Вы находитесь на странице: 1из 6

Available online at www.sciencedirect.


Procedia CIRP 63 (2017) 354 – 359

The 50th CIRP Conference on Manufacturing Systems

Robust Facility Layout Design under Uncertain Product Demands

Shanshan Zha*, Yu Guo, Shaohua Huang, Falin Wang, Xiao Huang
College of Mechanical and Electrical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, 210016, China
* Corresponding author. Tel.: +86-025-84895840; E-mail address: sszha@nuaa.edu.cn


Facility layout problem (FLP) is one of the classical and important problems in manufacturing system. It involves determining the optimal
placement of different types of facilities within the boundaries of workshop so that materials handling cost is optimized. Aiming at dynamic
facility layout problem caused by uncertainty of product demands, a novel continuous formulation of robust facility layout is presented on the
basis of fuzzy-random theory. The main factors causing uncertainty demands are analyzed and the uncertain demands are represented by fuzzy
random variables. To minimize material handling cost between the various departments, a robust layout model with unequal-area departments
is established under fuzzy random environments. Moreover, the position-based flexible particle swarm optimization algorithm is proposed to
obtain feasible optimization solutions. Finally, a case study of manufacturing system has been utilized for the verification of the above model
and optimization.
2017TheTheAuthors. Published
Authors. by Elsevier
Published B.V. This
by Elsevier B.V.is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the scientific committee of The 50th CIRP Conference on Manufacturing Systems.
Peer-review under responsibility of the scientific committee of The 50th CIRP Conference on Manufacturing Systems
Keywords: Robust Facility Layout; uncertain product demands; optimization model

1. Introduction interruption and relocation [4]. Robust approach is not

necessarily an optimal solution for every period of facility
Facility layout problem (FLP) has significant impact layout, but for entire time planning horizon. Unlike the
upon manufacturing costs, work in process, lead time and flexible layout needs rearrangement costs, robust layout has
productivity. FLP involves determining the optimal the advantage of low rearrangements and production
placement of different type of facilities within the space of interruption costs. It was proposed that robustness of a
workshop. Such facilities include machines, workstations, layout as the number of times that the layout falls within a
utilities etc. Material handling cost is the most critical pre-specified percentage of the optimal solution under
measure to determine the reasonability of a layout since it different material demands [5].
forms 20-50% of the total manufacturing costs and it can be
decreased 10-30% by efficient layout design [1,2]. Due to 2. Literature Review
the fact that FLP is an NP-hard problem and combinatorial
optimization problem (COP), various mathematical models Most of the researchers in the literature focused on the
and heuristic or intelligence methods have been developed changes of material demands from period to period under
in past two decades. deterministic environment [6,7]. With fuzzy production
Dynamic facility layout problem (DFLP) takes into information, changes of product volume, and product
account possible changes in the material demands over varieties, material demands for each period exist uncertainty
multiple periods [3]. The planning horizon is generally to some extent. Material demands is considered as a
divided into weeks, months, or years. There are two stochastic variable with the normal distribution, which could
approaches towards DFLP: flexible and robust. Flexible be predicted on the basis of past information relating to
approach assumes that layout will accommodate changes product [8]. Chan and Malmborg (2011) [9] adopted Monte
from time to time with low arrangements and easily Carlo simulation to empirically search for robust solution
relocated machines. It may delays the lead time and make under stochastic demand scenarios. G.Aiello et al. (2012)
customer satisfaction levels decline sharply with production [10] proposed that the uncertainty of material demands

