An extended study on an open-shop scheduling problem using the minimisation of the sum of quadratic completion times
Introduction
In an open-shop scheduling model, each available job has to be processed on m machines on a route decided by a decision-maker to optimise an objective, such as makespan (the completion time of the last job) or the sum of completion times. Since the pioneering work of Gonzalez and Sahni [1], many researchers have been interested in this area of study. For a comprehensive survey of open-shop scheduling problems, readers can refer to Chen et al. [2] or Potts and Strusevich [3].
We consider an open-shop scheduling problem to minimise the sum of quadratic completion times. Open-shop scheduling problem can arise in many practical applications, such as in a network of diagnostic testing facilities in a hospital, automobile repair, satellite communications and quality control centers. The classical formulations of these problems assume that parameters, such as the processing time, are constant values known in advance. This assumption is inadequate when processing times vary. For example, in a hospital, the treatment cost and time for an emergency patient can increase significantly if the patient’s condition becomes more serious due to waiting for diagnostic testing facilities and operation rooms. Studies have concluded that the sum of quadratic completion time is a suitable objective function for the aforementioned cases [4], i.e., the more time it takes a job to finish, the greater the cost per unit of elapsed time, which motives a study of open-shop scheduling to minimise the sum of quadratic completion times.
Using the three-field notation of Graham et al. [5], the open shop problem to minimise the sum of quadratic completion times can be denoted by , where m is the number of machines. The objective of minimising total quadratic completion time was first presented by Townsend [6], who showed that using this objective, the single-machine problem can be solved by the Shortest Processing Time first (SPT) rule in polynomial time, and he also provided a branch-and-bound algorithm for the weighted version. Cheng and Liu [4] showed that the quadratic objective is more appropriate for cases where the more time it takes a job to finish, the greater the cost per unit of elapsed time. Additionally, the sum of quadratic completion times as the objective function is a trade-off between the makespan and the total completion time, which are two important objective functions in scheduling problems. The authors proved the strong NP-hardness of and the asymptotical optimality of the SPT heuristic for the problem. Koulamas and Kyparisis [7] showed that is strongly NP-hard and proved that the SPT heuristic with worst case ratio is asymptotically optimal for the problem. Koulamas and Kyparisis [8] introduced a general objective, minimisation of the total loss of job values, , where the job value is a power function of the completion time job j (clearly, if and for all jobs, the general objective becomes the quadratic objective). The authors established the NP-hardness for and strong NP-hardness for and also solved several special cases of this general objective in single-machine and parallel-machine settings in polynomial time, respectively, and constructed a branch-and-bound algorithm and several heuristic algorithms for . Janiak et al. [9] presented a job-scheduling problem with deteriorating job processing times and decreasing job values. The objective was to minimise the weighted sum of quadratic completion times in a parallel processor environment. Moreover, the other types of general objective functions dependent on job completion times are also extensively considered in scheduling problems. The interested reader can refer to the work of Raut et al. [10], Janiak et al. [11], Janiak and Krysiak [12], [13], Janiak et al. [14] and Janiak and Krysiak [15].
Thus far, the related studies on open-shop scheduling problem are rare. In this paper, we extend a previous study on the open-shop problem to minimise the sum of quadratic completion times. To avoid duplication, the effectiveness of the quadratic objective and the NP-hardness of this problem can be found in a companion paper of Bai and Zhang [16]. Although Bai and Zhang [16] proved the asymptotic optimality of the SPTB heuristic for large-sized problems, practical tests showed inferior performance for small-sized problems.
In our work, a Lagrangian relaxation method is proposed to obtain the near optimal solution for small-sized instances. In addition, we present optimal properties for two special cases of the problem. Numerical experiments show the superior performance of the Lagrangian relaxation method.
The remainder of the paper is arranged as follows. The formulation of the open-shop problem is presented in Section 2. The Lagrangian relaxation method and the optimal properties are provided in Sections 3 A Lagrangian relaxation based solution approach for small-scale instances, 4 The optimal properties for two special cases, respectively. Section 5 presents the computational results, and the discussion and conclusions appear in Section 6.
