You are on page 1of 10

International Journal of Production Research

ISSN: 0020-7543 (Print) 1366-588X (Online) Journal homepage: http://www.tandfonline.com/loi/tprs20

No-wait two-machine permutation flow shop


scheduling problem with learning effect, common
due date and controllable job processing times

Fu Gao, Mengqi Liu, Jian-Jun Wang & Yuan-Yuan Lu

To cite this article: Fu Gao, Mengqi Liu, Jian-Jun Wang & Yuan-Yuan Lu (2017): No-wait
two-machine permutation flow shop scheduling problem with learning effect, common due
date and controllable job processing times, International Journal of Production Research, DOI:
10.1080/00207543.2017.1371353

To link to this article: http://dx.doi.org/10.1080/00207543.2017.1371353

Published online: 12 Sep 2017.

Submit your article to this journal

Article views: 26

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tprs20

Download by: [Purdue University Libraries] Date: 27 September 2017, At: 02:19
International Journal of Production Research, 2017
https://doi.org/10.1080/00207543.2017.1371353

No-wait two-machine permutation flow shop scheduling problem with learning effect, common
due date and controllable job processing times
Fu Gaoa , Mengqi Liub∗ , Jian-Jun Wanga and Yuan-Yuan Luc
a Faculty of Management and Economics, Dalian University of Technology, Dalian, China; b Business School, Hunan University
Changsha, China; c College of Mathematics, Jilin Normal University, Siping, China
(Received 25 April 2017; accepted 11 August 2017)

We consider a two-machine no-wait permutation flow shop common due date assignment scheduling problem where the
processing time of a job is given as a function of its position in the sequence and its amount of resource allocated to this
Downloaded by [Purdue University Libraries] at 02:19 27 September 2017

job. The common due date (CON) assignment method means that all the jobs are given a common due date. We need to
make a decision on the common due date, resource allocation and the sequence of jobs to minimise total earliness, tardiness,
common due date cost and total resource cost. We show that the problem remains polynomially solvable under the proposed
model.
Keywords: flow shop; learning effect; scheduling; resource allocation; optimization

1. Introduction
For the majority of traditional scheduling models and problems in the literature, the job processing times are fixed and have
constant values. However, we often encounter production systems in which job processing times may be changing due to
the phenomenon of learning effects and/or resource allocation. A more recent survey can be found in Biskup (2008) and
Azzouz, Ennigrou, and Said (2017). Machine scheduling models and problems with learning effects have been paid more
attention in recent years; they include Wang and Wang (2011), Kuo (2012), Janiak, Kovalyov, and Lichtenstein (2013), Sun
et al. (2013), Niu, Wan, and Wang (2015), Lu, Teng, and Feng (2015), Shiau et al. (2015), Wang and Zhang (2015), Wu and
Wang (2016), Wang, Liu, and Wang (2016), Wang, Wei, and Sun (2017b), among others. An extensive survey of research
related to resource allocation (i.e. controllable processing times) scheduling was provided by Shabtay and Steiner (2007).
For more recent studies on this stream of research, the reader may refer to Yang, Lee, and Guo (2013), Mor and Mosheiov
(2014), Yang, Lai, and Yang (2014), Hsieh, Yang, and Yang (2015), Liu, Wang, and Wang (2016).
On the other hand, Wang, Wang, and Wang (2010), Zhu et al. (2011), Lu et al. (2014), Wang and Wang (2014), Liu and
Feng (2014), Wang and Wang (2015), and Li et al. (2015) considered scheduling problems with learning effects and resource
allocation at the same time. Wang, Wang, and Wang (2010) considered
  the following scheduling model: pjr (uj ) = p̄j r a −b j uj
p̄ r a m
(i.e. a linear resource allocation function) and pjr (uj ) = uj j , uj > 0 (a convex resource allocation function), where
p̄j is the normal processing time of job J j , a ≤ 0 is a learning index, r is the position of job J j is scheduled in a sequence,
uj is the amount of resource that can be allocated to job J j , and m > 0 is a positive constant. Zhu et al. (2011) considered
single-machine scheduling with resource allocation and a learning effect under the group technology  aj environment. Lu et
p̄j r m
al. (2014) considered the following scheduling model: pjr (uj ) = p̄j r − bj uj , and pjr (uj ) =
a j
uj , uj > 0, where
a j ≤ 0 is the job-dependent learning factor of job J j . For two due date assignment (Gordon, Proth, and Chu 2002a, 2002b)
methods (i.e. the common (CON, Panwalker, Smith, and Seidmann 1982) due date, and the slack (SLK) due date), they
provided a polynomial-time algorithm to minimise an integrated objective function, which includes earliness, tardiness, due
date assignment, and total resource consumption costs. Wang and Wang (2014) considered the same scheduling models as
Lu et al. (2014), for a common due-window assignment scheduling problem, they proved that the problem can be solved in
polynomial time. Liu and Feng (2014) considered the two-machine no-wait flowshop scheduling with learning effects and
resource allocation. Li et al. (2015) considered the same scheduling models as Lu et al. (2014), for a due-window assignment
scheduling problem based on a common flow allowance (i.e. all jobs have a slack due window (SLKW)), they proved that the
problem can be solved in polynomial time. Wang and Wang (2015) considered the same scheduling models as Lu et al. (2014),
for three due-date assignment methods (i.e. the common (CON) due date, the slack (SLK) due date, and the unrestricted (DIF)

