneously, the
flexibility index (FI) is incorporated to quantify the uncertain factors of the STHE operation. The
STHE load capacity and the corresponding uncertain disturbance factors affecting the STHE
operability are evaluated. In addition, design workflows incorporating flexibility index are also
discussed. Simultaneous consideration of the flexibility index in the optimization procedure can
improve the STHE design operability under the expected range of disturbance factors and lower
the total cost.
* Corresponding author.
E-mail address: vska@nchu.edu.tw (V.S.K. Adi).
1
Independent Scholar, Taichung, 40249, Taiwan
https://doi.org/10.1016/j.csite.2022.102163
Received 30 January 2022; Received in revised form 27 May 2022; Accepted 28 May 2022
Available online 5 June 2022
2214-157X/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
1. Introduction
Shell and tube heat exchanger (STHE) is widely used in many industries, i.e., industrial power plants as condensers, chemical and
petrochemical plants as preheating systems, or cooling systems. Shell and tube heat exchanger consists mainly of tubes, shell, front
head, rear head, baffles, and other components. Traditionally, [1,2] incorporate geometric features to design the shell and tube heat
exchanger, the so-called Kern, and the Bell-Delaware method. These approaches establish the foundation of the current shell and tube
heat exchanger structure.
In recent years, many studies have analyzed STHE design from different angles and used various optimization techniques to achieve
better design improvements. [3] consider minimizing the entropy as an objective function to estimate the optimal heat exchanger
conditions in multiple aspects. [4] indicate that the STHE should be connected in series to increase the temperature correction factor
(Ft) by more than 0.8 to avoid infeasible situations and temperature cross. [5] believe that the objective function should minimize the
heat transfer surface area or the equipment capital cost.
The optimization method in designing STHE has recently received widespread attention. Various stochastic algorithms have been
developed that involve particle swarm optimization [6–8], falcon optimization algorithm [9] and other related algorithms. The
metaheuristic algorithm can provide a better or satisfactory solution in the case of incomplete information. Many successful meta
heuristic algorithms are based on biological phenomena and run by simulating different processes in a computer. Some of which are
well-known, for example, genetic algorithm (GA) [10], differential evolution (DE) [11], particle swarm optimization (PSO) [12]. New
alternative algorithms perform beyond the previous optimization methods. Models with nonlinearity and multimodality are also
addressed [13]. [14] optimized the structure of the STHE with 11 discrete variables through a binary coded genetic algorithm. The
heat load is optimized while minimizing the cost function. Later, [15] used the statistical software MATLAB genetic algorithm toolbox
to minimize STHE costs. Using three optimization variables, the study selects equipment construction cost and operating cost as the
cost targets. The results show that compared with the traditional design method, it can save up to 7.4% of the construction cost and
93% of the operating cost. The total cost is reduced by 52%. In addition, the design problem can be solved quickly. The results show
that the genetic algorithm is promising in achieving optimal results. In later studies, multi-objective optimization of shell and tube heat
exchangers is addressed [7,16], with concerns on effectiveness and minimization of total cost, pressure drop, and other objectives.
Although this study addresses the STHE design, there are studies on the optimization of different types of heat exchangers based on
metaheuristic algorithms, i.e., plate-fin heat exchanger [7], air-fin cooler [17], and many more. Most of the above approaches are
based on user-given process requirements without considering any process uncertainties.
In fact, uncertain issues always exist in various chemical processes and need to be addressed. These uncertainties may come from
random external interference, such as raw material quality, product composition, environmental factors, or internal interference, such
as overall heat transfer coefficient, chemical reaction rate constant, and other physical properties [18]. Because the traditional design
does not consider these factors, the resulting configuration may not achieve the required target or reduce economic efficiency when
operating in actual conditions. Therefore, the ability of a given system to maintain operation without deviating from normal condi
tions, called operational flexibility [19], is obviously a crucial factor. Any working process or design should take operational flexibility
into consideration. This way, additional operating costs from an inoperable design due to process uncertainty can be reduced. [20,21]
first defined the steady-state flexibility index (FIs), which is used as a parameter measurement of feasible regions in the parameter
space. Under the assumption of convexity, the feasible operating critical point must meet the apex of the uncertain parameter and
inequality limit space to determine the operating limit or load of the system. This method is called the vertex method. [22] also
developed the active constraints method in the form of mixed-integer nonlinear programming (MINLP). Similar approaches are also
carried out in subsequent studies to improve the design [23,24].
In this study, the issues of STHE design under process uncertainty are addressed. The rigorous operating constraints are taken into
account to establish the structural design of the STHE while simultaneously incorporating the flexibility index. This study can be
divided into the following three steps:
Specifically, for the accurate and smooth operation of STHE, the process uncertainty is considered. Therefore, a steady-state
flexibility index with vertex method is implemented to evaluate the impact and load of STHE design under uncertainties. In-depth
analysis is then conducted for improvement and optimization.
This work is organized as follows: Section 2 is dedicated to the STHE mathematical models. Section 3 describes the optimization
strategy of STHE. The case study and optimization results are discussed in Section 4. Finally, conclusions are given in the last section.