2212-8271 © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the scientific committee of The 50th CIRP Conference on Manufacturing Systems
Shanshan Zha et al. / Procedia CIRP 63 (2017) 354 – 359 355

between facilities and defined the material demands for each material demands in this paper. Essentially, a facility layout
period as fuzzy numbers. Kaveh and Dalfard (2012) [11] plan is drawn up usually before the implementation.
defined product demand (i.e. material flow) as a fuzzy Uncertain fluctuation leads decision-makers difficulty for
number and modeled in fuzzy programming including estimate precise material demands changes in the process of
expected value, chance-constrained programming and production. Therefore, the material demands between
dependent-chance programming. workstations cannot be determined in advance for ambiguity
The first robust layout model was presented as a QAP of renewal information related to production, which leads to
model with stochastic material demands for a single period consider uncertainty. Because of multi-varieties and rapid
[5]. Afterwards, some researchers have established a robust response to market, it is appropriate to use fuzzy variables
layouts model for multi-production scenarios in to model the precise material demands history data and
manufacturing system [12,13]. Pillai and Subbarao (2008) decision-makers’ forecasting data. At the same time,
[14] developed a robust layout design methodology with material demands may to be adjusted from day to day or
equal-area department and adopted total penalty cost to test week by week, which indicate material demands is a
the suitability of the suggested layout model for DFLP. Liu variable in a period that exists probability to some extent.
(2014) [15] improved the existing robust indicator and As mentioned above, probability theory has been applied to
design a robust constrain to improve robustness of final solving FLP in previous researches, in which the material
layout. Neghabi et al. (2015) [16] proposed a robust layout demands approximately follows a normal distribution.
under uncertain environment with dynamic and uncertain Consequently, the material demands between any two
value for departments’ dimensions. Izadinia (2016) [17] departments or workstations at any period need to consider
proposed a mixed integer programming model to generate a both fuzziness and randomness. In a word, a fuzzy random
robust solution for multi-floor layout problem. variable is suggested in this paper to demonstrate the dual
Although previous studies have significantly improved uncertainty in the material demands.
FLP with uncertainty, it is difficult to reflect the subjective
and objective imprecision and complexity simultaneously. 4. Mathematic Modeling for DFLP
Decision-making processes often stay in a hybrid uncertain
environment. So it is hard to know precisely whether the Based on the problem statement above, a dynamic
uncertain environments is fuzzy or random. Therefore, this facility layout model with unequal area under fuzzy random
study proposes fuzzy random variables to handling with the environment established in this paper seeks to find a robust
dual uncertainty, which makes more comprehensive and layout. The model is formulated under the following
practical for facility layout design. assumptions:
(1) There are multi-periods consist of entire planning
3. Key Problem Statement horizon.
(2) The profile of department area is seen as rectangle
As mentioned in the literature, presence of ambiguity in ignoring its outline details.
material flow information and changes of product for each (3) Each facilities are placed parallel to x, y axis.
sections lead the dynamic facility layout problem. Realizing (4) The material demands between departments are
the idealistic of considering material demands as a certain defined as fuzzy random variables.
variable, most researchers consider its uncertainty of the Nomenclature
problem. The approaches solving uncertain facility layout
problems can be classified into two general groups of d tjk manhattan distance between facility j to facility k
probabilistic and fuzzy approaches [3]. d l , d u material demands maximum value and minimum
In practice, FLP decision-maker cannot estimate the
value respectively
precise product demand in advance for rapid changes of
f Dt ^D ,J ` ( x ) increasing functions
market and scarcity of accurate statistical demand data. p

Therefore, it is difficult to confirm precisely whether the g Dt ^D ,J ` ( x ) decreasing functions

environment is fuzzy or random. For the former, fuzzy logic
has been introduced to handle this imprecision in a certain rjt binary variable , denote whether the facility j is to be
range. For the latter, product demands fluctuate from day arranged at period t
to day or week by week with process adjusting, product x j the centroid of facility i at the horizontal coordinate
quality, and so on. As mentioned above, probability theory
y j the centroid of facility i at the vertical coordinate
has been applied to handle this kind uncertainty. Under this
situation, there is need to stress the twofold uncertainty, as it D pt demands of part p in period t
allows for a more comprehensive analysis of the FLP
EV ª Dpt ^D , J `º denotes the expected value of material
product demands uncertainty. ¬ ¼
There are two distinct characteristics considered in this demands at t period
paper. First, unequal-area facilities need to be considered. If Fpt delivery frequency of part p in period t
each facilities have been assigned with uniformly area, it L,W the length and the width of the workshop
will reduce the rate of area utilization. Second, fuzzy
MC1 total material handling costs for flexible model
random variables are applied to replace the deterministic
356 Shanshan Zha et al. / Procedia CIRP 63 (2017) 354 – 359

