Optimal Overcurrent Relay Coordination in

Distribution Networks
Mert Bekir Atsever Ozgur Karacasu Mehmet Hakan Hocaoglu
Department of Electronics Engineering Marmara Research Centre Department of Electrical-Electronics Engineering
Gebze Technical University TUBITAK Istanbul Commerce University
Kocaeli, TURKEY Kocaeli, TURKEY Istanbul, TURKEY
mbatsever@gtu.edu.tr ozgur.karacasu@tubitak.gov.tr mhhocaoglu@ticaret.edu.tr

Abstract—Over-current relay coordination plays a critical role proper constraints. In 1988, Urdaneta [1] proposed optimum
for effective protection of networks since primary protection is relay coordination for three-bus power systems by setting
based on over-current detection in distribution systems. Thus, IP as constant values and solving optimal TDS settings via
optimal relay coordination must be ensured between relays
in faulted condition, providing selectivity and fast operation. linear programming (LP). Although, LP have some advantages
Therefore, relay coordination can be designed as an optimization such as simple solution procedures and faster convergence
problem with highly constrained equations in such networks. time, better results may be obtained in other IP settings
Linear programming, which is simple to be implemented and by setting IP values as variables and applying non-linear
requires less computation time when compared other optimisa- programming (NLP) techniques. However, results may fall
tion techniques, is widely employed for solving relay coordination
problems. To achieve faster and selective protection a number into local optimum since single point search methods or IP
of optimisation algorithms have, also, been customised for the value may not be suitable setting for relay type [2]. Thus,
purpose. Water Cycle Algorithm (WCA) is among them and has optimization problem formulated between suitable relay IP
promising results, already, depicted in the literature. settings. That convert the optimization problem to a mixed
However, capacitive currents may jeopardize traditional opti- integer non-linear programming (MINLP) problem [3]- [4].
misation process for earth faults. Thus, in this paper, performance
analysis is carried out for earth faults in sample distribution For optimizing relay coordination problem, heuristics and
networks constructed with overhead lines and underground swarm based/nature inspired meta-heuristic methods have,
cables, separately. The results show that in cabled networks, also, been implemented. These meta heuristic methods can
additional constraints are needed for the optimisation process be listed as; evolutionary algorithm [5], genetic algorithm [6],
to provide selectivity. Consequently, optimal relay settings and seeker algorithm [7], chaotic firefly algorithm [8], gravitational
selective protection are ensured by adding proper constraints into
optimisation problem in this paper. search algorithm [9], grey-wolf optimization [10], root tree
Index Terms—relay coordination, optimization, water cycle algorithm [11], particle swarm optimization [12], electromag-
algorithm, earth fault netic field optimization [13], hybridized whale optimization al-
gorithm [14], water cycle algorithm (WCA) [15]. Performance
I. I NTRODUCTION analysis of some meta-heuristic optimization approaches can,
Distribution systems may face faults or abnormal condi- also, be found in [16]. A good and recent review of this
tions that can be hazardous for system safety and supply methods is provided in [17].
security. However, there is no global solution to prevent In this paper, optimal relay coordination in distribution
faults occurrence or in other words, fully uninterrupted power networks for earth faults is carried out implementing WCA.
delivery is an impossible mission in a foreseeable future. Results are compared with traditional LP techniques. Two dif-
Thus, main feature of distribution system protection is to ferent case studies are selected to apply both method;i.e. WCA
locate and isolate the system faults without affecting other and LP. First case study represents typical rural distributions
system components. Over-current protection relays constitute networks, constructed with over head lines whose phase-to
the brain of the system since the over-current protection is earth capacitance is negligible. Second case represents typical
the primary protection strategy in distribution networks also urban network with underground cables and has high phase-
back-up protection for the distance protection in transmission to earth capacitances. In this scenario, capacitive currents
systems. circulate in the system under earth fault conditions and effects
To minimize hazardous effects of the faults, it is required the current magnitudes seen by the relays. Thus, traditional
that they should be isolated as quickly and selectively as optimisation constraints may fail on preserving selectivity. To
possible. The selectivity and fast isolation can be achieved ensure selectivity, additional constraints added to the opti-
through optimal coordination between protection relays during mization process. Earth fault simulations carried out using
the faults. Mostly, optimal relay coordination includes two ATP/EMTP (Alternative Transient Program/ElectroMagnetic
parameters to optimize; Time Dial Setting (TDS) and PSM Transient Program) which is highly capable of modelling
(Plug Setting Multiplier) or pick-up current setting (IP ) with overhead lines and underground cables based on their electrical