2
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
where, do and di are outside diameter and inside diameter of the tube, hs and ht are the heat transfer coefficient for the fluid flowing in
the shell side and tube side, Rs and Rt are the scaling or fouling resistance of the shell side and the tube side.
The energy balance equation of the heat exchanger can be expressed by the following equation:
( )
Q = ṁh Cp,h Th,i − Th,o (2)
( )
Q = ṁc Cp,c Tc,o − Tc,i (3)
where Q is the heat duty between two workflows, ṁh and ṁc are the hot stream and the cold stream mass flow rate, Cp,h and Cp,c are the
hot stream and cold stream heat capacity, Th,i and Tc,i are the inlet hot and cold stream temperatures, Th,o and Tc,o are outlet hot and
cold stream temperatures.
The logarithmic mean temperature difference (ΔTlm ) of the heat exchanger expressed as follows:
( ) ( )
Th,i − Tc,o − Th,o − Tc,i
ΔTlm = ( ) (4)
(Th,i − Tc,o )
ln T − T
( h,o c,i )
The heat exchanger design equation is generally used to estimate the required heat transfer surface area as follows:
Q = Uo AFt ΔTlm (5)
where A is the required heat exchanger area, Ft is the correction factor. In multiple flow heat exchangers, Ft is generally required to be
above 0.8. Note that this study is limited to 1-1 and 1–2 STHE pass configurations.
The following is the relevant equation for the correction factor Ft of the 1–2 shell and tube heat exchanger in this study:
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅ [( )/( )]
R2 + 1ln 1 − P1,2 1 − RP1,2
Ft = {[ ( √̅̅̅̅̅̅̅̅ )]} for R ∕
=1 (6)
2− P1,2 R+1− R2 +1
(R − 1)ln [ ( √̅̅̅̅̅̅̅̅ )]
2
2− P1,2 R+1+ R +1
√̅̅̅
2P
Ft = { 1,2 √̅̅ } for R = 1 (7)
( ) [2− P1,2 (2− 2 )]
1 − P1,2 ln 2− P 2+√̅̅2
[ 1,2 ( )]
where P1,2 is the thermal effectiveness of each 1–2 shell in the series, which can be expressed by the following equations (8) and (9). R
is the heat capacity ratio, as expressed by the following equation (10).
( 1−RP
)N1
1− s
P1,2 = 1− P
( 1− ) 1 for R ∕
=1 (8)
RP N
R− 1− P
s
P
P1,2 = for R = 1 (9)
P − Ns P + Ns
where Ns is the number of heat exchangers, P is the thermal effectiveness. The equation for heat capacity ratio (R) and thermal
effectiveness (P) mentioned above are as follows:
Cpc Thi − Tho
R= = (10)
Cph Tco − Tci
3
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Table 1
Colburn factor j coefficient [14].
Tco − Tci
P= (11)
Thi − Tci
where hid is the ideal heat transfer coefficient of pure cross-flow (stream B) at the center of the shell. Jc is the correction coefficient for
the distance between the baffle cut and the baffle and considers heat transfer into the window zone. Jl is the baffle leakage correction
coefficient and considers the shell-to-baffle leakage (stream E) and the tube-to-baffle hole leakage (stream A). Jb is the correction
coefficient of the split flow between the tube bundle and the shell (stream C) and the split flow between the tube and the cross-flow
placed baffle (stream F), and consider the number of side sealing strips in the split flow. Js is the correction coefficient for the distance
between the entrance and the exit baffle. Jr is the correction coefficient of the adverse temperature gradient at low Reynolds number
laminar flow.
The ideal heat transfer coefficient hid of the pure cross-flow (stream B) in the center of the shell side is as follows:
(13)
− 2
hid = jGs Cps Prs 3
where j is the Colburn factor, Cps is the shell-side heat capacity, Gs is the shell-side mass velocity, Prs is the shell side Prandtl number.
The relation expression of j, Gs is as follows:
( )a
1.33
j = a1 *
Res a2 (14)
Pt
ṁs
Gs = (15)
Ao,cr
where Res is the shell side Reynolds number, Pt * is the tube pitch ratio, ṁs is the shell-side mass flowrate, Ao,cr is the cross-flow section
at or near the shell centerline for one cross-flow section. a1 , a2 , a3 , and a4 can be obtained in Table 1, and the related expressions of Res ,
a and Ao,cr are as follows:
a3
a= (16)
1 + 0.14Res a4
ṁs do Gs do
Res = = (17)
μs Ao,cr μs
4
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
[ ]
Dctl
Ao,cr = Lb,c (Ds − Dotl ) + (Xt − do ) (18)
Xt
where Lb,c is the baffle spacing at the center, Ds is the shell diameter, Dotl is the tube bundle outer diameter, Dctl is a specified diameter
of the circle through the centers of the outermost tubes, Xt is the transverse tube pitch. The details of baffle geometry can be found in
the literature [26,31].
The relation between the correction coefficient Jc of the baffle cut and the baffle spacing is as follows:
Jc = 0.55 + 0.72Fc (19)
where Fc defined as the fraction of all the tube bundles in the cross-flow section.