∗ Corresponding author. Email: liumengqi163@126.com

© 2017 Informa UK Limited, trading as Taylor & Francis Group


2 F. Gao et al.

due date), they proved that the problem can be solved in polynomial time. Li et al. (2015) considered the same scheduling
models as Wang and Wang (2014), for a slack due-window assignment based on the common flow allowance scheduling
problem, they proved that the problem can be solved in polynomial time. Wang et al. (2017a) considered single-machine
resource allocation scheduling with truncated job-dependent learning and deterioration effects.
On the other hand, scheduling problems with the due dates’ assignment have received considerable attention by many
researchers (Panwalker, Smith, and Seidmann 1982; Gordon, Proth, and Chu 2002a, 2002b; Dolgui, Gordon, and Strusevich
2012; Gordon, Strusevich, and Dolgui 2012; Ji et al. 2014; Liu et al. 2017). Hence, in this paper, we consider resource
allocation and the common due date assignment scheduling problem with learning effects in the two-machine no-wait
permutation flow shop environment. The objective is to find a schedule, a common due date and resource allocation that
minimises the total earliness, tardiness, common due date cost and total resource cost. The rest of the paper is organised as
follows: In the next section we introduce and formulate the problem. In Section 3 we provide some properties to optimally
solve the problem. In Section 4, we consider a special case of the common due date assignment problem. We conclude the
paper in Section 5.
Downloaded by [Purdue University Libraries] at 02:19 27 September 2017

2. Problem description
To facilitate the problem formulation, we introduce the following notation, which are used throughout the paper:
J = {J1 , J2 , . . . , Jn }
a set of n jobs;
[ j]
the job in the jth position in any sequence;
the machine i, i = 1, 2;
Mi
the operation of job J j on machine Mi , i = 1, 2 and j = 1, 2, . . . , n;
O ji
the basic (normal) processing time of operation O ji (i.e. the processing time without any resource
p̄ji
allocation and learning effect);
pji the actual processing time of operation O ji ;
C ji the completion time of job J j on machine Mi ;
C j = C j2 the completion time of job J j ;
d the common due date for all jobs;
E j = max{0, d − C j } the earliness of job J j ;
T j = max{0, C j − d} the tardiness of job J j ;
The problem under study can be formally described as follows: We are given a set of independent jobs J = {J1 , J2 , . . . , Jn }
to be processed on a two-machine no-wait permutation flow shop, i.e. each job must be processed first on machine M1 and
then on machine M2 , and the jobs are not allowed to wait between the two machines (i.e. with a constraint C[ j+1]1 ≥ C[ j]2 ).
All the jobs are simultaneously available at time zero and job preemption is not allowed.
As in Wang, Wang, and Wang (2010) and Liu and Feng (2014), we consider the following model
 
pji r a m
pji = , (1)
uji
where pji is the basic (normal) processing time of operation O ji , uji is the amount of a non-renewable resource allocated
to operation O ji and m is a positive constant. The problem is to determine the common due date d, the optimal resource
allocations u and the optimal sequence π of jobs so that the following cost function is optimal:

n 
2 
n
Z (π, u, d) = (α E j + βT j + γ d) + θ Gji uji , (2)
j=1 i=1 j=1

where weights α ≥ 0, β ≥ 0, γ ≥ 0, θ ≥ 0 are given constants, and Gji is the cost associated with the resource allocation
per time unit. Using
 the extended three-field
 a notation
 scheme (Graham et al. 1979) for scheduling problems, the problem can
 pji r m  n 2 n
be denoted as F2 no − wait, pji = uji  j=1 (α E j + βT j + γ d) + θ i=1 j=1 Gji uji .

3. Common due date assignment problem


  a m  
 p r  n
First, we give some lemmas of an optimal schedule for the problem F2 no − wait, pji = ujiji  j=1 (α E j + βT j +
2 n
γ d) + θ i=1 j=1 Gji uji . As in Liu and Feng (2014), for the no-wait permutation flowshop scheduling, we have p[ j]2 =
p[ j+1]1 .
International Journal of Production Research 3

Lemma 1 For any specified schedule π , there exists an optimal common due date with the property that the common due
date value d coincides with the completion time of a job.
Proof By contradiction, suppose that an optimal schedule exists such that C[k] < d < C[k+1] .
(1) If C[k] < d < C[k+1] , then the total cost is given by

k 
n 
n
Z =α (d − C[ j] ) + β (C[ j] − d) + γ nd + θ G j uj .
j=1 j=k+1 j=1

(2) If d = C[k] , then the total cost is given by



k−1 
n 
n
Z1 = α (C[k] − C[ j] ) + β (C[ j] − C[k] ) + γ nC[k] + θ G j uj .
j=1 j=k+1 j=1

(3) If d = C[k+1] , then the total cost is given by


Downloaded by [Purdue University Libraries] at 02:19 27 September 2017


k 
n 
n
Z2 = α (C[k+1] − C[ j] ) + β (C[ j] − C[k+1] ) + γ nC[k+1] + θ G j uj .
j=1 j=k+2 j=1

Let x = d − C[k] , y = C[k+1] − d, then


Z − Z 1 = x(αk − β(n − k) + γ n),
Z − Z 2 = −y(αk − β(n − k) + γ n).
Obviously, if αk − β(n − k) + γ n ≥ 0, then Z ≥ Z 1 , if αk − β(n − k) + γ n ≤ 0, then Z ≥ Z 2 . Hence, an optimal
schedule exists in which d coincides with the completion time of some job. 
  a m    
 p r  n 2 n
Lemma 2 For the F2 no − wait, pji = ujiji  j=1 (α E j + βT j + γ d) + θ i=1 j=1 Gji uji problem, an optimal
sequence exists such that d = C[k] , where


(β − γ )n
k = min max ,0 ,n . (3)
α+β
Proof Consider an optimal schedule such that d = C[k] for some job J
[k] . Using
2 n
the classical small perturbation technique,
we measure the change in the total cost nj=1 (α E j + βT j + γ d) + θ i=1 j=1 G ji uji when moving d = C [k] .
(1) If d = C[k] , then the total cost is given by

k−1 
n 
2 
n
Z =α (C[k] − C[ j] ) + β (C[ j] − C[k] ) + γ nC[k] + θ Gji uji .
j=1 j=k+1 i=1 j=1

(2) If d = C[k] − x, then the total cost is given by



k−1 
n 
2 
n
Z1 = α (C[k] − x − C[ j] ) + β (C[ j] − C[k] + x) + γ n(C[k] − x) + θ Gji uji .
j=1 j=k i=1 j=1

(3) If d = C[k] + y, then the total cost is given by



k 
n 
2 
n
Z2 = α (C[k] + y − C[ j] ) + β (C[ j] − C[k] − y) + γ n(C[k] + y) + θ Gji uji .
j=1 j=k+1 i=1 j=1

Obviously,
Z − Z 1 = α(k − 1)x − β(n − k + 1)x + γ nx ≤ 0,
Z − Z 2 = −αky + β(n − k)y − γ ny ≤ 0,
then
(β − γ )n (β − γ )n
≤k ≤1+
α+β α+β
4 F. Gao et al.
  
(β−γ )n
From above analysis and since k must be a non-negative integer, we have k = max α+β , 0 . In addition k ≤ n,
   
hence k = min max (β−γ α+β
)n
,0 ,n . 
From Lemmas 1 and 2, we have


n 
2 
n
Z (π, u, d) = (α E j + βT j + γ d) + θ Gji uji
j=1 i=1 j=1

k−1 
n 
2 
n
=α (C[k] − C[ j] ) + β (C[ j] − C[k] ) + γ nC[k] + θ Gji uji
j=1 j=k+1 i=1 j=1
⎡ ⎤
k 
k−1
= α ⎣ ( j − 1) p[ j]1 + (k − 1) p[k]2 − p[ j]2 ⎦
j=2 j=1
⎡ ⎤

n 
n
Downloaded by [Purdue University Libraries] at 02:19 27 September 2017

+β ⎣ (n − j + 1) p[ j]1 − (n − k) p[k]2 + p[ j]2 ⎦


j=k+1 j=k+1
⎡ ⎤ ⎡ ⎤

k 
n 
n
+γ ⎣n p[ j]1 + np[k]2 ⎦ + θ ⎣ G[ j]1 u[ j]1 + G[ j]2 u[ j]2 ⎦
j=1 j=1 j=1

n   
n 
n  m 
n
p[ j]1 ja m p[ j]2 j a
= ωj +θ G[ j]1 u[ j]1 + νj +θ G[ j]2 u[ j]2 , (4)
u[ j]1 u[ j]2
j=1 j=1 j=1 j=1

where

⎨ nγ , if j = 1,
ω j = α( j − 1) + γ n, if j = 2, . . . , k, (5)

β(n − j + 1), if j = k + 1, . . . , n,

and

⎨ −α, if j = 1, . . . , k − 1,
νj = α(k − 1) − β(n − k) + γ n, if j = k, (6)

β, if j = k + 1, . . . , n.