978-1-6654-4389-0/21/$31.00 © 2021 IEEE

and physical parameters. Linear programming technique is
applied using GAMS (General Algebraic Modelling Systems).
WCA is applied using specially developed codes in MATLAB
environment
Main contribution of this study is to indicate that additional
constraint functions should be attached to the optimisation Fig. 1: Radial distribution system.
problem in order to provide selective protection when the
distribution system is highly cabled. This paper designed as
follows; Section II provides comprehensive explanations of main and backup relays known as Coordination Time Interval
optimisation formulation for distribution networks. Section III (CTI). Commonly used typical values for CTI are 0.2-0.6
presents WCA algorithms and its equations. Case studies are seconds in distribution levels. Relay coordination constraints
detailed in Section IV. Results show that two algorithms are can, mathematically, be integrated as constrains and given in
implemented successfully. Objective functions are converged (3).
in similar numerical values for overhead lines and under-
ground cables separately. Additional constraint equations for
underground cable usage results in higher total time when T (i,k)b − T (i,k)p ≥ CT I (3)
compared with overhead line usage. Both overhead lines
and underground cables scenarios are protected selectively Along with coordination constraints, relays have other con-
and optimally. The requirement for additional constraints to straints such as; TDS boundaries, pick-up current boundaries
preserve selectivity among relays are justified for heavily and operating time boundaries. Relay TDS, pick-up setting
cabled systems. and operating time constraints are stated in Equation (4) to
(6) respectively. TDS values vary between 0.025 to 1.2. Lastly,
II. P ROBLEM F ORMULATION FOR D ISTRIBUTION pick up current arrangement is different depending on faulty
N ETWORKS types. For phase faults, Ip is set 1.5 times higher of nominal
Distribution networks impose severe faults especially in load current or minimum fault current seen by the relays. For
directly earthed system for earth faults which is the most earth faults, Ip is set 15-30% of nominal load current [18],
occurred faults in the distribution networks [18]. Faults must [21].
be isolated from the system as fast as possible along with
preserving selectivity. Thus the sum of operating time of all
T DS imin ≤ T DSi ≤ T DS imax (4)
over-current relays must be minimized. Equation (1) states the
objective function of optimal over-current relay coordination.
X Ipimin ≤ Ipi ≤ Ipimax (5)
objective = min Wi ∗ Ti,k (1)
i
T imin ≤ Ti ≤ T imax (6)
Where;
Wi are the weight factors and set to unity since the fault III. WATER C YCLE A LGORITHMS
occurrence probability of each lines are considered as equal
[19]. In general, such optimization problems are being treated
Ti,k is the operating time of relay i for fault at zone k and with various methods heuristic or meta-heuristic algorithms
its given in (2). such as; evolutionary, stochastic, nature inspired methods so
α ∗ T DSi on. In relay coordination optimization problems many meta-
Ti,k = +c (2) heuristic methods have been implemented in both transmission
IF β
IP −1 and distribution level providing multi search point and escape
Constant values β, a and c are chosen based on relay char- from local optima .
acteristics. In this study, Standard Inverse (SI) characteristic Recently, one of the solution methods that has received a
is selected which is used, widely, in industrial applications. lot of attention in relay coordination optimization is called as
Thus β=0.14, a=0.02 and c =0 is set. Constants for other relay WCA which is well studied for solving constrained engineer-
characteristic can be found in [20]. ing optimisation problem. WCA is inspired by behaviour of
For selective protection, there must be a coordination be- rivers and streams which compose water cycles found in the
tween relays. Figure 1 shows a simple radial distribution nature [22]. This process mathematically expressed as follows
system where a single line to earth fault is occurred on the (7)-(18) [22];
remote end. For this fault two relays (RB and RP ) must Equation (7)-(9) presents initiation of algorithm called ’rain-
sense the fault. Since the RP is the primary relay for fault drops’. Cost function given in (10). Relation between stream
at F, RP trips before RB relay. If RP relay may fail to trip, and rivers given in (11)-(12).
RB relay must trip to ensure selectivity. This arrangement  
called as backup. Specific time delay is arranged between Raindrop = x1 , x2 , x3 , ..., xN (7)
  
Raindrop1
 Raindrop2  Bus 1
 
 . 
P opulation of raindrops = 
  (8) R1 R2 R3
 . 