The relation between the correction coefficient Jl of the shell-to-baffle leakage (stream E) and the tube-to-baffle hole leakage
(stream A) is as follows:
Jl = 0.44(1 − rs ) + [1 − 0.44(1 − rs )]exp( − 2.2rlm ) (20)
Ao,sb + Ao,tb
rlm = (22)
Ao,cr
where Ao,sb is the shell-to-baffle leakage area, Ao,tb is the tube-to-baffle leakage area.
The relation between the correction coefficient Jb of the tube bundle-to-shell bypass (stream C) and tube pass partition bypass
(stream F) is as follows:
⎧
⎪ + 1
⎪
⎨ 1 for Nss ≥
2
Jb = ( [ ]) (23)
⎪
⎩ exp − Cbp Fbp 1 − (2Nss + )1/3 for Nss + < 1
⎪
2
where Fbp is for a fraction of cross-flow area available for flow bypass and the relation between Cbp , Fbp , and Nss + is as follows:
{
1.35 for Res ≤ 100
Cbp = (24)
1.25 for Res > 100
( )
Lbc Ds − Dotl + 0.5Np wp
Fbp = (25)
Ao,cr
Nss
Nss + = (26)
Nr,cc
Ds (1 − 2lc )
Nr,cc = (27)
Xl
where Np is the number of pass divider lanes, wp is the width of the pass divider lane, Nss is the number of sealing strip pairs, Nr,cc is the
number of tube rows crossed during flow through one cross-flow section, Xl is the longitudinal tube pitch as described by [26].
The following will describe the relationship between the correction coefficient Js for the baffle spacing in the outlet region as
follows:
where Nb is the number of baffles calculated for all the number of cross-flow and window zone, which should be determined from
drawing or direct counting. Otherwise Nb can be expressed by the following equation:
L − Lb,i − Lb,o
Nb = +1 (29)
Lb,c
where L is the tube length, Lb,i is the baffle spacing in the inlet region, Lb,o is the baffle spacing in the outlet region.
Finally, the following will depict the relation equation for the correction coefficient Jr of the adverse temperature gradient at low
Reynolds number laminar flow as follows:
5
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
( )0.18
10
Jr = for Res ≤ 20 (30)
Nr,c
where Nr,c is the number of effective tube rows crossed during flow through one baffle section. The relation of Nr,c is as follows:
Nr,c = Nr,cc + Nr,cw (33)
where Nr,cw is the number of effective tube rows crossed during flow through one window zone.
where kt is the tube-side thermal conductivity, di is the tube inner diameter, Prt is Prandtl number, Ret is the tube-side Reynolds
number, μ is the fluid viscosity at the bulk fluid temperature, μw is the fluid viscosity at the wall. The relation between Prt and Ret is as
follows:
ρt Vt di
Ret = (35)
μt
Cpt μt
Prt = (36)
kt
where ρt is tube-side density, Vt is tube-side velocity, μt is tube-side viscosity, Cpt is tube-side heat capacity. The relation Vt is as
follows:
ntp ṁ
Vt = ( /t ) (37)
Nt π di 2 4 ρt
where ṁt is the tube-side mass flowrate, ntp is number of tube pass.
where ΔPio is the pressure drop in the inlet and outlet sections, ΔPcr is the pressure drop in the cross-flow section, ΔPw is the pressure
drop in the window area. The relation between ΔPio , ΔPcr , and ΔPw are as follow:
( )
Nr,cw
ΔPio = 2ΔPb,id 1 + ζζ (39)
Nr,cc b s
where ΔPb,id is the ideal pressure drop in the cross-flow section, ζb is the correction factor for stream C and stream F, ζs is the correction
factor of the baffle spacing in the inlet and outlet region and in the central region. ζl is the correction factor of the tube-to-baffle and
shell-to-baffle leakage (stream A and stream E). ΔPw,id is the ideal pressure drop in the window area. The relations of ΔPb,id , ζb , ζs , ζl ,
and ΔPw,id are in the following order:
6
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
4f id Gs 2 Nr,cc
ΔPb,id = (42)
2ρs
⎧
⎪ + 1
⎪
⎨ 1 for Nss ≥
2
ζb = ( [ ]) (43)
⎪
⎩ exp − Dbp Fbp 1 − (2Nss + )1/3 for Nss + < 1
⎪
2
( )2− n
′ ( )2− n
′
Lb,c Lb,c
ζs = + (44)
Lb,o Lb,i
where f is the friction factor. As the preferred Ret is in turbulent region, f is defined as follows [32]:
f = 0.046Ret − 0.2
for Ret > 4000 (47)
where A is the equipment capacity or size. K1 , K2 , and K3 are equipment constants to be calculated.
According to different operating pressure, the equipment designed in this study will have different pressure factors to be
considered. According to [33]; the equipment pressure factor Fp is expressed with the general equation as follows:
where P is operating pressure for equipment and the unit is gauge or barg. C1 , C2 , and C3 are the equipment pressure constants to be
calculated.
The material and bare module cost CBM is defined according to the restrictions on the types of materials used in the equipment. The
general equation is expressed as follows:
7
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Table 2
STHE bare module cost constant.
K1 K2 K3 C1 C2 C3
Table 3
Centrifugal pump bare module cost constant.