MC2 total material handling costs for robust model D pt

Fpt (4)
N p number of part p per transportation N pt
R kt relocation cost of facility k at period t
T number of periods in planning horizon
¦D t
t 1
Fp t
TN p

4.1. Mathematical method for fuzzy random environments

Based on the above transforming process, the new model
As mentioned above, the uncertainty demands between of the dynamic approach that has been proposed with
facilities was defined as a fuzzy random variable. The unequal-area under fuzzy random environment is seen as
follows respectively:
introduction of the fuzzy random variable D pt describes
uncertainty of the actual production, which leads model T q q EV ¬ª D p ^D , J `¼º

more difficult to solve. The expected value method first min MC2 ¦¦ ¦ TN p
d jk (6)
transformed the fuzzy random variable into fuzzy variable t 1 k 1 j k 1

and then into deterministic value, so that it can get

deterministic optimal solution. In this paper, the fuzzy In order cater for real production condition and avoid
random variable D pt is converted into trapezoidal fuzzy overlapping between facilities, the constraints should be
consider as follows:
number W pt , where Dpt (dl ,W (d ), d u ) , and W (d ) is
estimated follow a normal distribution. The fuzzy random x j  l j / 2 t bx (7)
variable is transformed into a D , J -level trapezoidal fuzzy
variable Wpt (dl , d , d , du ) , where D is a given possibility x j  l j / 2 d L  bx (8)
level of the fuzzy random variable and J is a given
possibility level of the random variable, both of it represents y j  w j / 2 t by (9)
decision makers’ degree of confidence [18]. Based on the
method of expected-value operator proposed by Kruse and y j  w j / 2 d W  by (10)
Meyer theory and Heilpern theory [19], the fuzzy variable
transformed into a deterministic value EV ª¬ Dpt ^D , J `º¼ , (l j  lk ) / 2  d fx d x j  xk d L  (l j  lk ) / 2  2bx (11)
(l j  lk ) / 2  d fy d y j  yk d W  ( w j  wk ) / 2  2by (12)
d du
EV ª D ^D , J `º
¼ 2 [(d  ³ f D ^D ,J ` ( x )dx )  (d  ³ g D ^D ,J ` ( x )dx )] (1)
¬ p t
Where Eq. (7) and (10) are used to guarantee that each
d l d
facility is assigned within the boundary workshop
respectively. Where Eq. (11) and (12) are used to guarantee
4.2. Formulation for robust facility layout planning that the space for install, operation, and material
transportation between facilities.
As mentioned above, robust approach has more
advantages when the facilities are difficulty to shift, 5. Position-based on particle swarm optimization for
production disruption and rearrangement costs hard to bear. solving Robust Facility Layout Problem
Minimizing the total cost of material handling cost is an
important measure for rationality of layout. Therefore, a Because of easy to implement, efficient global search
typical model with flexible approach and robust approach approach, and few parameters adjustment, many particle
are described as follows respectively: swarm optimization (PSO) and its improved algorithm has
been applied in FLP. However, little research has been
T q q T q found on proposed a suitable PSO in robust facility layout
¦¦ ¦ F d ¦¦ ¦ R

t t t
min MC1 p jk k
(2) problem (RFLP) with uncertain environment. In this paper,
t 1 k 1 j k 1 t 2 k 1 m 1
a new algorithm position-based PSO is proposed, which
system layout planning method (SLP) is introduced to
T q q

min MC2 ¦¦ ¦ F d p
t t
(3) improve the quality of initial solution. The whole procedure
for the position-based PSO for solving a RFLP with unequal
t 1 k 1 j k 1

departmental areas is shown in Fig1 stated as follows:

Subjected to Step1: According to the method described in section 4.1,
transformed the fuzzy random demands into deterministic
demands before algorithm programming.
Shanshan Zha et al. / Procedia CIRP 63 (2017) 354 – 359 357