 . 
RaindropNpop 20 km F2 20 km F1

x11 x12 x13 . . . x1Nvar

 
R5 R4
 x21 x22 x23 . . . x2Nvar  F3
  Bus 2 40 km F6
=
 . . . ... .   (9)
R6
 . . . ... . 
N N N N Load
x1 pop x2 pop x3 pop . . . xNpop
var

20 km F4

Ci = f (xi1 , xi2 , ....., xiNvar ) where i = 1, 2, ..., Npop (10)

R7 R8

F5 Bus 3

Nsr = N umber of rivers + |{z}

1 (11) Load
sea
Fig. 2: Multi-loop single end distribution network.

NRaindrops = NP op − Nsr (12)

IV. C ASE S TUDY
New position of rivers and stream after one iteration given in To analyse the performances of WCA and LP methods for
(13)-(14). Where rand is a random number between 0 and 1. the earth fault protection optimisation, a sample distribution
C is a constant value also between 1 and 2. Usually offered case which is used for similar problems and shown in Figure
setting is 2 [21]. 2 is selected. The chosen networks has three non-directional
relay and five directional relay as indicated in Figure 2. This
xi+1 i i i
Stream = xStream + rand ∗ C ∗ (xRiver − xStream ) (13) topology named as multi-loop single end distribution network
and widely used in optimization studies [8], [11] and [19].
In order to show the effect of system capacitance which
is, generally, neglected in protection studies, a modification is
xi+1 i i i
River = xRiver + rand ∗ C ∗ (xSea − xRiver ) (14) undertaken in terms of used conductors. It is assumed that the
system consists overhead line, hawk, as a first case. And, It is
Equation (15) and (16) represents evaporation condition which made up solely with underground cables i.e. 240 mm2 single
prevent rapid converge. Where dmax can be considered as core XLPE in flat formation. Overhead line configuration and
tolerance number. Underground cable construction are provided in the Figure 3.
Both configurations are used for coordination analysis.
if |xiRiver − xiRiver |< dmax i = 1, 2, .., Nsr − 1 (15) Six different earth faults are occurred in the middle of the
lines referring Figure 2. Total operating time of the relays
related these six faults should be minimized. Since networks
i+1 dimax have loop topology earth fault seen more than one primer-back
dmax = dimax − (16)
itermax up relay pairs. Table 1 listed relays that sense each earth fault
scenario. LP using GAMS and WCA applied to the Figure 2
Equation (17) and (18) compose ’raining processes’ of the
according to Table 1 to find the optimum operating time for
algorithm. Where LB is the lower boundaries of variable
√ relays.
U B is the upper boundaries of variables. Also , µ is the
Initial population of WCA are selected as 200. Iteration
coefficient.
number is 200. CTI between relays set as 0.3 seconds. Number
of the sea and rivers are selected as 4. WCA solved with
xN ew
Stream = LB + rand ∗ (U B − LB) (17)
MATLAB, LP solved with GAMS. Constraints of optimization
problem are handled with penalty method.
Since analysis carried for earth fault lower and upper
N ew √
XStream = XSea + µ ∗ rand(1, NV ar ) (18) boundaries of TDSs set as 0.1 and 1.2 respectively. Optimal
 sured for the case where the system constructed with overhead
lines. Also three is no coordination constraint violation.
However, capacitive currents circulation during earth faults
1.3 m
hinder selectivity and make traditional constraint equations are
21 m insufficient for the faults at F1 , F2 and F6 listed in Table 3. For
17 m F1 primary relays are R2 and R4 . Back-up protection of R4
is R7 . Thus, R7 should not operate before R4 . For selectivity
also R7 should not operate before R2 . Circulating capacitive
Ground Surface
currents affect current magnitude seen by R7 causing a trip
before R2 .
(a)
False tripping times due to capacitive currents are marked
Ground Surface with different colour in Table 3. It can be seen from Table 3,
R7 trips before R2 for fault at F1 , as similar R7 trips before
R1 for fault at F2 .
1m
R4 Thus, to ensure the selectivity special constrains are de-
R3 70 mm 70 mm signed to prevent false tripping. The equation for providing
R2 selectivity fault at F1 given in (19).
R1 2.083 ∗ T DS7 − 1.188 ∗ T DS2 > 0.01 (19)
(b)
With the help of (19) false trips are prevented between relay
Fig. 3: Geometric structure of a) overhead line b) cable. R2 and relay R7 . Similar logic persistent for other fault path
as well, thus, equation (20) added for fault at F2 . Equation
TABLE I: FAULT SENSED BY RELAYS (21) added for fault at F6 . Selectivity constraints equations
provides selective protection as seen in Table 5.
Relay Name
Fault Location R1 R2 R3 R4 R5 R6 R7 R8
2.083 ∗ T DS7 − 1.188 ∗ T DS1 > 0.01 (20)
F1 X X X X - - X -
F2 X X X - X - X - 1.560 ∗ T DS6 − 1.275 ∗ T DS3 > 0.01 (21)
F3 X X X - - - X -
F4 X X X - - X X - Additional constraints are incorporated into traditional con-
F5 X X X - - X - -
F6 X X X - - X - X
straints equations. Table 4 listed new TDS settings of cable
X Fault sensed by the corresponding relay.
scenarios. Compared with previous solutions CPU time has
- Fault not seen by the corresponding relay. increased both GAMS solver and WCA. It is acceptable since
the total number of equations has increased. Also, additional
constraints cause total operating time to be increased in the
TDS settings of eight relays listed in Table 2. From the Table expence of fully selective protection as indicated Table 5. Both
2 it can be clearly shown that both optimisation technique GAMS and WCA give similar total operating times.
implemented successfully. Total operating time of relays and
TDS settings give similar results. Comparing line types; over- 45
Overhead Line 24
Cable
head lines result in more total operating time than underground Cable with Add. Eq.
22
cable usage. Using obtained TDS setting operating time of 40
20
each relays are listed in Table 3.
Total operating time (s)