K1 K2 K3 C1 C2 C3
where FBM is the material and bare module factor, FM is the material factor, B1 and B2 is the bare module factor consists of direct and
indirect cost factors.
The heat exchanger design in this study involves a single tube pass and a double tube pass. The fixed tube sheet heat exchanger is
selected to facilitate broader utilization. It is commonly applicable to high temperature and high pressure, simple in structure, and
relatively cheap. According to [33], the constant factor corresponding to equations (48)–(50) are listed and summarized as shown in
Table 2.
K1 , K2 , and K3 are suitable for fixed tube sheet heat exchanger areas in the range of 10–1000 m2 C1 , C2 , and C3 are suitable for an
operating pressure range of 5–140 barg.
9.8 × Q × H × ρ P×Q P × ṁ
Wpump = = = (52)
η η η×ρ
where Wpump is the pump power, Q is the volume flow rate, H is the head, η is the pump efficiency. In this study, η is set to 0.8.
Similarly, according to [33], the constant factors correspond to the centrifugal pump required by equation (50)–(52) are listed in
Table 3.
K1 , K2 , and K3 are suitable for centrifugal pump operating power in the range 1–300 kW, C1 , C2 , and C3 are suitable for operating
pressure range of <10 barg.
OC = (Ws + Wt ) × op × ec (54)
where Ws and Wt are the work done by the centrifugal pump for the shell and tube side, op is the operating time of the whole year. In
this study, 365 days a year is deducted with 11 days of Taiwan national holidays, and the equipment is operated 24 h a day, so it is
(365 − 11) × 24 = 8496 hr.
i(1 + i)n
(55)
′
IC = CBM
(1 + i)n − 1
8
L.-Y. Chen et al.
9
TAC = IC + OC (56)
where CBM includes the heat exchanger in section 2.3.1 and two centrifugal pumps in section 2.3.2, namely CBM = CBM,HE +
′ ′ ′
2CBM,pump . i is the annual interest rate set to 5%. n is the service life set to 20 years. IC is the equipment cost including interest in
′
$/year.
gj (d, z, x, θ) ≤ 0, ∀j ∈ J (60)
where hi is the ith equality constraint in the design model, such as the mass or energy balance equation in the steady-state process unit.
gj is the jth inequality constraint in the design model, such as the upper and lower constraint of operating flow or temperature.
Equations 59 and 60 can represent all of the equations as mentioned previously (1)–(56). d is the design variable for all design
specifications, which is the fixed-parameter vector given by design. z is expressed as a vector parameter of the adjustable control
variable. x is the vector parameter of the passively affected state variable. θ is defined as the vector parameter of the uncertain
parameter.
The following mathematical formula can be used to determine the so-called feasibility function ψ(d, θ) expressed as follows:
ψ(d, θ) = minmaxgj (d, z, x, θ) (61)
x,z j∈J
Subject to the constraint of equation (60), it can be noted that the above mathematical formula is intended for a fixed design parameter
(d) and a fixed value of the uncertain parameter (θ). The maximum gj (∀j ∈ J) is achieved by adjusting the control variable (z) and
maintaining hi = 0(∀i ∈ I). For ψ(d, θ) = 0, the system model needs to touch one of the inequalities, that is gj = 0(∃j ∈ J). Any region
with ψ(d, θ) ≤ 0 is a feasible operating region and ψ(d, θ) > 0 is an infeasible operating region.
Because the feasibility function ψ(d, θ) is evaluated based on the fixed range of uncertain parameter θ, it is necessary to
comprehensively consider all of the possible values of all uncertain parameters. Here, the hypercube T in the parameter space is
defined as follows:
{ ⃒ }
T = θ⃒θN − Δθ− ≤ θN ≤ θN + Δθ+ (62)
where θN represent as a vector value of the nominal parameter. Δθ− and Δθ+ are the expected deviation in the negative and positive
directions. Therefore, an additional optimization problem is rigorously formulated as follows:
χ(d) = maxψ(d, θ) (63)
θ∈T
where χ(d) is expressed as the feasibility function on the fixed design variable d in the hypercube T. If χ(d) ≤ 0, the given system is
feasible, otherwise it is not feasible, as shown in Fig. 2(a).
[20,21] introduce the measurement of the maximum tolerance under each uncertain parameter by expanding or contracting the
hypercube T through the scalar variable δ described as follows:
{ ⃒ }
T(δ) = θ⃒θN − δΔθ− ≤ θN ≤ θN + δΔθ+ (64)
where δ is the steady-state flexibility index FIs which is maximized to expand the hypercube T so that χ(d) = 0. The mathematical
formula is expressed:
FIs = maxδ (65)
subject to
χ(d) ≤ 0 (66)
Note that the steady-state flexibility index FIs is maximized to determine the maximum value of δ, ensuring gj ≤ 0(∀j ∈ J) and χ(d) ≤ 0.
A δ ≥ 1 actually implies that the system design is feasible under the constraint of the original equation (62).
There have been several developments [35] for the optimization problem in equations 65 and 66. The model can be significantly
10
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Table 4
STHE design variables.