Lemma 3 For a given
 sequence
  π = (J[1] , J[2] , . . . , J[n] ), the optimal resource allocation u ∗ (π ) for the problem
 pji r a m  n 2 n
F2 no − wait, pji = uji  j=1 (α E j + βT j + γ d) + θ i=1 j=1 Gji uji is:

  1
mω1 m+1 m
u ∗[1]1 = ( p[1]1 ) m+1 (7)
θ G[1]1
  1
mνn m+1 m
u ∗[n]2 = ( p[n]2 n a ) m+1 (8)
θ G[n]2
  1
m(ω j + νj−1 ) m+1 m
u ∗[j]1 = ( p[ j]1 j a ) m+1 , j = 2, 3, . . . , n (9)
θ G[ j]1
  1
m(ω j+1 + νj ) m+1 m
u ∗[j]2 = ( p[ j]2 j a ) m+1 , j = 1, 2, . . . , n − 1. (10)
θ G[ j]2
Proof From (4), we have:


n  m 
n 
n  m 
n
p[ j]1 j a p[ j]2 j a
Z (π, u, d) = ωj +θ G[ j]1 u [ j]1 + νj +θ G[ j]2 u [ j]2
u [ j]1 u [ j]2
j=1 j=1 j=1 j=1
= Z 11 (π, u, d) + Z n2 (π, u, d) + Z ji (π, u, d),
International Journal of Production Research 5

where
 
p [1]1 m
Z 11 (π, u, d) = ω1 + θ G[1]1 u [1]1 , (11)
u [1]1
 
p [n]2 n a m
Z n2 (π, u, d) = νn + θ G[n]2 u [n]2 . (12)
u [n]2
n    n 
n−1   
n−1
p [ j]1 j a m p[ j]2 j a m
Z ji (π, u, d) = ωj +θ G[ j]1 u [ j]1 + νj +θ G[ j]2 u [ j]2 . (13)
u [ j]1 u [ j]2
j=2 j=2 j=1 j=1

Since each of the objectives is a convex function, differentiating (11) and (12) with respect to u [1]1 and u [n]2 ,respectively,
the necessary and sufficient conditions for an optimal solution are
 m
∂ Z 11 (π, u, d) p [1]1
= θ G[1]1 − mω1 × = 0, (14)
∂u [1]1 (u [1]1 )m+1
  m
Downloaded by [Purdue University Libraries] at 02:19 27 September 2017

∂ Z n2 (π, u, d) p [n]2 n a
= θ G[n]2 − mνj × = 0. (15)
∂u [n]2 (u [n]2 )m+1

From (14) and (15), we have (7) and (8).


By using C[ j+1]1 ≥ C[ j]2 ( j = 1, 2, . . . , n − 1), we have


n   
n 
n−1
 
a m 1 1
Z ji (π, u, d) = (ω j + νj−1 ) p[ j]1 j + +θ G[ j]1 u[ j]1 + θ G[ j]2 u[ j]2 . (16)
(u [ j]1 )m (u [ j]2 )m
j=2 j=2 j=1

Similarly, we have
 m
∂ Z ji (π, u, d) p [ j]1 j a
= θ G[ j]1 − m(ω j + νj−1 ) × = 0, ∀ j = 2, 3, . . . , n, (17)
∂u [ j]1 (u [ j]1 )m+1
 m
∂ Z ji (π, u, d) p [ j]1 j a
= θ G[ j]2 − m(ω j + νj−1 ) × = 0, ∀ j = 1, 2, . . . , n − 1. (18)
∂u [ j]2 (u [ j]2 )m+1
  1
m(ω j +νj−1 ) m
From (17), we have u ∗[j]1 =
m+1
θ G[ j]1 ( p [ j]1 j a ) m+1 , j = 2, 3, . . . , n. From (18) and C[ j+1]1 ≥ C[ j]2 ( j =
  1
m(ω j+1 +νj ) m
1, 2, . . . , n − 1), we have u ∗[j]2 =
m+1
θ G[ j]2 ( p[ j]2 j a ) m+1 , j = 1, 2, . . . , n − 1. 
n
Substituting (7)–(10) into (4), we obtain that a new unified expression for Z (π, u, d) = j=1 (α E j + βT j + γ d) +
2  n
θ i=1 j=1 Gji u ji under an optimal resource allocation and as a function of π is