Step2: Input position-based PSO relative parameters 6. Case study

(position boundary, velocity boundary, popsize, maximum
iterations, constraint parameters in section4.2 and so on). In order to verify effectiveness of proposed model and
Step3: Decode for each particles into solution, a position- algorithm, the case study presents a practical application of
based solution representation is proposed. SLP method help the RFLP in a new assembly workshop of large-scale
acquire w part of higher-quality initial particles. Combined aeronautic equipment. From the view of economical and
with randomly generated n  w part of initial particles safety, the workshop decision maker’s objective were to
compose of initial population n . The three-dimension save material handling costs and prevent overlapping
particles with different position values are used for represent between facilities. In this section, the flexible approach is
candidate solution. The position formulation can be compared with the robust approach, the optimal solution can
represented as follows: be obtained with minimum material handling cost.

I ^( x , y , r ,..., x , y , r ,..., x
1 1
t t
n 1 `
, ynt 1 , rnt1 ) (13)
6.1. Case study representation

Step4: Evaluate the each solution based on the This study attempts to choose 10 facilities to be placed
calculation of fitness value f A according to in workshop, including special equipment, transporting
equipment, large fixture for assisting assembly, deposit area
f A 1/ MCl (14) beside assembly line for storing standard parts, consumable,
attachment and so on. The new assembly workshop had a
Step5: Update the personal best solution and global best large planning area with its length L 220m and its
solution with iteration running. Check if the personal best width W 160m . The entire period of production is make up
solution and global best solution have changed with of 5 different periods. Each period has different product
iteration running or not. If yes, then go to step6 continue demand.
update particles, otherwise, check if particles fall into a Essentially, the aeronautic equipment demands can be
local optimum. acquired by market prediction in planning stage, but
Step 6: Update the particles velocity and position with fluctuate within a certain range. Therefore, the uncertainty
iteration running according to caused by market volatility can be described in fuzzy logic.
In production stage, since the limited type of parts,
p(k  1) p(k )  v(k  1) (15) components, attachment consist the large aeronautic
equipment, the product demands between facilities can be
v H  1 Z v H  c1r1 ( pHpbest  pH )  c2 r2 ( pHgbest  pH ) (16) predicted by history data such as BOM, quality test data, et
al, which follows a normal distribution. Based on the above
statement, the product demands between facilities at the 1st
Then on the basis of step 5 and step6, update the personal period are partly listed in Table1due to the short text. Using
best solution and global best solution. the method introduced in Section4.1, the fuzzy random
Step7: Check the feasibility of the solution acquired. demand was transformed into a deterministic value.
Whether all the particles are obtained finally met the Before the proposed position-based PSO running in the
constraints, if not, then return to step 3, re-initialization. If MATLAB program, the algorithm-related parameters were
yes, output the global best solution. set as follows: popsize K 100 ; maximum iteration H 150 ;
acceleration constant c1 c2 2.

6.2. Model analysis

Based on the proposed models, more computational

results have been obtained with the same position-based
PSO. The results can provide the theoretical optimal plan
for decision maker and layout designer. Further analysis on
the advantages in comparison with above models.
The results (showed in Table2) demonstrated that the
robust model had more advantages than the flexible solution.
For the material handling cost object, the best solution for
flexible approach was higher 4.42% than the robust
approach in this case, the worst and the average solutions
were 3.77% and 4.08% higher, respectively. Moreover, the
flexible solution involves 6 re-arrangement times and costs
in the entire planning horizon, while robust solution
(showed in Table3) suggests only one solution. Re-
Fig.1. The position-based PSO flowchart arrangement may leads production interruption and
358 Shanshan Zha et al. / Procedia CIRP 63 (2017) 354 – 359

increases unnecessary layout costs. Therefore, it can be with minimum material handling costs and improves the
concluded that the RFLP in this case has better solutions dynamic layout adaptability.