Tube outer diameter, do (m) 8 values: 0.01587, 0.01905, 0.02222, 8 values: 0.01587, 0.01905, 0.02222, 0.0254, 0.03175, 0.0381,
0.0254,0.03175, 0.0381, 0.0508, 0.0635 0.0508, 0.0635
Tube wall thickness, Twt(m) fix value: 0.120 (BWG = 11) 8 values: 0.0030, 0.0028, 0.0024, 0.0021, 0.0018, 0.0017, 0.0015,
0.0012, 0.0011, 0.0009 (BWG ¼ 11–20)
Shell inner diameter, Ds (m) 8 values: 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1 21 values: 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8,
0.85, 0.9, 0.95, 1, 1.05, 1.1, 1.15, 1.2, 1.25, 1.3
Baffle cut, Bc (%) 4 values: 25, 30, 40, 45 4 values: 25, 30, 40, 45
Tube pitch, Pt (m) × (do) 2 values: 1.25, 1.5 2 values: 1.25, 1.5
Baffle spacing at the center, Lbc 8 values: 0.2, 0.25, 0.3, 0.35,0.4, 0.45, 0.5, 0.55 Continuous variables: 0.2–0.55
(m) × (Ds)
Baffle spacing at the inlet and 4 values: 1, 1.2, 1.4, 1.6 Continuous variables: 1–1.6
outlet, Lboi (m) × (Lbc)
Tube-to-baffle diametrical 4 values: 0.01, 0.04, 0.07, 0.1 4 values: 0.01, 0.04, 0.07, 0.1
clearance, δtb (m) × (do)
Shell-to-baffle diametrical 4 values: 0.01, 0.04, 0.07, 0.1 4 values: 0.01, 0.04, 0.07, 0.01
clearance, δsb (m) × (Ds)
Tube bundle outer diameter, Dotl 4 values: 0.8, 0.85, 0.9, 0.95 4 values: 0.8, 0.85, 0.9, 0.95
(m) × (Ds-δsb)
Tube layout pattern (deg) 3 values: 30, 45, 90 3 values: 30, 45, 90
Tube pass, ntp 2 values: 1, 2 2 values: 1, 2
Tube length, L or Lass (m) Length change with exchanger area 5 values: 2.438, 3.048, 3.658, 4.877, 6.096
Lass = Areal ÷ (πdo Nt )
Number of seal strip pairs, Nss fix value: 2 3 values: 0, 1, 2
width of the pass divider lane, wp fix value: 0.05 fix value: 0.05
(m) × (Ds)
s.t.
simplified by assuming that the optimal solution is always related to one of the vertices in the hypercube parameter space [19].
Suppose Δθk (∀k ∈ V) is defined as a vector from the nominal point θN to the kth vertex and V is the set of all vertices, then there are
maximum values of δk from θN to all directions of the vertices (Δθk ). The smallest δk from all sets can be deemed as the steady-state
flexibility index FIs , as shown in Fig. 2(b).
The following will introduce the detailed mathematical programming model of the vertex method:
δk = maxδ (67)
x, z,δ
θ = θN + δΔθk (68)
From the resulting hypercube space, namely T(θk ) for k ∈ V, it is clear that only the smallest T(θk ) which corresponds to the smallest δk
can be fully described in the feasible region by equations 59 and 60, so:
{ }
FIs = min θk (69)
k∈V
The following are the simple optimization steps of the vertex method:
Step 1: Set the nominal point θN for the system and determine the type of uncertain parameter θ. Consider the coordinate dimension
then set the expected upper and lower constraint of the uncertain parameter Δθk . Determine the control variable of the system and
analyze the system’s feasibility under the available constraints by adjusting the control variable.
Step 2: Solve the optimization problem described by equations 59 and 60 and equations 67 and 68 for each vertex k ∈ V.
Step 3: Use equation (69) to determine FIs of the particular system design parameter d.
11
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
approach estimates the state of the operating fluid outlet under the given conditions of the internal geometry, heat exchange area, and
heat load of the STHE. In this study, the optimization approach in the traditional literature is improved significantly.
(70)
′
TAC = IC + OC + [10(1 − FIs )OC]
where TAC is the total annual cost with the penalty. IC is the average annual equipment cost with an annual interest rate, as shown in
′
equation (55). OC is the operating cost in equation (54). FIs is the steady-state flexibility index from the vertex method as shown in
equation (69).
The reasoning for equation (70) formulation is that the TAC is defined as equipment cost + operation cost + failed operation cost.
′
Suppose STHE’s FIs are equal to 1. In that case, that means the STHE can meet all the required operating conditions under the given
uncertainty range. No additional cost is necessary to trade off any unexpected offset. On the other hand, if STHE’s FIs is lower than 1,
that means the STHE may fail at some range of uncertainty range to meet the required operating conditions. (1 − FIs ) represents the
degree of failure, and the costs associated with the inability to return to normal would require 10 times the OC price, which is the
penalty function.
12
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Fig. 3. (a) Flexibility analysis computational procedure using vertex method; (b) The computational procedure for optimal STHE design with inherent flexibility.
13
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Table 5
Two case studies for STHE design.