Z (π, u ∗ (π ), d)
m
 m 1
 1   m
= θ m+1 m − m+1 + m m+1 (ω1 ) m+1 G[1]1 p [1]1 m+1
m
 m 1
 1 ma   m
+θ m+1 m − m+1 + m m+1 (νn ) m+1 n m+1 G[n]2 p [n]2 m+1
m
 m 1

+θ m+1 m − m+1 + m m+1
⎡ ⎤
n
  1 ma   m 
n−1
  1 ma   m
×⎣ ω j + νj−1 m+1 j m+1 G[ j]1 p [ j]1 m+1 + ω j+1 + νj m+1 j m+1 G[ j]2 p [ j]2 m+1 ⎦ . (19)
j=2 j=1
  a m  
 p r  n
In order to obtain the optimal job sequence, we formulate the problem F2 no − wait, pji = ujiji  j=1 (α E j +
2 n
βT j + γ d) + θ i=1 j=1 Gji u ji as a linear assignment problem. Let x jr be a 0/1 variable such that x jr = 1 if
job  J j ( j = 1, 2, . . ., n) is scheduled
 at position r (r = 1, 2, . . . , n), and x jr = 0, otherwise. Then, the problem
 pji r a m  n 2  n
F2 no − wait, pji = u ji  j=1 (α E j + βT j + γ d) + θ i=1 j=1 Gji u ji can be solved by the following linear
assignment problem:
6 F. Gao et al.

n 
n
m
 m 1

LAP: Min θ m+1 m − m+1 + m m+1 λ jr x jr (20)
r =1 j=1
st

n
x jr = 1, j = 1, 2, . . . , n, (21)
r =1
 n
x jr = 1, r = 1, 2, . . . , n, (22)
j=1
x jr = 0 or 1, j = 1, 2, . . . , n, r = 1, 2, . . . , n, (23)
where
1 ma   m 1 ma   m
λ jr = (ηr ) m+1 r m+1 Gj1 pj1 m+1 + (ϑr ) m+1 r m+1 Gj2 pj2 m+1 , (24)

ω1 , if r = 1,
ηr =
Downloaded by [Purdue University Libraries] at 02:19 27 September 2017

(25)
ωr + νr −1 , if r = 2, 3, . . . , n,
and
ωr +1 + νr , if r = 1, 2, . . . , n − 1
ϑr = (26)
νn , if r = n.
Once the assignment problem has been solved, we will obtain an optimal matching job with positions (i.e. an optimal
schedule).   a m  
 p r  n
Based on the above analysis, we can determine the optimal solution for F2 no − wait, pji = ujiji  j=1 (α E j +
2 n
βT j + γ d) + θ i=1 j=1 Gji u ji by the following algorithm:
Algorithm 1
   
By Lemma 2, calculate k = min max (β−γ
(Step 1). α+β
)n
,0 ,n .
Calculate the values λ jr by using (24).
(Step 2).
(Step 3).
Solve the linear assignment problem LAP (20)–(23) to determine the optimal job sequence.
(Step 4).
Calculate the optimal resource allocation by using Lemma 3.
(Step 5).
Calculate the actual processing times by using (1).
Calculate the due date d = C[k] .
(Step 6).
  a m   2 n
 p r  n
Theorem 1 For the F2 no − wait, pji = ujiji  j=1 (α E j +βT j +γ d)+θ i=1 j=1 Gji u ji problem, an optimal
solution can be obtained by Algorithm 1 in O(n 3 ) time.
Proof The correctness of Algorithm 1 follows from Lemmas 1–3 and the linear assignment problem LAP (20)–(23). The
time complexity of (Step 2) is O(n 2 ) and (Step 3) is O(n 3 ) time. Steps 1, 4, 5 and 6 can be performed in linear time. Thus
the overall computational complexity of Algorithm 1 is O(n 3 ). 
  a m   2 n
 pji r  n
In order to illustrate Algorithm 1 for F2 no − wait, pji = u ji  j=1 (α E j + βT j + γ d) + θ i=1 j=1 Gji u ji ,
we present the following instance.
Example 1 Consider n =7, m = 2, a = −0.322, α = 7, β = 17, γ = 5, θ = 1 and the other corresponding parameters
shown in Table 1.
   
Solution: From Algorithm 1, k = min max (β−γ α+β
)n
, 0 , n = 4, the optimal schedule is [J1 , J3 , J2 , J5 , J4 , J7 , J6 ]
(The values ωr , νr , ηr , ϑr and λ jr are given in Tables 2 and 3), the optimal resource allocations are u 11 = 8.2426, u 31 =
9.8499, u 21 = 6.6801, u 51 = 8.7455, u 41 = 14.7635, u 71 = 13.7611, u 61 = 4.1087, u 12 =  2.8575, u 32 = 8.7587, u 22 =
4.7722, u 52 = 5.2749, u 42 = 6.8797, u 72 = 2.8896, u 62 = 7.7624, d = 1.7224, and nj=1 (α E j + βT j + γ d) +
2  n
θ i=1 j=1 G ji u ji = 598.8596.

4. A special case
n
n we consider a special case, i.e. Gj1 pj1 = Gj2 pj2 = Gj pj ( j = 1, 2, . . . , n). From (19), j=1 (α E j + βT j +
In this section,
γ d) + θ j=1 Gj u j can be expressed as:
International Journal of Production Research 7

Table 1. The data of Example 1.

Jj J1 J2 J3 J4 J5 J6 J7

pj1 4 6 8 9 10 5 18
pj2 1 3 5 6 4 17 3
G j1 2 5 3 1 6 7 4
G j2 3 4 2 4 5 6 8

Table 2. The weights of Example 1.

r 1 2 3 4 5 6 7

ωr 35 42 49 56 51 34 17
νr –7 –7 –7 5 17 17 17
ηr 35 35 42 49 56 51 34
Downloaded by [Purdue University Libraries] at 02:19 27 September 2017

ϑr 35 42 49 56 51 34 17

Table 3. The λ jr values of Example 1.

j\r 1 2 3 4 5 6 7

λ jr = 1 19.8884 17.5061 16.9955 16.7794 16.2933 14.6844 12.0559


2 48.7270 42.9161 41.6606 41.1280 39.9069 35.9307 29.4735
3 42.3993 37.3572 36.2622 35.7971 34.7179 31.2390 25.6105
4 41.3695 37.1195 35.9303 35.3965 33.5594 29.2702 23.3197
5 74.2344 65.2725 63.3795 62.5811 60.8485 54.9366 45.1737
6 106.4095 95.5537 92.4812 91.0992 86.2851 75.1512 59.7934
7 83.8287 73.7085 71.5708 70.6693 68.7128 62.0367 51.0121

Note: The bold numbers are the optimal solution.

Z (π, u ∗ (π ), d)
 n 
  
m m 1   1   1 ma   m
= θ m+1 m − m+1 + m m+1 ηj m+1 + ϑ j m+1 j m+1 G[ j] p[ j] m+1
j=1

m
 m 1

n
=θ m+1 m − m+1 + m m+1 W j P[ j] , (27)
j=1
 
  1   1 ma   m
where W j = ηj m+1 + ϑj m+1 j m+1 and P[ j] = G[ j] p[ j] m+1 .
  
n   1   1 ma   m 
Obviously, the term j=1 ηj m+1 + ϑj m+1 j m+1 G[ j] p[ j] m+1 = nj=1 W j P[ j] can be minimized by the HLP

rule (that is ‘The sum of products nj=1 x j y j is minimized if sequence x1 , x2 , . . . , xn is ordered nondecreasingly and sequence
y1 , y2 , . . . , yn is ordered nonincreasingly or vice versa, and it is maximized if the sequences are ordered in the same way’,
Hardy, Littlewood, and Polya (1967)) in O(n log n) time.  Hence   for the special case (i.e. Gj1 pj1= Gj2 pj2 = Gj pj for
 pji r a m  n n
j = 1, 2, . . . , n) of the problem F2 no − wait, pji = uji  j=1 (α E j + βT j + γ d) + θ j=1 Gj uj , the optimal
solution can be solved by the following algorithm:
8 F. Gao et al.

Algorithm 2
   
(Step 1). By Lemma 2, calculate k = min max (β−γα+β
)n
,0 ,n .
 
  1   1 ma
(Step 2). Calculate ηj m+1 + ϑ j m+1 j m+1 for j = 1, 2, . . . , n.
(Step 3). Assign the job with the longest P j to the position with the smallest value of W j , the job with the second longest
P j to the position with the second smallest value of W j , etc. Then, obtain an optimal job sequence and the total
cost.
(Step 4). Calculate the optimal resource allocation by using Lemma 3.
(Step 5). Calculate the processing times by using (1).
(Step 6). Calculate the due date d = C[k] .