Table1. The fuzzy random material demands between any two facilities (1~5) at the first period
Facility index F1 F2 F3 F4 F5
F1 —— (50, W 12 ,70) (42, W 13 ,65) (7, W 14 ,10) (35, W 15 ,46)

W 12 ~N(62,10) W13 ~N(58,8) W14 ~N(8,3) W15 ~N(41,6)

F2 (50, W 21 ,70) —— (10, W 23 ,25) (20, W 24 ,32) (5, W 25 ,10)

W 21 ~N(62,10) W 23 ~N(19,4) W 24 ~N(28,5) W 25 ~N(8,5)

F3 (42, W 31 ,65) (10, W 32 ,25) —— (20, W 34 ,32) 0

W 31 ~N(58,8) W 32 ~N(19,4) W 34 ~N(28,5)

F4 (7, W 41 ,10) (20, W 42 ,32) (20, W 43 ,32) —— 0

W 41 ~N(8,3) W 42 ~N(28,5) W 43 ~N(28,5)

F5 (35, W 51 ,46) (5, W 52 ,10) 0 0 ——

W 51 ~N(41,6) W 52 ~N(8,5)
F6 (5, W 61 ,8) (4, W 62 ,7) (15, W 63 ,25) (6, W 64 ,9) (2, W 65 ,7)

W 61 ~N(6,2) W 62 ~N(6,2) W 63 ~N(19,4) W 64 ~N(7,2) W 65 ~N(4,2)

F7 (4, W 71 ,7) (5, W 72 ,8) (15, W 73 ,25) (20, W 74 ,28) (6, W 75 ,9)

W 71 ~N(6,2) W 72 ~N(6,2) W 73 ~N(19,2) W 74 ~N(23,5) W 75 ~N(7,2)

F8 (90, W 81 ,120) (45, W 82 ,70) (55, W 83 ,75) (25, W 84 ,42) (5, W 85 ,10)

W 81 ~N(110,20) W 82 ~N(60,6) W 83 ~N(65,7) W 84 ~N(35,4) W 85 ~N(35,4)

F9 (120, W 91 ,150) (200, W 92 ,220) (90, W 93 ,120) (60, W 94 ,80) (20, W 95 ,30)

W 91 ~N(142,20) W 92 ~N(217,10) W 93 ~N(117,8) W 94 ~N(75,6) W 95 ~N(25,6)

F10 (7, W101 ,10) (5, W102 ,8) 0 (5, W104 ,8) (4, W105 ,7)

W101 ~N(8,3) W102 ~N(6,3) W104 ~N(6,3) W105 ~N(6,2)

Table2. Model comparison

MHC best result MHC Worst result MHC average result Relocation time
Model type
( 1 u 10 CNY) ( 1 u 10 CNY) ( 1 u 10 CNY)
5 5 5
Flexible 7.46 7.71 7.59 6
Robust 7.13 7. 42 7.28 0
F10 165.21 145.37 r 0
Table 3. Robust model for optimal position value at all periods
Facility index Dimension( x ) Dimension( y) Relocation time
7. Conclusion and Future Work
F1 35.43 30.25 r 0
F2 40.08 105.24 r 0 A new novel robust method was proposed in this paper to
F3 90.65 62.96 r 0 solve DFLP in manufacturing system. First, the main factors
F4 119.28 92.31 r 0 causing uncertainty demands are analyzed and the uncertain
F5 136.18 91.01 r 0 demands are represented on the basis of fuzzy random
F6 78.05 80.79 r 0 theory. Second, a novel continuous formulation of robust
facility layout with un-equal area under uncertain product
F7 101.80 127.52 r 0
demands has been first developed. Subsequently, the
F8 182.36 166.12 r 0 developed model has been verified using a realistic
F9 75.34 102.78 r 0 industrial case study. This study developed a position-based
Shanshan Zha et al. / Procedia CIRP 63 (2017) 354 – 359 359