For all fault location scenarios, selective protection is en- 35 18

16
30
TABLE II: OPTIMAL TDS SETTINGS 20 40 60

Overhead line Underground cable 25

Relay GAMS WCA GAMS WCA

20
T DS1 0.341 0.341 0.455 0.455
T DS2 0.341 0.341 0.455 0.455
T DS3 0.302 0.305 0.408 0.407 15
T DS4 0.100 0.100 0.100 0.100 0 20 40 60 80 100 120 140 160 180 200
T DS5 0.100 0.100 0.100 0.100 Iteration number
T DS6 0.237 0.236 0.293 0.292
T DS7 0.166 0.168 0.213 0.213 Fig. 4: Converge characteristic of cases.
T DS8 0.100 0.100 0.100 0.100
CPU Time (s) 1.128 1.4128 0.950 1.597 Converge characteristic of WCA shown in Figure 4 for three
Total Operating Time (s) 16.505 16.594 15.557 15.554
different cases. Optimal results are found in approximately 60
 TABLE III: OPERATING TIMES OF RELAYS BOTH OVERHEAD LINE AND CABLE
Overhead line / Hawk Underground cable
Relay R1 R2 R3 R4 R5 R6 R7 R8 R1 R2 R3 R4 R5 R6 R7 R8
F1 0.684 0.498 0.919 0.184 - - 0.509 - 0.695 0.540 0.824 0.144 - - 0.443 -
F2 0.498 0.684 0.919 - 0.184 - 0.509 - 0.540 0.695 0.824 - 0.144 - 0.443 -
F3 0.585 0.585 0.752 - - - 0.407 - 0.608 0.608 0.699 - - - 0.371 -
F4 0.697 0.697 0.697 - - 0.397 0.370 - 0.673 0.673 0.634 - - 0.373 0.334 -
F5 0.791 0.791 0.620 - - 0.440 - - 0.730 0.730 0.594 - - 0.401 - -
F6 0.973 0.973 0.512 - - 0.519 - 0.219 0.846 0.846 0.520 - - 0.457 - 0.156

