MPE Mathematical Problems in Engineering 1563-5147 1024-123X Hindawi Publishing Corporation 10.1155/2014/682738 682738 Research Article Due Date Single Machine Scheduling Problems with Nonlinear Deterioration and Learning Effects and Past Sequence Dependent Setup Times Ceylan Hüseyin Arik Sabri Kırıkkale Vocational School Kırıkkale University 71451 Kırıkkale Turkey kku.edu.tr 2014 2372014 2014 26 02 2014 07 07 2014 09 07 2014 23 7 2014 2014 Copyright © 2014 Hüseyin Ceylan. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

We present some problems against due dates with nonlinear learning and deterioration effects and past sequence dependent setup times. In this study, two effects (learning and deterioration) are used for the same processing time. The processing time of a job is shorter if it is scheduled later, rather than in the sequence. This phenomenon is known in the literature as a “learning effect.” On the other hand, in many realistic scheduling settings, a job processed later consumes more time than the same job processed earlier—this is known as scheduling with deteriorating jobs. In the past sequence dependent setup times approach, the setup time of a job is proportionate to the sum of processing times of the jobs already scheduled. In this study, we demonstrated that some problems with due dates remain polynomially solvable. However, for some other problems, we concentrated on finding polynomially solves under their special cases.

1. Introduction

This paper addresses several problems against due dates with nonlinear learning and deterioration effects and past sequence dependent setup times. A scheduling problem is very important for a manufacturing system. So, numerous scheduling problems have been studied by researchers. In classical scheduling theory, the processing times of jobs are assumed to be known and fixed. However, in many realistic settings, workstations improve continuously as a result of repeating the same or similar activities such as the production of glass crafts by a skilled craftsman , the car repair or maintenance by a worker, and the patient diagnosis and treatment by a doctor . Thus, the processing time of a job is shorter if it is scheduled later, rather than in the sequence. Mosheiov  determined that this phenomenon is known in the literature as a “learning effect.” Biskup  was the first to investigate the learning effect in the scheduling problems. He assumed that the production time as a function of the number of repetitions of the production of a single item with a learning effect decreases. A number of researchers  studied E/T problem with learning effect, which is with a single machine and no deterioration effect. They solved unrestricted common due date problem as an assignment problem. In this work, we used Biskup’s learning effect (see ). Consider (1) p * = p i × r a , where p * is basic processing time of job i performed at position r when a 0 is learning index.

There is a growing interest in the literature to study scheduling problems of deterioration jobs; that is, jobs whose processing times are increasing functions of their starting times. Mosheiov  pointed out another case searching for an object under worsening weather or growing darkness. Although J. N. D. Gupta and S. K. Gupta  provided an example of steel rolling miles where the temperate of an ingot, while waiting to enter the rolling machine, drops below a certain level, which required the ingot to be reheated before rolling, Kunnathur and Gupta  gave a fire fighting example where the time and effort required to control a fire increase if there is a delay in the start of the fire-fighting effort. Alidaee and Womer  classified deteriorating jobs models into three different types: linear, piecewise linear, and nonlinear. In this paper, we are interested in nonlinear deterioration effect. One has (2) P r ( t ) = ( p r + α t r b ) , where t i is starting time of J i and α    ( α > 0 ) and ( b > 0 ) are nonlinear deterioration effect, which is the amount of increase in the processing time of a job per unit delay in its starting time.

In the scheduling literature, many researchers have worked different combinations of these three special cases to solve single machine, parallel machine, and flow shop scheduling problems. In this paper, we introduce firstly position dependent learning effect, nonlinear deterioration effect, and past sequence dependent setup time to solve due date problems in the single machine scheduling environment.