CASE I CASE II
Tube side (Cold fluid) Shell side (Hot fluid) Tube side (Cold fluid) Shell side (Hot fluid)
14
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Table 6
Optimal STHE design based on TAC objective function.
Table 7
Cost breakdown for optimal STHE design based on TAC objective function.
The first method is implemented for sequential optimization with genetic algorithm and flexibility analysis. Similarly, as the second
method, simultaneous optimization using genetic algorithm with embedded flexibility analysis can be formulated accordingly.
The modified steps for the second method are as follows:
【Step 13】: Replace Step 11 and use equation (70) as the fitness function to find penalized TAC for each design candidate.
′
Fig. 3(a) is a schematic diagram of the operation flow of flexibility analysis using the vertex method.
【Step 14】: Find the maximum annual total cost with penalized TAC for a given structure of STHE
′
By using the vertex method and adjusting the control variable, the FIs from each vertex can be measured from equation (70) to
obtain the total annual cost with penalized TAC . The maximum TAC among all vertices is considered from the most conservative
′ ′
point of view, representing the worst state condition, as the final result of this step.
【Step 15】: Return TAC to the genetic algorithm
′
Since the design calculation process has two layers, the outer genetic algorithm is to find the minimum value of TAC for STHE with
′
the minimum cost among various geometric structures generated from the genetic algorithm. The inner layer uses flexibility
analysis to find the maximum TAC from all the vertices of the STHE structure design by using the vertex method. Finally, the results
′
are sent to the outer genetic algorithm for fitness value evaluation. In other words, different structure of STHE is compared in term
of conservative TAC s for to find the best solution with minimum cost.
′
【Step 16】: When the genetic algorithm termination condition is met, the overall computational procedure of STHE design is
completed.
The complete design procedure is depicted in Fig. 3(b) for clarity.
4. Case study
Two design strategies are elaborated in the case study to highlight the advantages of the proposed design method. The traditional
design strategy is compared in the first part of the case study. The flexibility index’s consideration is included in the design strategy and
15
L.-Y. Chen et al.
16
Table 8
Relevant parameters for the flexibility analysis of STHE design.
Table 9
The physical upper and lower bound of cold and hot fluid inlet temperatures.
CASE I CASE II
Tube side (Cold fluid) Shell side (Hot fluid) Tube side (Cold fluid) Shell side (Hot fluid)
discussed in the latter part of the case study. In addition, the computer resources used for the computational works in this study is
Intel® Core™ i7-8700 CPU @ 4.60 GHz, 64 GB RAM, Windows 10 64 bit, and MATLAB version 2018a. The MATLAB’s built-in GA
Toolbox is used for genetic algorithm. The source codes for the case studies are available upon request.
where θmin and θmax are the minimum and the maximum limits of the tolerance values. It can be noted that θ− and θ+ indicate the range
of uncertain parameters, and θmin and θmax are the corresponding physical upper and the lower bounds.
The flexibility analysis results of both old and new designs are shown in Fig. 6. According to equation (76), this study assumes that
the cold and the hot fluid inlet temperatures can be changed through the additional heating network, but the minimum temperature
θmin and the maximum θmax are constrained based on the following Table 9.
18
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Fig. 6. Flexibility analysis results of (a) Case 1 and (b) Case 2.
19
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Table 10
Cost breakdown for Case 1 and Case 2 with flexibility analysis.
Table 11
Optimal STHE design based on TAC objective function.
′
Table 12
Cost breakdown for optimal STHE design based on TAC objective function.
′
In Fig. 6, FIsk is the value of the flexibility index of the corresponding vertex k. TAC k is the value of the total annual cost with
′
penalty in equation (70). Nominal is the original design nominal point located at the coordinate points (Tci, Thi). zmck is the mass flow
rate of cold fluid at each vertex k and the value is indicated for the required cold fluid mass flow rate to achieve the temperature target
shown in Table 8. Consequently, there are a total of four vertices in FIs and TAC represented by numbers.
′
20
L.-Y. Chen et al.
21
In addition, the TAC evaluation concept of the four vertices is decribed as follows:
′
Due to the influence of the uncertain parameters of the XY axis, the STHE will adjust the control parameters shown in Table 8 on
four vertices. At the same time the control range is limited as shown in Table 9 to achieve the target Tho. The FIs of the STHE design will
be determined based on the given uncertainty range. The changes of control parameters will be computed into operating cost OC, i.e.,
cooling costs. The four vertices of TAC can be calculated through equation (70).
′
As observed from Fig. 6(a), the third vertex in Case 1 is the worst case, represented in red. Increasing the cold fluid mass flow rate
mc or velocity Vt is necessary when the cold fluid feed temperature Tci and the hot fluid feed temperature Thi both rise. Due to the
constraints shown in Table 8, the control parameters have been adjusted to the upper limit Vt = 2 m/s, that is, mc = 62.56 kg/s, which
renders the flexibility index FIs unable to increase. Showing that FIs < 1 is partially feasible, to make up the failed part of the operation
to succeed, an additional penalty cost of 10 times is required. Therefore the TAC is affected by the penalty value and increased rapidly.