TABLE IV: OPTIMAL TDS SETTINGS Results produced by WCA are compared with LP imple-
Cable Additional Constraints
mented using GAMS. Similar total operating time have been
obtained overhead lines and underground cables separately.
Relay GAMS WCA
Both overhead lines and underground cables scenarios, system
T DS1 0.585 0.584 is protected selectively and optimally using LP and WCA.
T DS2 0.585 0.584
T DS3 0.534 0.533
T DS4 0.100 0.100 R EFERENCES
T DS5 0.100 0.101
T DS6 0.443 0.442 [1] A. J. Urdaneta, R. Nadira and L. G. Perez Jimenez, “Optimal coordina-
T DS7 0.338 0.338 tion of directional overcurrent relays in interconnected power systems,”
T DS8 0.100 0.100 in IEEE Transactions on Power Delivery, vol. 3, no. 3, pp. 903-911,
CPU Time (s) 1.846 2.364 July 1988, doi: 10.1109/61.193867.
Tot (s) 20.767 20.717 [2] P. P. Bedekar and S. R. Bhide, “Optimum Coordination of Directional
Overcurrent Relays Using the Hybrid GA-NLP Approach,” in IEEE
Transactions on Power Delivery, vol. 26, no. 1, pp. 109-119, Jan. 2011,
TABLE V: RELAY OPERATING TIME AFTER ADDI- doi: 10.1109/TPWRD.2010.2080289.
TIONAL SELECTIVITY [3] H. Zeienldin, E. F. El-Saadany and M. A. Salama, “A novel prob-
lem formulation for directional overcurrent relay coordination,” 2004
Relay Name Large Engineering Systems Conference on Power Engineering (IEEE
Cat. No.04EX819), Halifax, NS, Canada, 2004, pp. 48-52, doi:
Fault R1 R2 R3 R4 R5 R6 R7 R8 10.1109/LESCPE.2004.1356265.
F1 0.893 0.694 1.077 0.144 - - 0.704 - [4] H. Zeineldin, E. F. El-Saadany and M. A. Salama, “Optimal coordination
F2 0.694 0.893 1.077 - 0.146 - 0.704 - of directional overcurrent relay coordination,” IEEE Power Engineering
F3 0.781 0.781 0.916 - - - 0.589 - Society General Meeting, 2005, San Francisco, CA, USA, 2005, pp.
F4 0.865 0.865 0.830 - - 0.565 0.530 - 1101-1106 Vol. 2, doi: 10.1109/PES.2005.1489233.
F5 0.938 0.938 0.777 - - 0.606 - - [5] C.W. So, K.K. Li, “Overcurrent relay coordination by evolutionary
F6 1.087 1.087 0.680 - - 0.690 - 0.156 programming,” Electric Power Systems Research, Volume 53, Issue 2,
2000, Pages 83-90, https://doi.org/10.1016/S0378-7796(99)00052-8.
[6] Prashant P. Bedekar, Sudhir R. Bhide, Optimum coordination of over-
current relay timing using continuous genetic algorithm, Expert Systems
iterations. Also, there is no local optima problem. As seen with Applications, Volume 38, Issue 9, 2011, Pages 11286-11292,
https://doi.org/10.1016/j.eswa.2011.02.177.
in Figure 4 minimum operating time is obtained with cabled [7] T. Amraee, “Coordination of Directional Overcurrent Relays Using
case without additional equations. However, the results do not Seeker Algorithm,” in IEEE Transactions on Power Delivery, vol. 27,
ensure selectivity. no. 3, pp. 1415-1422, July 2012, doi: 10.1109/TPWRD.2012.2190107.
[8] S.S. Gokhale, V.S. Kale, An application of a tent map initiated Chaotic
Firefly algorithm for optimal overcurrent relay coordination, Interna-
V. C ONCLUSION tional Journal of Electrical Power & Energy Systems, Volume 78, 2016,
Along with traditional optimization techniques, meta- Pages 336-342, https://doi.org/10.1016/j.ijepes.2015.11.087.
[9] Astha Chawla, Bhavesh R. Bhalja, Bijaya Ketan Panigrahi and Manohar
heuristic optimization techniques attract the attention for pro- Singh (2018) Gravitational Search Based Algorithm for Optimal Coor-
tection relay coordination studies. WCA is an optimization dination of Directional Overcurrent Relays Using User Defined Char-
algorithm which is widely used for optimal relay coordination acteristic, Electric Power Components and Systems, 46:1, 43-55, DOI:
10.1080/15325008.2018.1431982
studies to select suitable setting of relays and minimum [10] A. Korashy, S. Kamel, A. Youssef and F. Jurado, ”Solving Optimal