Section snippets
Problem formulation
Generally, an open shop consists of m machines, each of which processes a different operation, and n jobs having m operations. The order in which a job passes through the m machines is arbitrary. The operation of job , on machine , is denoted by . Each operation has a processing time . No machine processes more than one operation, and no job is assigned to more than one machine at any time. Preemption and delaying of an operation is not allowed, and jobs
A Lagrangian relaxation based solution approach for small-scale instances
In this section, we propose a Lagrangian relaxation method to solve the program for small-scale instances.
The optimal properties for two special cases
This section derives the optimal properties for two special cases of the problem.
Consider the problem, where denotes denotes and . This problem can be solved by the algorithm presented by Dror [18]. Algorithm 1 Continuously schedule job 1 on machines . Continuously schedule job , on machines , and then on machine 1. All jobs begin processing on the m machines without any delay. Property 1 For problem , [18]
Computational results
We tested the performance of the proposed method to demonstrate the effectiveness of the proposed algorithm. The number of jobs (n) is set to 3, 4, 5, 6, 7, 8, 9, and 10, and the number of machines (m) is set to 3, 4, 5, and 6. However, when the number of machines is six, only combinations of (6, 3), (6, 4), (6, 5) (where in the bracket, the first element denotes the number of machines and the second element denotes the number of jobs) are tested due to time limitation. Additionally, the
Discussion and conclusions
In this paper, we modelled as an integer programming and solved it with the Lagrangian relaxation technique. The simulation results revealed that most of the average gaps between the lower bound and upper bound can be controlled to 20% for a given combination of number of machines and jobs. Additionally, extended experiments demonstrated that the Lagrangian relaxation procedure surpasses the heuristics mentioned in [16] for small-scale instances. In addition, properties were introduced
Acknowledgements
The authors would like to thank the anonymous reviewers for their insightful comments and suggestions that have significantly improved the article. This research is partly supported by National Natural Science Foundation of China (Grant No. 71201107).
References (18)
- et al.
Optimization and approximation in deterministic sequencing and scheduling: a survey
Ann. Discrete Math.
(1979) - et al.
A scheduling problem with job values given as a power function of their completion times
Eur. J. Oper. Res.
(2009) - et al.
Scheduling jobs with values dependent on their completion times
Int. J. Prod. Econ.
(2012) Openshop scheduling with machine dependent processing times
Discrete Appl. Math.
(1992)- et al.
Open shop scheduling to minimize finish time
J. ACM (JACM)
(1976) - B. Chen, C. Potts, G. Woeginger, A review of machine scheduling: complexity, algorithms and approximability, in:...
- et al.
Fifty years of scheduling: a survey of milestones
J. Oper. Res. Soc.
(2009) - et al.
Parallel machine scheduling to minimize the sum of quadratic completion times
IIE Trans.
(2004) The single machine problem with quadratic penalty function of completion times: a branch-and-bound solution
Manage. Sci.
(1978)
Cited by (7)
Dynamic Programming algorithms and their applications in machine scheduling: A review
2022, Expert Systems with ApplicationsCitation Excerpt :Regarding the methods used, one can notice that they vary frequently. Also, combinations have been used in order to improve the efficiency of a given method or DP’s itself, as it can be seen in Guéret and Prins (1999) and Zhang and Bai (2014). It might be interesting to investigate some properties related to dominance rules and also the results of combining DP with heuristics in order to reduce the effects of storage intrinsic to DP.
Multi-objective open shop scheduling by considering human error and preventive maintenance
2019, Applied Mathematical ModellingAn improved bat optimization algorithm to solve the tasks scheduling problem in open shop
2021, Neural Computing and ApplicationsOssp with release dates to minimize the total completion time-a hypothetical case
2020, Advances in Mathematics: Scientific JournalOnline scheduling for hybrid flexible flow shop with Open-Shop
2019, Jisuanji Jicheng Zhizao Xitong/Computer Integrated Manufacturing Systems, CIMSExtended Genetic Algorithm for solving open-shop scheduling problem
2019, Soft Computing