PSO algorithm to obtain a flexible and robust satisfactory Integrated Manufacturing Systems , 2014, 20 (8) : 1879-1886 (in
solution respectively. Compared with flexible solution, Chinese)
[16] Salmani M H, Eshghi K, Neghabi H. A bi-objective MIP model for
robust solution had advantage of minimum material facility layout problem in uncertain environment. The International
handling cost in all periods, without interrupting production. Journal of Advanced Manufacturing Technology, 2015, 81(9):1563-
For future work, the quantitative index for choosing robust 1575.
approach will be focused and a multi-objective such as [17] Izadinia N, Eshghi K. A robust mathematical model and ACO
minimum actual route distance, area utilization, and so on solution for multi-floor discrete layout problem with uncertain
locations and demands. Computers & Industrial Engineering, 2016,
will be considered. In addition, the three different models 96:237-248.
that include stochastic model, fuzzy model and fuzzy [18] XU J, Song X. Multi-objective dynamic layout problem for temporary
random model will be compared to further elucidating the construction facilities with unequal-area departments under fuzzy
advantages of the fuzzy random environment. random environment.Knowledge-Based Systems, 2015, 81(C): 30-45.
[19] S. Heilpern. The expected value of a fuzzy number.Fuzzy Set

This research was supported by the National Natural

Science Foundation of China (Grant No. 51575274), and
National Defense Basic Scientific Research (Grant


[1] Tompkins J A, White J A, Bozer Y A, et al. Facilities Planning. New

work: John Wiley & Sons, Inc,2003.
[2] Neghabi H, Eshghi K, Salmani M H. A new model for robust facility
layout problem. Information Sciences, 2014, 278:498-509.
[3] Braglia M, Zavanella S Z L. Layout design in dynamic environments:
Strategies and quantitative indices. International Journal of Production
Research, 2003, volume 41(41):995-1016.
[4] Pillai V M, Hunagund I B, Krishnan K K. Design of robust layout for
Dynamic Plant Layout Problems. Computers & Industrial Engineering,
2011, 61(3):813-823.
[5] MEIR J. ROSENBLATT, HAU L. LEE. A robustness approach to
facilities design. International Journal of Production Research, 1987,
[6] Hosseini S, Khaled A A, Vadlamani S. Hybrid imperialist competitive
algorithm, variable neighborhood search, and simulated annealing for
dynamic facility layout problem.Neural Computing & Applications,
2014, 25 (7-8) :1871-1885.
[7] Bozorgi N, Abedzadeh M, Zeinali M. Tabu search heuristic for
efficiency of dynamic facility layout problem.International Journal of
Advanced Manufacturing Technology,2014, 77(1-4):689-703.
[8] Fariborz J, Reza T, Mohammad T. A multi-objective particle swarm
optimisation algorithm for unequal sized dynamic facility layout
problem with pickup/drop-off locations. International Journal of
Production Research, 2012,50(15):4279-4293.
[9] Wai Kin (Victor) Chan, Charles J. Malmborg. A Monte Carlo
simulation based heuristic procedure for solving dynamic line layout
problems for facilities using conventional material handling devices.
International Journal of Production Research, 2010, 48(10):2937-2956.
[10] Aiello G, Scalia G L, Enea M. A multi objective genetic algorithm for
the facility layout problem based upon slicing structure encoding.
Expert Systems with Applications, 2012, 39(12):10352-10358.
[11] Kaveh M, Dalfard V M, Amiri S. A new intelligent algorithm for
dynamic facility layout problem in state of fuzzy constraints. Neural
Computing & Applications, 2014, 24(5):1179-1190.
[12] Benjaafar S, Sheikhzadeh M. Design of flexible plant layouts. IIE
Transactions, 2000, 32(4):309-322.
[13] Aiello G, Enea M.: Fuzzy approach to the robust facility layout in
uncertain production environments. International Journal of Production
Research, 2010, 39(18):4089-4101.
[14] V. Madhusudanan Pillai, Kankata Subbarao. A robust cellular
manufacturing system design for dynamic part population using a
genetic algorithm. International Journal of Production Research, 2008,
[15] LIU QiongˈXU Jinhui, ZHANG Chaoyong. Robust layout of floor
shop based on improved shuffled frog leaping algorithm. Computer