Theorem 2 For the special case (i.e. Gj1 pj1 = Gj2 pj2 = Gj pj for j = 1, 2, . . . , n) of the problem F2no − wait, pji
 a m   n
p r  n
= ujiji  j=1 (α E j + βT j + γ d) + θ j=1 Gj uj , an optimal solution can be obtained by Algorithm 2 in O(n log n)
time.
Downloaded by [Purdue University Libraries] at 02:19 27 September 2017

5. Conclusions
In this study, the learning effect and controllable processing times are taken into consideration for the two-machine no-wait
permutation flow shop scheduling problem. For a common due date (CON) assignment method, our objective function is
to minimize total earliness, tardiness, common due date cost and total resource cost, we proved that the problem can be
solved in polynomial time. Further research may explore more realistic settings such as the maintenance activity (two-agent
scheduling, flexible flow shop, due window assignment, parallel machines and job rejection) and optimize other performance
measures.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This research was supported by the National Natural Science Foundation of China [grant number 71672019], [grant number 71271039],
[grant number 71471057], [grant number 71501082]; the New Century Excellent Talents in University [NCET-13-0082].

References

Azzouz, A., M. Ennigrou, and L. B. Said. 2017. “Scheduling Problems under Learning Effects: Classification and Cartography.”
International Journal of Production Research. doi:10.1080/00207543.2017.1355576.
Biskup, D. 2008. “A State-of-the-art Review on Scheduling with Learning Effects.” European Journal of Operational Research 188:
315–329.
Dolgui, A., V. Gordon, and V. Strusevich. 2012. “Single Machine Scheduling with Precedence Constraints and Positionally Dependent
Processing Times.” Computers and Operations Research 39: 1218–1224.
Gordon, V. S., J. M. Proth, and C. B. Chu. 2002a. “A Survey of the State-of-the-art of Common due Date Assignment and Scheduling
Research.” European Journal of Operational Research 139: 1–25.
Gordon, V. S., J. M. Proth, and C. B. Chu. 2002b. “Due Date Assignment and Scheduling: SLK, TWK and Other Due Date Assignment
Models.” Production Planning and Control 13: 117–132.
Gordon, V., V. Strusevich, and A. Dolgui. 2012. “Scheduling with Due Date Assignment under Special Conditions on Job Processing.”
Journal of Scheduling 15: 447–456.
Graham, R. L., E. L. Lawler, J. K. Lenstra, and A. H. G. Rinnooy Kan. 1979. “Optimization and Approximation in Deterministic Sequencing
and Scheduling: A Survey.” Annals of Discrete Mathematics 5: 287–326.
Hardy, G. H., J. E. Littlewood, and G. Polya. 1967. Inequalities. Cambridge: Cambridge University Press.
Hsieh, P.-H., S.-J. Yang, and D.-L. Yang. 2015. “Decision Support for Unrelated Parallel Machine Scheduling with Discrete Controllable
Processing Times.” Applied Soft Computing 30: 475–483.
Janiak, A., M. Y. Kovalyov, and M. Lichtenstein. 2013. “Strong NP-hardness of Scheduling Problems with Learning or Aging Effect.”
Annals of Operations Research 206: 577–583.
Ji, P., G. Li, Y. Huo, and J.-B. Wang. 2014. “Single-machine Common Flow Allowance Scheduling with Job-dependent Aging Effects and
a Deteriorating Maintenance Activity.” Optimization Letters 8: 1389–1400.
International Journal of Production Research 9

Kuo, W.-H. 2012. “Single-Machine Group Scheduling with Time-dependent Learning Effect and Position-based Setup Time Learning
Effect.” Annals of Operations Research 196: 349–359.
Li, G., M.-L. Luo, W.-J. Zhang, and X.-Y. Wang. 2015. “Single-machine Due-window Assignment Scheduling based on Common Flow
Allowance, Learning Effect and Resource Allocation.” International Journal of Production Research 53 (4): 1228–1241.
Liu, Y. and Z. Feng. 2014. “Two-machine no-wait Flowshop Scheduling with Learning Effect and Convex Resource-dependent Processing
Times.” Computers & Industrial Engineering 75: 170–175.
Liu, L., J.-J. Wang, and X.-Y. Wang. 2016. “Due-window Assignment Scheduling with Resource Processing Times to Minimise Total
Resource Consumption Cost.” International Journal of Production Research 54: 1186–1195.
Liu, M., S. Wang, F. Zheng, and C. Chu. 2017. “Algorithms for the Joint Multitasking Scheduling and Common Due Date Assignment
Problem.” International Journal of Production Research 55: 6052–6066.
Lu, Y.-Y., G. Li, Y.-B. Wu, and P. Ji. 2014. “Optimal Due-dateAssignment Problem with Learning Effect and Resource-dependent Processing
Times.” Optimization Letters 8: 113–127.
Lu, Y.-Y., F. Teng, and Z.-X. Feng. 2015. “Scheduling Jobs with Truncated Exponential Sum-of-logarithm-processing-times based and
Position-based Learning Effects.” Asia-Pacific Journal of Operational Research 32 (3): 1550026 (20p).
Mor, B. and G. Mosheiov. 2014. “Batch Scheduling of Identical Jobs with Controllable Processing Times.” Computers & Operations
Research 41: 115–124.
Niu, Y.-P., L. Wan, and J.-B. Wang. 2015. “A note on Scheduling Jobs with Extended Sum-of-processing-times-based and Position-based
Downloaded by [Purdue University Libraries] at 02:19 27 September 2017