′
A similar explanation applies to the remaining vertices 1, 2, and 4. Note that on the fourth vertex, the STHE can handle such extreme
condition. Therefore FIs > 1 is entirely feasible, and there is no need to spend another 10 times the penalty cost. Hence TAC is not
′
sequent approach, the simultaneous optimization strategy is carried out by directly embedding equation (70) for the genetic algo
rithm’s fitness function. Each individual generated from the genetic algorithm are evaluated based on equation (70) to find the lowest
conservative TAC among the candidates. Accordingly, the optimal STHE design can be summarized in Table 11. The costs breakdown
′
is shown in Table 12. The simultaneous optimization results with flexibility analysis are shown in Fig. 5.
From Fig. 7, the third vertices in Case 1 and Case 2 are the worst case. Note that the control variable mc ranges are now different
since the corresponding optimal STHE designs are significantly different from the previous sequential approach (see Tables 6 and 11).
Based on the results of Fig. 7, the third vertices are the most conservative point of view. With the simultaneous optimization strategy,
the corresponding TAC of Case 1 and Case 2 are significantly lower than the one with sequential approach (see Tables 10 and 12). In
′
Table 12, the most conservative operating cost OC of the latter method is significantly lower due to the difference in STHE structure.
With such different values of OC, the OC + OCpenalty is consequently affected as part of the total annual cost TAC . Note that the STHE
′
equipment cost ICHE for Case 1 and Case 2 under different approaches are pretty similar. However, the second approach simultaneously
considers the influence of uncertain parameters to determine STHE structure. Consequently, the simultaneous approach can reduce the
corresponding OC + OCpenalty and the tube-side pumping equipment cost ICt, pump, thereby reducing The TAC significantly for the
′
simultaneous method. Depending on the design approach, it can be inferred that with more design parameters optimized at one time,
the better the results of the STHE structure can be obtained. Due to the simultaneous optimization, all possible combinations can be
explored at the most comprehensive level, instead of the sequential approach. Moreover, by considering rigorous operational con
straints, the operational feasibility is ensured for any STHE designs from the optimization procedure.
5. Conclusions
This study has successfully implemented a set of shell and tube heat exchanger design strategies with a flexibility index to evaluate
the design’s operability from a given uncertainty range. The sequential design approach may result in low TAC. Still, when process
uncertainty is considered, the overall operational costs may tell otherwise. By introducing multilevel optimization with flexibility
analysis, the STHE design and operability issues can be simultaneously addressed to obtain the lowest overall costs. The simultaneous
approach is expected to be adapted further on a larger scale, i.e., heat exchanger network, so that the overall plant performance can be
improved accordingly.
22
L.-Y. Chen et al. Case Studies in Thermal Engineering 37 (2022) 102163
Acknowledgments
This work was partially supported by the Ministry of Science and Technology, Taiwan, under contract no. MOST 106-2221-E-005-
095-, MOST 107-2221-E-005-034-, the 105科技部補助大專校院延攬特殊優秀人才award, and a grant from National Chung-Hsing
University 10617003G 新進教師經費補助.
References
[1] D.Q. Kern, Process Heat Transfer, McGraw-Hill, New York, NY, 1950.
[2] K.J. Bell, Final Report of the Cooperative Research Program on Shell-And-Tube Heat Exchangers. University of Delaware Engineering Experiment Station
Bulletin No. 5, University of Delaware, Newark, DE, 1963.
[3] E. Johannessen, L. Nummedal, S. Kjelstrup, Minimizing the entropy production in heat exchange, Int. J. Heat Mass Tran. 45 (2002) 2649–2654.
[4] J.M. Ponce-Ortega, M. Serna-González, A. Jiménez-Gutiérrez, Design and optimization of multipass heat exchangers, Chem. Eng. Process: Process Intensif. 47
(2008) 906–913.
[5] A.L.H. Costa, M. Queiroz, Design optimization of shell-and-tube heat exchangers, Appl. Therm. Eng. 28 (2008) 1798–1805.
[6] R.V. Rao, V.K. Patel, Design optimization of shell and tube heat exchangers using swarm optimization algorithms, Proc. IME J. Power Energy 225 (2010)
619–634.
[7] R.V. Rao, V.K. Patel, Multi-objective optimization of heat exchangers using a modified teaching-learning-based optimization algorithm, Appl. Math. Model. 37
(2013) 1147–1162.
[8] V.K. Patel, R.V. Rao, Design optimization of shell-and-tube heat exchanger using particle swarm optimization technique, Appl. Therm. Eng. 30 (2010)
1417–1425.
[9] E.H. de Vasconcelos Segundo, V.C. Mariani, L. dos Santos Coelho, Design of heat exchangers using falcon optimization algorithm, Appl. Therm. Eng. 156 (2019)
119–144.
[10] J.H. Holland, Genetic algorithms, Sci. Am. 267 (1992) 66–72.
[11] R. Eberhart, J. Kennedy, A new optimizer using particle swarm theory, in: Proceedings of the Sixth International Symposium on Micro Machine and Human
Science, 1995, pp. 39–43.
[12] R. Storn, K. Price, Differential evolution – a simple and efficient heuristic for global optimization over continuous spaces, J. Global Optim. 11 (1997) 341–359.