In the literature, there are a few studies on scheduling problems with effects of learning and deterioration simultaneously. Wang and Cheng  studied single machine scheduling problem with deteriorating jobs and learning effects to minimize the makespan when Wang  solved polynomially single machine scheduling problems with deteriorating jobs and learning effects. Toksari and Güner  proposed mixed nonlinear integer model for parallel machine earliness/tardiness scheduling problem with sequence dependent setup time and the effects of learning and deterioration. In another paper, they  proposed that the optimum solution of the same problem is V-shaped. Toksari and Güner  consider a parallel machine earliness/tardiness (ET) scheduling problem with different penalties under the effects of position based learning and linear and nonlinear deteriorations. The same authors  also worked on some scheduling problems with the nonlinear effects of learning and deterioration. Cheng et al.  derived polynomial-time optimal solutions for some scheduling problems with deteriorating jobs and learning effect. Linear deterioration effect was used in all studies on scheduling problems effects of learning and deterioration simultaneously. Güner and Toksari  studied the same scheduling problems with position based learning effect and linear deterioration simultaneously. We extended this study using nonlinear deterioration effect for the same problems.

2. Problem Formulation

Güner and Toksari  considered the position dependent learning effect of a job and linear deterioration effect for the same processing time. In this study, these two effects are combined as follows: (3) p * = ( p r + ( α × t r b ) ) r a . There are n jobs to be scheduled on single machine. If job i , i = 1,2 , , n , is being scheduled in position r in a sequence, its actual processing time is p * . p r is basic processing time of job scheduled in position r . α    ( α > 0 ) and ( b > 0 ) are nonlinear deterioration effect, which is the amount of increase in the processing time of a job per unit delay in its starting time. a    ( a < 0 ) is the learning index. t r is starting time of job scheduling in position r , and ( C r - 1 * ) is the actual completion time of the job scheduled in position ( r - 1 ) . Thus, the actual processing time p * is formulized as follows: (4) p * = p r + ( α × ( C r - 1 * b ) ) r a , where ( C r - 1 * ) is the actual completion time of the job scheduled in position ( r - 1 ) .

We also take setup times into consideration in the scheduling model by adopting the notion of Koulamas and Kyparisis  that setup times are past sequence dependent. Koulamas and Kyparisis  motivate the assumption of p-s-d setup times by the manufacturing of integrated circuit (IC) boards. Specifically, an IC board consists of a number of electronic components mounted on it and the processing of any electronic component will have an adverse effect on the “readiness” of other components on the board due to the passage of electricity through the board. Consequently, each component prior to processing requires a setup operation to restore it to “full-readiness” status and the setup time depends on the component’s degree of “unreadiness,” which is proportional to the actual processing times of the already processed components. Thus, the p-s-d setup time ( s r psd ) of job j if it is scheduled in the r th position in a sequence is given as follows: (5) s 1 psd = 0 , s r psd = γ i = 1 r - 1 p * , where γ 0 is a normalizing constant. The value of the normalizing constant γ determines the actual lengths of the required setups and, when γ = 0 , there is no need for any p-s-d setups (see ). We denote the effects of the learning and deterioration in (3) by learning effect and nonlinear deterioration effect and denote the p-s-d setup given in (5) by s p-q-d .

3. Some Single Machine Scheduling Problems with Individual Due Dates under Nonlinear Effects of Learning and Deterioration

Before presenting the main results, we first present a lemma, which will be used in the proofs of the theorems in the sequel.

Lemma 1.

Consider (6) C * = [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × ( C * r - 1 b ) ) ] ( j ) a ] + i = 2 r [ [ p i + ( α × ( C r - 1 * b ) ) ] ( i ) a ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × ( C r - 1 * b ) ) ] ( j ) a ] ] , where (7) p * = [ p r + ( α × ( C r - 1 * b ) ) ] ( r ) a , C 1 * = s 1 * + p 1 * = p 1 + M + ( α × M b ) , when M is starting time of the schedule.

Proof.

One has (8) s 1 * = γ ( 0 ) = 0 ; p 1 * = [ p 1 + ( α × M b ) ] ( 1 ) a = p 1 + ( α × M b ) , C 1 * = s 1 * + p 1 * = M + p 1 + ( α × M b ) , s 2 * = γ ( p 1 * ) , p 2 * = [ p 2 + ( α × ( C 1 * b ) ) ] ( 2 ) a , C 2 * = [ C 1 * + s 2 * + p 2 * = M + [ [ p 1 + ( α × M b ) ] ( 1 ) a ] + [ [ p 2 + ( α × ( C 1 * b ) ) ] ( 2 ) a ] + γ ( p 1 * ) ] , s 3 * = γ ( p 1 * + p 2 * ) , p 3 * = [ p 3 + ( α × ( C 2 * b ) ) ] ( 3 ) a , C 3 * = C 2 * + s 3 * + p 3 * = M + [ [ [ p 1 + ( α × M b ) ] ( 1 ) a ] k + [ [ p 2 + ( α × ( C 1 * b ) ) ] ( 2 ) a ] k + [ [ p 3 + ( α × ( C 2 * b ) ) ] ( 3 ) a ] k [ [ p 1 + ( α × M b ) ] ( 1 ) a ] + γ ( p 1 * ) + γ ( p 1 * + p 2 * ) ] , then, we can obtain the following: (9) p * = [ p r + ( α × ( C r - 1 * b ) ) ] ( r ) a , C * = [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × ( C r - 1 * b ) ) ] ( j ) a ] + i = 2 r [ [ p i + ( α × ( C r - 1 * b ) ) ] ( i ) a ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × ( C r - 1 * b ) ) ] ( j ) a ] ] , where (10) C 1 * = s 1 * + p 1 * = p 1 + M + ( α × M b ) . This completes the proof.

Let L j , L max = max { d j - C j j = 1,2 , , n } , T j , and T max = max { { d j - C j j = 1,2 , , n } , 0 } represent the total lateness, total tardiness, maximum lateness, and maximum tardiness of a given permutation, respectively. These problems with past sequence dependent setup times were studied by Biskup and Herrmann  and Güner and Toksari  studied the same scheduling problems with position based learning effect and linear deterioration simultaneously. We used the same approach with Güner and Toksari  to prove the theorems.

Theorem 2.

The problem can be solved optimally by sequencing jobs in nondecreasing order of their processing times (SPT rule).

Proof.

Consider an optimal schedule π . Assume, under π , there are two adjacent jobs, job J u , whose sequence is r , followed by job J v , whose sequence is r + 1 , such that p u < p v . C u * and C v * express completion times of J u and J v scheduled at positions u and v    ( v = u + 1 ) under the nonlinear deterioration and learning effect, respectively. M is starting time of the schedule.

Lateness of job in position r can be calculated by (11) L r * = ( C r * - d r ) ; then, from Lemma 1, (12) L u * ( π ) = [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] k + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] k + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] - d u , L v * ( π ) = [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] + [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] + [ + γ [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p v + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] k + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + γ [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] - d v , where ( C u b * ( π ) ) r explains the completion time of J u when it is in position r at schedule π . If we assume that t is total lateness up to position r , then (13) L ( π ) = t + 2 [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 b * ) ] ( j ) a ] + i = 2 r - 1 [ [ p i + ( α × C i - 1 b * ) ] ( i ) a ] + [ [ p u + ( α × C r - 1 b * ) ] ( r ) a ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 b * ) ] ( j ) a ] ] + [ + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p v + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] k + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + γ [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] - d v - d u . Performing a pairwise interchange on jobs J u and J v , we obtain schedule π . The completion times of the jobs processed before jobs J u and J v are not affected by interchange; then, (14) L v * ( π ) = [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] k + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] k + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] - d v , L u * ( π ) = [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] k + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] k + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] + + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p u + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] k + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + γ [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] - d u . If we assume that t is total lateness up to position r , then (15) L ( π ) = t + 2 [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 b * ) ] ( j ) a ] b + i = 2 r - 1 [ [ p i + ( α × C i - 1 b * ) ] ( i ) a ] + [ [ p u + ( α × C r - 1 b * ) ] ( r ) a ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 b * ) ] ( j ) a ] ] + + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p u + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] + γ [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] - d v - d u . The difference between the values of L ( π ) and L ( π ) is (16) ( L ( π ) - L ( π ) ) = [ + ( α ( C r b * ( π ) - C u b * ( π ) ) ( r + 1 ) a ) ( ( γ + 2 ) ( p v - p u ) ( r ) a ) - ( ( p v - p u ) ( r + 1 ) a ) k + ( α ( C r b * ( π ) - C u b * ( π ) ) ( r + 1 ) a ) ] = [ ( ( γ + 2 ) ( p v - p u ) ( r ) a ) - ( ( p v - p u ) ( r + 1 ) a ) k + ( α ( p v - p u ) ( r ) a ( r + 1 ) a ) ( ( γ + 2 ) ( p v - p u ) ( r ) a ) - ( ( p v - p u ) ( r + 1 ) a ) ] , L ( π ) - L ( π ) 0 , where p u < p v , ( a < 0 ) , ( b > 0 ) , and ( α > 0 ) .

Consequently, L ( π ) > L ( π ) .

The total lateness under π is strictly less than that under π . This completes the proof.

Example 3.

Let p 1 = 100 , p 2 = 90 , d 1 = 101 , d 2 = 102 , a = - 0.322 when learning rate % 80 , α = 0.1 , b = 0.2 and γ = 0.2 . Then, sequence π    ( S = ( 1,2 ) ) leads to p 1 * ( π ) = 100 , s 1 * ( π ) = 0 , C 1 * ( π ) = 100 , p 2 * ( π ) = 73.26 , s 2 * ( π ) = 20 , C 2 * ( π ) = 193.26 , and i = 1 2 L i ( π ) = 91.26 . The SPT sequence π    ( S = ( 2,1 ) ) leads to p 1 * ( π ) = 90 , s 1 * ( π ) = 0 , C 1 * ( π ) = 90 , p 2 * ( π ) = 81.24 , s 2 * ( π ) = 18 , C 2 * ( π ) = 189.24 , and i = 1 2 L i ( π ) = 88.24 .

Theorem 4.

The problem 1 [ p r + ( α × t r b ) ] ( r ) a + s r * p s d T can be solved optimally by sequencing jobs in nondecreasing order of their processing times (SPT rule).

Proof.

Tardiness T is defined as T = { 0 , L } . The results of Theorem 2 can be transferred directly to the case of 1 | [ p r + ( α × t r b ) ] ( r ) a + s r * psd | T . This completes the proof.

Theorem 5.

The problem both 1 [ p r + ( α × t r b ) ] ( r ) a + s r * p s d L max and 1 [ p r + ( α × t r b ) ] ( r ) a + s r * p s d T max can be solved optimally by sequencing jobs in nondecreasing order of their processing times (SPT rule) (or equivalently earliness due date (EDD) rule) if the due dates and the processing times are agreeable; that is, if p u p v implies d u d v .

Proof.

We would not duplicate L u * ( π ) , L v * ( π ) , L u * ( π ) , and L v * ( π ) due to knowing them from Theorem 4.

The difference between the values of L u * ( π ) and L v * ( π ) is (17) ( L v * ( π ) - L u * ( π ) ) = + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p v + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] + γ [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] - [ d v - d u ] , L v * ( π ) - L u * ( π ) 0 , where p u p v , d u d v , ( a < 0 ) , and ( α , b > 0 ) .

Consequently, L v * ( π ) L u * ( π ) .

The difference between the values of L u * ( π ) and L v * ( π ) is (18) ( L u * ( π ) - L v * ( π ) ) = [ + ( α ( C v b * ( π ) - C u b * ( π ) ) ( r + 1 ) a ) ( ( γ + 1 ) ( p v - p u ) ( r ) a ) - ( ( p v - p u ) ( r + 1 ) a ) k + ( α ( C v b * ( π ) - C u b * ( π ) ) ( r + 1 ) a ) ] - ( d u - d v ) , L u * ( π ) - L v * ( π ) 0 , where p u < p v , ( a < 0 ) , and ( α > 0 ) .

Consequently, L u * ( π ) L v * ( π ) .

The total lateness under π is strictly less than that under π .

The maximum tardiness T max is defined as T max = { 0 , L max } . The obtained results can be transferred directly to the case of 1 [ p r + ( α × t r b ) ] ( r ) a + s r * psd T max .

This completes the proof.

Example 6.

Let p 1 = 90 , p 2 = 100 , d 1 = 101 , d 2 = 102 , a = - 0.322 when learning rate % 80 , α = 0.1 , b = 0.2 and γ = 0.2 . Then, the SPT sequence π    ( S = ( 1,2 ) ) leads to p 1 * ( π ) = 90 , s 1 * ( π ) = 0 , C 1 * ( π ) = 90 , p 2 * ( π ) = 80.2 , s 2 * ( π ) = 18 , C 2 * ( π ) = 188.19 , and L max ( π ) = 86.19 . Sequence π    ( S = ( 2,1 ) ) leads to p 1 * ( π ) = 100 , s 1 * ( π ) = 0 , C 1 * ( π ) = 100 , p 2 * ( π ) = 72.2 , s 2 * ( π ) = 20 , C 2 * ( π ) = 192.2 , and L max ( π ) = 91.2 .