Learning Effect.” Asia-Pacific Journal of Operational Research 32 (2): 1550001 (18p).


Panwalker, S. S., M. L. Smith, and A. Seidmann. 1982. “Common Due-date Assignment to Minimize Total Penalty for the one Machine
Scheduling Problem.” Operations Research 30: 391–399.
Shabtay, D. and G. Steiner. 2007. “A Survey of Scheduling with Controllable Processing Times.” Discrete Applied Mathematics 155 (13):
1643–1666.
Shiau, Y.-R., M.-S. Tsai, W.-C. Lee, and T. C. E. Cheng. 2015. “Two-agent Two-machine Flowshop Scheduling with Learning Effects to
Minimize the Total Completion Time.” Computers & Industrial Engineering 87: 580–589.
Sun, L.-H., K. Cui, J.-H. Chen, J. Wang, and X.-C. He. 2013. “Some Results of the Worst-case Analysis for Flow Shop Scheduling with a
Learning Effect.” Annals of Operations Research 211: 481–490.
Wang , D., Wang , M.-Z., and Wang , J.-B.. 2010. “Single-machine Scheduling with Learning Effect and Resource-dependent Processing
Times.” Computers & Industrial Engineering 59: 458–462.
Wang, J.-B., F. Liu, and J.-J. Wang. 2016. “Research on m-machine Flow Shop Scheduling with Truncated Learning Effects.” International
Transactions in Operational Research. doi:10.1111/itor.12323.
Wang, J.-B., M. Liu, N. Yin, and P. Ji. 2017a. “Scheduling Jobs with Controllable Processing Time, Truncated Job-dependent Learning
and Deterioration Effects.” Journal of Industrial and Management Optimization 13 (2): 1025–1039.
Wang, J.-B. and M.-Z. Wang. 2011. “Worst-case Behavior of Simple Sequencing Rules in Flow Shop Scheduling with General Position-
dependent Learning Effects.” Annals of Operations Research 191: 155–169.
Wang, J.-B. and M.-Z. Wang. 2014. “Single-machine Due-window Assignment and Scheduling with Learning Effect and Resource-
dependent Processing Times.” Asia-Pacific Journal of Operational Research 31 (5): 1450036 (28p).
Wang, J.-B. and J.-J. Wang. 2015. “Research on Scheduling with Job-dependent Learning Effect and Convex Resource Dependent
Processing Times.” International Journal of Production Research 53: 5826–5836.
Wang, Z., C.-M. Wei, and L. Sun. 2017b. “Solution Algorithms for the Number of Tardy Jobs Minimisation Scheduling with a Time-
Dependent Learning Effect.” International Journal of Production Research 55: 3141–3148.
Wang, J.-J. and B.-H. Zhang. 2015. “Permutation Flowshop Problems with bi-criterion Makespan and Total Completion Time Objective
and Position-weighted Learning Effects.” Computers & Operations Research 58: 24–31.
Wu, Y.-B. and J.-J. Wang. 2016. “Single-machine Scheduling with Truncated Sum-of-processing-times-based Learning Effect Including
Proportional Delivery Times.” Neural Computing and Applications 27: 937–943.
Yang, D.-L., C.-J. Lai, and S.-J. Yang. 2014. “Scheduling Problems with Multiple due Windows Assignment and Controllable Processing
Times on a Single Machine.” International Journal of Production Economics 150: 96–103.
Yang, S.-J., H.-T. Lee, and J.-Y. Guo. 2013. “Multiple Common Due Dates Assignment and Scheduling Problems with Resource Allocation
and General Position-dependent Deterioration Effect.” The International Journal of Advanced Manufacturing Technology 67: 181–188.
Zhu, Z., L. Sun, F. Chu, and M. Liu. 2011. “Single-machine Group Scheduling with Resource Allocation and Learning Effect.” Computers
& Industrial Engineering 60: 148–157.

You might also like