[13] S. Salcedo-Sanz, Modern meta-heuristics based on nonlinear physics processes: a review of models and design procedures, Phys. Rep. 655 (2016) 1–70.
[14] P. Wildi-Tremblay, L. Gosselin, Minimizing shell-and-tube heat exchanger cost with genetic algorithms and considering maintenance, Int. J. Energy Res. 31
(2007) 867–885.
[15] A.C. Caputo, P.M. Pelagagge, P. Salini, Heat exchanger design based on economic optimization, Appl. Therm. Eng. 25 (2008) 1151–1159.
[16] B.D. Raja, R.L. Jhala, V.K. Patel, Many-objective optimization of shell and tube heat exchanger, Therm. Sci. Eng. Prog. 2 (2017) 87–101.
[17] O.E. Turgut, A novel chaotic manta-ray foraging optimization algorithm for thermo-economic design optimization of an air-fin cooler, SN Appl. Sci. 3 (2021) 3.
[18] F.V. Lima, C. Georgakis, J.F. Smith, P.D. Schnelle, D.R. Vinson, Operability-based Determination of Feasible Control Constraints for Several High-dimensional
Nonsquare Industrial Processes, vol. 56, American Institute of Chemical Engineers, 2010, pp. 1249–1261.
[19] K.P. Halemane, I.E. Grossmann, Optimal Process Design under Uncertainty, vol. 29, American Institute of Chemical Engineers, 1983, pp. 425–433.
[20] R.E. Swaney, I.E. Grossmann, An index for operational flexibility in chemical process design, Part I: Formulation and theory. American Institute of Chemical
Engineers 31 (1985) 621–630.
[21] R.E. Swaney, I.E. Grossmann, An index for operational flexibility in chemical process design, Part II: Computational algorithms. American Institute of Chemical
Engineers 31 (1985) 631–641.
[22] I.E. Grossmann, C.A. Floudas, Active constraint strategy for flexibility analysis in chemical processes, Comput. Chem. Eng. 11 (1987) 675–693.
[23] C.T. Chang, B.H. Li, C.W. Liou, Development of a generalized mixed integer nonlinear programming model for assessing and improving the operational
flexibility of water network designs, Ind. Eng. Chem. Res. 48 (2009) 3496–3504.
[24] E. Riyanto, C.T. Chang, A heuristic revamp strategy to improve operational flexibility of water networks based on active constraints, Chem. Eng. Sci. 65 (2010)
2758–2770.
[25] R.K. Shah, Heat exchanger basic design methods, in: S. Kakaç, R.K. Shah, A.E. Bergles (Eds.), Low Reynolds Number Flow Heat Exchangers, Hemisphere,
Washington, DC, 1983, pp. 21–72.
[26] R.K. Shah, D.P. Sekulić, Fundamentals of Heat Exchanger Design, John Wiley & Sons, Hoboken, NJ, 2003.
[27] T. Tinker, Shell Side Characteristics of Segmentally Baffled Shell-And-Tube Heat Exchangers, Parts I, II, III, General Discussion On Heat Transfer, Institute of
Mechanical Engineers, London, UK, 1951, pp. 89–116.
[28] T. Tinker, Shell side characteristics of shell-and-tube heat exchangers–– a simplified rating system for commercial heat exchangers, Trans. ASME (1958) 36–52.
[29] J.W. Palen, J. Taborek, Solution of shell side flow, pressure drop and heat transfer by stream analysis method, Chem. Eng. Prog. Symp. Ser. 65 (1969) 53–63.
[30] K.J. Bell, Delaware method for shell-side design, in: R.K. Shah, E.C. Subbarao, R.A. Mashelkar (Eds.), Heat Transfer Equipment Design, Hemisphere,
Washington, DC, 1988, pp. 145–166.
[31] J. Taborek, Shell-and-tube heat exchangers: single phase flow, in: G.F. Hewitt (Ed.), Handbook of Heat Exchanger Design, Begell House, NY, 1998.
[32] R. Smith, Chemical Process Design and Integration, John Wiley & Sons, West Sussex, UK, 2005.
[33] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, D. Bhattacharyya, in: Analysis, Synthesis, and Design of Chemical Processes, fourth ed., Prentice Hall, Upper
Saddle River, NJ, 2012.
[34] Taiwan Power Company, 2018年各國平均電價比較, 2019. Retrieved from, https://www.taipower.com.tw/tc/page.aspx?mid=213&cid=351&cchk=1b3221ee-
37c3-4811-9d4d-a1bb215f33c8.
[35] L.T. Biegler, I.E. Grossmann, A.W. Westerberg, Systematic Methods for Chemical Process Design, Prentice Hall PTR, Upper Saddle River, NJ, 1997.
[36] TEMA, in: Standards of the Tubular Exchanger Manufacturers Association, ninth ed., Tubular Exchanger Manufacturers Association, Tarrytown, NY, 2007.
[37] J.E. Edwards, Design and Rating of Shell and Tube Heat Exchangers, P&I Design, Teesside, UK, 2008.
[38] R. Mukherjee, Effectively design shell-and-tube heat exchangers, Chem. Eng. Prog. 94 (1998) 21–37.
23