Abstract

This paper considers two resource constrained single-machine group scheduling problems. These problems involve variable job processing times (general position-dependent learning effects and deteriorating jobs); that is, the processing time of a job is defined by the function that involves its starting time and position in the group, and groups’ setup time is a positive strictly decreasing continuous function of the amount of consumed resource. Polynomial time algorithms are proposed to optimally solve the makespan minimization problem under the constraint that the total resource consumption does not exceed a given limit and the total resource consumption minimization problem under the constraint that the makespan does not exceed a given limit, respectively.

1. Introduction

In classical scheduling theory, the jobs’ processing time is always assumed to be fixed and constant value. Actually, we often encounter the case that job’s processing time may be subject to change due to its position in the production sequence. Many authors have researched this aspect. Mosheiov [1] researched scheduling problems with a learning effect; that is, the production time of a given product is shorter if it is scheduled later rather than earlier in the sequence. Voutsinas and Pappis [2] discussed scheduling jobs with production times deteriorating exponentially over time in remanufacturing environments. The objective was to maximize total value of jobs. Cheng et al. [3] considered some scheduling problems with deteriorating jobs and learning effects. They derived polynomial time optimal solutions for the problems to minimize makespan and total completion time. Sun [4] considered a scheduling model in which deteriorating jobs and learning effect were both considered. The jobs’ actual processing time depended on the processing time of the jobs already processed and their positions. He showed that the problems of minimizing the makespan, total completion time, and the sum of the quadratic job completion times remain polynomially solvable. Toksarı [5] presented a single-machine problem with the unequal release times under learning effect and deteriorating jobs. The objective was to minimize the makespan. He developed a branch-and-bound algorithm to solve it and proposed a heuristic algorithm to obtain a near optimal solution. Wang et al. [6] considered a single machine scheduling problem with a time-dependent learning effect and deteriorating jobs. The objective was to determine an optimal schedule so as to minimize the total completion time. They proposed three heuristic algorithms by using the V-shaped property. Wang et al. [7] studied a single-machine scheduling and due window assignment problem, which included deteriorating and learning effect. The objective was to minimize costs for earliness, tardiness, the window location, window size and makespan. They showed that the problem was polynomially solvable. Low and Lin [8] considered the model that the processing time of a job was determined by a function of its starting time and positional sequence in each machine. They discussed different objectives, such as the makespan, total completion time, and weighted completion time. Single-machine scheduling problems could be solved polynomially. They also showed that flow shop scheduling problems were polynomially solvable under some certain conditions. Wang et al. [9] considered flowshop scheduling problems with a general exponential learning effect. The objective was to minimize the makespan, the total (weighted) completion time, the total weighted discounted completion time, and the sum of the quadratic job completion times, respectively. They proposed several simple heuristic algorithms to solve these problems and gave the tight worst-case bound of these heuristic algorithms.

In manufacturing industry, firms can increase production efficiency by using group technology (GT). GT is an approach in manufacturing that seeks to improve efficiency in high-volume production by exploiting the similarities of different products and activities in their production. Many researchers have paid great attention to this area and done a lot of work. Kuo and Yang [10] introduced a time-dependent learning effect into single-machine group scheduling problems. The time-dependent learning effect of a job is assumed to be a function of total processing times of jobs scheduled in front of it. They showed that the single-machine group scheduling problem with a time-dependent learning effect remains polynomially solvable for two objectives: minimizing the makespan and the total completion time. Wu et al. [11] considered two single-machine scheduling problems in the context of group technology where job processing times and setup times are simple linear functions of their starting times. The objectives were minimization of makespan and total completion time. Yan et al. [12] studied the single-machine scheduling problem with the effects of deterioration and learning, where the jobs were under group consumption and the setup time of each group was dependent on the resource it consumed. Zhang and Yan [13] introduced a deteriorated and learning effect into a single-machine problem where the learning effect not only depends on job position, but also depends on the group position; the deteriorated effect depends on its starting time of the job. They showed that the makespan, the total completion time, and maximum lateness problems remained polynomially optimally solvable under the proposed model. S. J. Yang and D. L. Yang [14] investigated single-machine group scheduling problems. The group setup time was assumed to follow a simple linear time-dependent deteriorating model. They examined two models of learning for the job processing time, provided polynomial time solutions for the makespan minimization problems, and showed that the total completion time minimization problems remained polynomially solvable under agreeable conditions. Zhu et al. [15] studied group scheduling problems with learning effects and resource allocation concurrently. They determined an optimal group sequence, internal job sequence, and the amount of resource allocated for two objectives, respectively. One was to minimize the weighted sum of makespan and total resource cost. The other was to minimize the weighted sum of total completion time and total resource cost. Lee and Lu [16] considered a single-machine scheduling problem in the context of group technology where the job processing times and the group setup times are simple linear functions of their starting times. The objective is to minimize the total weighted number of late jobs. They provided a branch-and-bound algorithm to solve the problem. Huang et al. [17] and Huang and Wang [18] considered two group scheduling problems with time and position-dependent processing times. Ji et al. [19] discussed group scheduling and job-dependent due window assignment problem and provided a polynomial algorithm to solve it. More recent papers have considered scheduling jobs with deteriorating jobs and/or learning effects including Wu et al. [11], Wu and Lee [20], Wang et al. [21], Wang [22], Yin et al. [23], Wang et al. [24], Lee and Wu [25], Li et al. [26], Zhu et al. [27], Huang et al. [28], Janiak and Rudek [29], Ji and Cheng [30], Yang [31], Yin and Xu [32], Bai et al. [33], Yin et al. [34], Yin et al. [35], J. B. Wang and J. J. Wang [36], Wu et al. [37], and Xu et al. [38].