4. Common Due Date Problem

In this section, we tackle common due date problem with past sequence dependent setup times under learning effect and nonlinear deterioration effect. An excellent introduction to common due date problems is given as (19) 1 [ p r + ( α × t r ) ] ( r ) a + s r * psd , d j = d ( E j + T j ) .

Theorem 7.

Consider 1 [ p r + ( α × t r ) ] ( r ) a + s r * p s d , d j = d ( E j + T j ) V-shape: jobs in set of jobs that are completed on or before the due date are sequenced in nonincreasing order of the p j ; jobs in set of jobs that are completed after the due date are sequenced in nondecreasing order of the p j .

Proof.

Consider an optimal schedule π . Assume, under π , there are two adjacent jobs, job J u , whose sequence is r , followed by job J v , whose sequence is r + 1 , such that p u < p v . C u * and C v * express completion times of J u and J v scheduled at positions u and v    ( v = u + 1 ) under the nonlinear deterioration and learning effect, respectively. M is starting time of the schedule.

Earliness of job in position r can be calculated by (20) E [ r ] = ( d - C r * ) ; then, from Lemma 1, (21) E u * ( π ) = d - [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C * j - 1 b ) ] ( j ) a ] + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] + [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] , E v * ( π ) = d - [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] + [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] + + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p v + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] + γ [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] . If we assume that t is total earliness up to position r , then (22) E * ( π ) = ( 2 d - t ) - 2 × [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] k + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] k + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] + + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p v + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] k + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + γ [ [ p u + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] . Performing a pairwise interchange on jobs J u and J v , we obtain schedule π . The completion times of the jobs processed before jobs J u and J v are not affected by interchange; then, (23) E v * ( π ) = d - [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] k + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] k + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] , E u * ( π ) = d - [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] + [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] + + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p u + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] k + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + γ [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] . If we assume that t is total earliness up to position r , then (24) E * ( π ) = ( 2 d - t ) - 2 × [ [ M + p 1 + ( α × M b ) ] + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] k + i = 2 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] k + γ i = 1 r - 1 j = 1 i [ [ p j + ( α × C j - 1 * b ) ] ( j ) a ] ] + [ + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] [ [ [ p u + ( α × C u * b ( π ) ) ] ( r + 1 ) a ] ] k + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] k + γ [ [ p v + ( α × C r - 1 * b ) ] ( r ) a ] + γ i = 1 r - 1 [ [ p i + ( α × C i - 1 * b ) ] ( i ) a ] ] . The difference between the values of E * ( π ) and E * ( π ) is (25) ( E * ( π ) - E * ( π ) ) = [ ( ( p v - p u ) ( r + 1 ) a ) + ( α ( C u * b ( π ) - C v * b ( π ) ) ( r + 1 ) a ) + ( ( γ + 2 ) ( p u - p v ) ( r ) a ) ] , E * ( π ) - E * ( π ) 0 , where p u < p v , ( a < 0 ) , and ( α , b > 0 ) .

Consequently, E * ( π ) E * ( π ) ; it means that jobs in set of jobs that are completed before the due date are sequenced in nonincreasing order of the p j .

If a similar argument follows for the jobs that are started after the common due date d , we will obtain the following: (26) T * ( π ) - T * ( π ) 0 , where p u < p v , ( a < 0 ) , and ( α > 0 ) .

Consequently, T * ( π ) T * ( π ) ; it means that jobs in set of jobs that are completed after the due date are sequenced in nondecreasing order of the p j .

5. Conclusions

In this paper, we introduced some problems against due dates with past sequence dependent setup times under learning effect and nonlinear deterioration effect. In this study, effects of learning and deterioration are considered simultaneously. We present that some problems remain polynomially solvable. These problems are minimizing total lateness, total tardiness, maximum lateness (with agreeable due dates), maximum tardiness (with agreeable due dates), and earliness/tardiness problem with common due date.

Conflict of Interests

The author declares that there is no conflict of interests regarding the publication of this paper.

Cheng T. C. E. Wu C. Lee W. Some scheduling problems with deteriorating jobs and learning effects Computers and Industrial Engineering 2008 54 4 972 982 10.1016/j.cie.2007.11.006 2-s2.0-41449089275 Yin Y. Xu D. Wang J. Some single-machine scheduling problems with past-sequence-dependent setup times and a general learning effect International Journal of Advanced Manufacturing Technology 2010 48 9–12 1123 1132 10.1007/s00170-009-2360-9 2-s2.0-77953613682 Mosheiov G. Scheduling problems with a learning effect European Journal of Operational Research 2001 132 3 687 693 10.1016/S0377-2217(00)00175-2 MR1832521 ZBL1017.90051 2-s2.0-0035427775 Biskup D. Single-machine scheduling with learning considerations European Journal of Operational Research 1999 115 1 173 178 10.1016/S0377-2217(98)00246-X ZBL0946.90025 2-s2.0-0345201636 Mosheiov G. Sidney J. B. Scheduling with general job-dependent learning curves European Journal of Operational Research 2003 147 3 665 670 10.1016/S0377-2217(02)00358-2 MR1965663 ZBL1037.90529 2-s2.0-0037449223 Mosheiov G. A-shaped policies to schedule deteriorating jobs The Journal of the Operational Research Society 1996 47 9 1184 1191 10.1057/jors.1996.146 2-s2.0-0030305471 Gupta J. N. D. Gupta S. K. Single facility scheduling with nonlinear processing times Computers and Industrial Engineering 1988 14 4 387 393 10.1016/0360-8352(88)90041-1 2-s2.0-0024137249 Kunnathur A. S. Gupta S. K. Minimizing the makespan with late start penalties added to processing times in a single facility scheduling problem European Journal of Operational Research 1990 47 1 56 64 10.1016/0377-2217(90)90089-T MR1067998 ZBL0717.90034 2-s2.0-0025700721 Alidaee B. Womer N. K. Scheduling with time dependent processing times: review and extensions The Journal of the Operational Research Society 1999 50 7 711 720 10.1057/palgrave.jors.2600740 2-s2.0-0032682389 Wang X. Cheng T. C. E. Single-machine scheduling with deteriorating jobs and learning effects to minimize the makespan European Journal of Operational Research 2007 178 1 57 70 10.1016/j.ejor.2006.01.017 MR2281468 ZBL1110.90045 2-s2.0-33750843991 Wang J. Single-machine scheduling problems with the effects of learning and deterioration Omega 2007 35 4 397 402 10.1016/j.omega.2005.07.008 2-s2.0-33748913175 Toksari M. D. Güner E. Minimizing the earliness/tardiness costs on parallel machine with learning effects and deteriorating jobs: A mixed nonlinear integer programming approach International Journal of Advanced Manufacturing Technology 2008 38 7-8 801 808 10.1007/s00170-007-1128-3 2-s2.0-49649118743 Toksari M. D. Güner E. Parallel machine earliness/tardiness scheduling problem under the effects of position based learning and linear/nonlinear deterioration Computers and Operations Research 2009 36 8 2394 2417 10.1016/j.cor.2008.09.012 ZBL1179.90158 2-s2.0-58549092975 Güner E. Toksari M. D. Single machine scheduling problems against due date with the past sequence dependent setup times under learning and deterioration effects 38 Proceedings of the 38th International Conference on Computers and Industrial Engineering 2008 Beijing, China 1923 1930 Koulamas C. Kyparisis G. J. Single-machine scheduling problems with past-sequence-dependent setup times European Journal of Operational Research 2008 187 3 1045 1049 10.1016/j.ejor.2006.03.066 MR2378328 ZBL1137.90498 2-s2.0-36849022486 Biskup D. Herrmann J. Single-machine scheduling against due dates with past-sequence-dependent setup times European Journal of Operational Research 2008 191 2 587 592 10.1016/j.ejor.2007.08.028 MR2435130 2-s2.0-43849098003