operating times of all possible fault cases. Coordination of Direction Overcurrent Relays Problem Using Grey Wolf
This paper present implementation of WCA in a loop type Optimization (GWO) Algorithm,” 2018 Twentieth International Middle
East Power Systems Conference (MEPCON), Cairo, Egypt, 2018, pp.
distribution networks under the earth faults. Since the earth 621-625, doi: 10.1109/MEPCON.2018.8635234.
fault is examined, both overhead lines and underground cables [11] A. Wadood, S. Gholami Farkoush, T. Khurshaid, C.-H. Kim, J. Yu, Z.
modelled taking into account of phase-to earth capacitance. Geem, and S.-B. Rhee, “An Optimized Protection Coordination Scheme
for the Optimal Coordination of Overcurrent Relays Using a Nature-
Capacitive currents of cables which are circulating in the Inspired Root Tree Algorithm,” Applied Sciences, vol. 8, no. 9, p. 1664,
system under the fault conditions may cause loss of selectivity. Sep. 2018.
To maintain selectivity, new constraints are added to the [12] H.H. Zeineldin, E.F. El-Saadany and M.M.A. Salama, Optimal coordina-
tion of overcurrent relays using a modified particle swarm optimization,
optimization problem. Thus, selective protection is ensured Electric Power Systems Research, Volume 76, Issue 11, 2006, Pages
with an increase in total operating time. 988-995,
[13] H.R.E.H. Bouchekara, M. Zellagui, M.A. Abido, Optimal coordination https://doi.org/10.1007/s11831-020-09516-z
of directional overcurrent relays using a modified electromagnetic field [18] Naser El-Naily, Saad M. Saad, Faisal A. Mohamed, “Novel approach
optimization algorithm, Applied Soft Computing, Volume 54, 2017, for optimum coordination of overcurrent relays to enhance microgrid
Pages 267-283, https://doi.org/10.1016/j.asoc.2017.01.037. earth fault protection scheme”, Sustainable Cities and Society, Volume
[14] T. Khurshaid, A. Wadood, S. Gholami Farkoush, J. Yu, C. Kim and S. 54, 2020,https://doi.org/10.1016/j.scs.2019.102006.
Rhee, “An Improved Optimal Solution for the Directional Overcurrent [19] Prashant P. Bedekar, Sudhir R. Bhide and Vijay S. Kale (2010) Optimum
Relays Coordination Using Hybridized Whale Optimization Algorithm Coordination of Overcurrent Relay Timing Using Simplex Method,
in Complex Power Systems,” in IEEE Access, vol. 7, pp. 90418-90435, Electric Power Components and Systems, 38:10, 1175-1193, DOI:
2019, doi: 10.1109/ACCESS.2019.2925822. 10.1080/15325001003652900
[15] Ahmed Korashy, Salah Kamel, Abdel-Raheem Youssef, Francisco Ju- [20] V. N. Rajput and K. S. Pandya, “On 8-bus test system for solving
rado, Modified water cycle algorithm for optimal direction overcurrent challenges in relay coordination,” 2016 IEEE 6th International Confer-
relays coordination, Applied Soft Computing, Volume 74, 2019, Pages ence on Power Systems (ICPS), New Delhi, India, 2016, pp. 1-5, doi:
10-25, https://doi.org/10.1016/j.asoc.2018.10.020. 10.1109/ICPES.2016.7584009
[16] Mahamad Nabab Alam, Biswarup Das, Vinay Pant, A comparative study [21] Gers, J. M., and Holmes, E. J. (2005). Protection of electricity distribu-
of metaheuristic optimization approaches for directional overcurrent tion networks, Vol. 47. IET.
relays coordination, Electric Power Systems Research, Volume 128, [22] Hadi Eskandar, Ali Sadollah, Ardeshir Bahreininejad, Mohd Hamdi,
2015, Pages 39-52, https://doi.org/10.1016/j.epsr.2015.06.018. Water cycle algorithm – A novel metaheuristic optimization method
[17] El-kordy, M., El-fergany, A. & Gawad, A.F.A. Various for solving constrained engineering optimization problems, Com-
Metaheuristic-Based Algorithms for Optimal Relay Coordination: puters & Structures, Volumes 110–111, 2012, Pages 151-166,
Review and Prospective. Arch Computat Methods Eng (2021). https://doi.org/10.1016/j.compstruc.2012.07.010.