Yan et al. [12] and Huang and Wang [18] considered the single-machine scheduling problems with the effects of learning and/or deterioration under the group technology consumption. They proved that the single-machine makespan minimization problem and total resource minimization problem under the group technology consumption remain polynomially solvable. This paper extends the results of Yan et al. [12] and Huang and Wang [18], by considering more general position-dependent learning effects and deteriorating jobs model that includes the one given in Yan et al. [12] and Huang and Wang [18] as a special case. The remaining part of this paper is organized as follows. In Section 2 we describe and formulate the problem. In Sections 3 and 4 we consider two resource-dependent scheduling problems and present polynomial algorithms to solve them. The last section includes conclusions.

2. Problem Formulation

There are jobs to be processed on a single-machine, and these jobs grouped into groups. A setup time is required if the machine switches from one group to another and all setup times of groups for processing are at time 0. The machine can handle one job at a time and job preemption is not allowed. Let denote the th job in group , ; , where denote the number of jobs belonging to group   (i.e., ). Let be the (actual) setup time of group , and let be the (actual) processing time of job if it is started at time and scheduled in position . In this paper, we consider the following model:where denotes the normal processing time for job , , . is the general case of positional learning for group , specially ( is the polynomial learning index for group ) and ( is the exponential learning index for group ). In addition, the setup time iswhere is the amount of a nonrenewable resource allocated to group , with , is a technological constraint, and ( is the set of nonnegative real numbers) is a strictly decreasing continuous function with . Let be the resource allocation vector satisfying the resource constraint , where is the total amount of the resource available.

For a given schedule , let be the completion time of job under , and let be the makespan (the maximal completion time) of a given schedule. We consider two problems: (1) minimizing the makespan with a resource consumption constraint and (2) the total resource consumption with a makespan constraint. Using the three-field notation (Graham [39]) for scheduling problems, these two problems can be denoted by , , , and , , , .

3. The Problem

Lemma 1. For a given schedule , if the first job starts at time , then the completion time is equal towhere means job is the th one to be processed.

Proof (by mathematical induction method). For job , we haveSuppose Lemma 1 holds for job ; that is,Then, for job , we haveHence, Lemma 1 holds.

Theorem 2. For the , , problem, and a given schedule , , the makespan is equal to

Proof (by mathematical induction method). Stemming from Lemma 1, for group , we haveSuppose Theorem 2 holds for group ; that is,Then, for group , we have Hence, Theorem 2 holds.

Theorem 3. For the , , problem, the optimal job sequence in each group can be obtained by the smallest processing time (SPT) first rule (i.e., arranging jobs in nondecreasing order of ).

Proof. It is similar to the proof of Bai et al. [33].

Theorem 4. For the , , problem, the groups scheduled earlier should be given the priority to resource allocation for any schedule of the groups.

Proof. We assume is any schedule of all the groups. Stemming from (7), first allocate the resource to the group, whose setup time is bigger than ; the makespan can be minimized by the same amount of resource. ObviouslyHence, for the objective , the groups scheduled earlier should be given the priority to resource allocation.

Theorem 5. For the , , problem, the groups scheduled later should be given the priority to resource allocation for any schedule of the groups.

Proof. We assume is any schedule of all the groups. Stemming from (7), first allocate the resource to the group, whose setup time is bigger than ; the makespan can be minimized by the same amount of resource. ObviouslyHence, for the objective , the groups scheduled later should be given the priority to resource allocation.

Theorem 6. For the , , problem, if resource amount of the groups in each position is fixed and in each group the jobs are scheduled by the SPT rule, then the optimal schedule can be obtained by scheduling the groups in nonincreasing order of .

Proof. Without loss of generality, we assume that the resource amount of the group scheduled at the th position is denoted by   . Let be a schedule and let be a new schedule by changing the sequence of and . Obviously, for schedule , is scheduled in the th (th) position; then the resource allocation amount of is . Then, under the schedules and , we haveThenif and only if .

Based on Theorems 3, 4, and 6, an algorithm can be proposed to solve , , , .

Algorithm 7. We have the following.
Step 1. Jobs in each group are scheduled by the SPT rule; that is,Step 2. Groups are scheduled in nonincreasing order of ; that is,Step 3. For the obtained schedule , set    and .
Step 4. Set , , and .
Step 5. If or , stop; else go to Step 4.

Based on Theorems 3, 5, and 6, an algorithm can be proposed to solve , , , .

Algorithm 8. We have the following.
Step 1. Jobs in each group are scheduled by the SPT rule; that is,Step 2. Groups are scheduled in nonincreasing order of ; that is,Step 3. For the obtained schedule , set    and .
Step 4. Set , , and .
Step 5. If or , stop; else go to Step 4.

Obviously the total time for Algorithm 7 (Algorithm 8) is .

The following example is used to illustrate Algorithm 7.

Example 9. Consider , , , and . The normal processing times , for each group are given in Table 1, , , , and .

Solution. According to Algorithm 7, we solve Example 9 as follows.

Step 1. In group , the optimal job sequence is . In group , the optimal job sequence is . In group , the optimal job sequence is .

Step 2. Consider .

Step 3. The obtained schedule is .

Steps 4 and 5. The optimal resources , , .

Therefore, the optimal schedule is , and the optimal completion times are calculated as follows: