Exact methods for order acceptance and scheduling on unrelated parallel machines
Introduction
To maintain operational agility and flexibility, many companies from different industries, such as engineering tooling, industrial boilers, construction and contracting, adopt the make-to-order (MTO) operational philosophy, thereby laying more focus on customer satisfaction (Mestry et al., 2011). Additionally, there exists an increase in popularity of the MTO philosophy in the service industry, particularly with regard to E-commerce and O2O takeout & catering services offered within restaurants. For example, the ele.me online platform for O2O takeout & catering service covered more than 200 cities in China and had served more than 260 million customers by June 2017 (source from www.ele.me). In this context, how to coordinate operations and sales for effective use of available resource (or limited capacity) is a big challenge for improving customer satisfaction meanwhile obtaining high profit margins.
The order acceptance and scheduling (OAS) problem arises in different MTO production and/or service systems, wherein limited production and/or service capacity and order-delivery requirements necessitate the use of selective order acceptance to satisfy distinct requirements of customers whilst also maximizing total revenue (profit) (Cesaret, Oğuz, Salman, 2012, Rom, Slotnick, 2009, Silva, Subramanian, Pessoa, 2018, Slotnick, Morton, 2007, Wang, Huang, Hu, Cheng, 2015).
The OAS problem requires one to simultaneously determine which orders should be accepted for processing as well as their corresponding schedule. The complexity of the problem due to their combinatorial nature and intertwined decisions makes optimization extremely difficult, as the problem typically is NP-hard (Ghosh, 1997). Such problems, however, invariably capture the rich and realistic classes of MTO processing, thereby making them easy to explain and tempting to be attempted and solved optimally.
In this paper, a deterministic OAS problem on unrelated parallel machines is addressed, since the prevalence of actual manufacturing environments and service industries are typically equipped with unrelated parallel machines. Typical applications of unrelated parallel machine scheduling include—but are not limited to—semiconductor manufacturing (Şen, Bülbül, 2015, Detienne, Dauzère-Pérès, Yugma, 2011, Shim, Kim, 2007), multiprocessor computer (Fanjul-Peyro and Ruiz, 2010), operating rooms in hospitals (Fanjul-Peyro and Ruiz, 2012), and car factories and food processing plant (Fanjul-Peyro et al., 2017). For the problem under study, there exists a pool of potential orders (jobs) with known processing times, due dates, revenues, and penalty tardiness weights. The objective here is to maximize the total net revenue, which refers to the difference between the sum of revenues obtained from accepted jobs and total weighted tardiness. This paper investigates the problem from an exact solution viewpoint. The contributions of this research are as follows:
- (1)
Two MIP formulations are proposed: one is with a dummy job and another is based on linear ordering variables. Formulation tightening and valid inequalities are proposed to improve the efficiency of the two MIP formulations.
- (2)
A formulation-based branch-and-bound algorithm is developed based on the idea of “divide and conquer”, in which the branch is to determine how many jobs should be accepted, followed by the subproblem of unrelated parallel machine scheduling.
- (3)
The computational results on various instances show the efficiency of the formulations with enhancement techniques, and demonstrate the efficacy of the proposed branch-and-bound algorithm.
The rest of this paper is organized as follows. Section 2 presents literature review relevant to the problem. In Section 3, two MIP models are formulated. In Section 4, some enhancement techniques for the MIP models are presented. The proposed formulation-based branch-and-bound algorithm is described in Section 5. In Section 6, extensive computational experiments are conducted to evaluate the performance of the developed models and the branch-and-bound algorithm. Lastly, Section 7 concludes the paper and suggests some future research directions.
Section snippets
Research streamlines of OAS problems
The OAS problem and its variants have been extensively investigated for more than two decades (Esmaeilbeigi et al., 2016). Interested readers can refer to the survey by Slotnick (2011) and references herein for further details. The scheduling problem with rejection costs is an equivalent version of the OAS problem. The reader is referred to the survey by Shabtay et al. (2013) for more details. Equivalently, the prize-collecting scheduling problem, wherein job acceptance is not mandatory, but
Mathematical description of OAS problem
In this section, after the problem description, two MIP formulations are presented.
There is a set of jobs denoted by which are all ready at time zero. There is a set of machines which are all available from time zero. Each job is processed non-preemptively on exactly one of the machines and the processing times of jobs on machines are (where denotes the set of positive integers), corresponding to the time required to process job j on machine i . If
Enhancement of proposed MIP formulations
The OAS problem on unrelated parallel machines is NP-hard, since it can be reduced to a classical unrelated parallel machine problem with a dummy machine (see Appendix), which is known to be NP-hard in the strong sense (Fanjul-Peyro, Ruiz, 2012, Fleszar, Hindi, 2018). It is, therefore, necessary to develop enhancement techniques to improve the performance of proposed MIP models.
It is easy to know that there is no idle time between any two consecutive accepted jobs on a machine. In addition, the
A formulation-based branch-and-bound algorithm
For the exact optimization, a formulation-based branch-and-bound (B&B) algorithm is designed based on the idea of “divide and conquer”. The proposed algorithm branches on the number of accepted jobs nA, which may assume values between 1 and n. Once the number of accepted jobs, nA, is determined, the problem is reduced to an unrelated parallel machine scheduling problem with the objective of minimizing the total weighted tardiness, which can be explored by the formulations mentioned above. A
Computational experiments
To evaluate the performance of the MIP models, the enhancement techniques, and the formulation-based branch-and-bound algorithm, an extensive computational analysis is conducted. Seven methods, i.e., the MIP1 method described in Section 3.1, the MIP2 in Section 3.2, the MIP1V and MIP2V in Section 4 solved with the IP solver CPLEX, the B&B algorithm with (denoted as “BB1”, “BB5”, and “BB10”, respectively), are used to test problem instances to compare the performance of the models with
Conclusions
In this paper we have studied an order acceptance and scheduling problem on unrelated parallel machines, which has not yet been thoroughly explored in available literature. Two different formulations that can be solved with general-purpose IP solvers have been developed. Formulation tightening and valid inequalities have also been proposed to improve the efficiency of the formulations. A formulation-based branch-and-bound algorithm has been developed based on the idea of “divide and conquer”,
Acknowledgment
The authors would like to thank the anonymous referees for their constructive comments which contributed to improve the quality of this paper. This work was supported by the National Science Foundation of China (NSFC) with Grant No. 71571135. The work was also supported by the Fundamental Research Funds for the Central Universities.
References (54)
- et al.
A tabu search algorithm for order acceptance and scheduling
Comput. Oper. Res.
(2012) - et al.
Hybrid evolutionary approaches for the single machine order acceptance and scheduling problem
Appl. Soft Comput.
(2017) - et al.
Order acceptance and scheduling problems in two-machine flow shops: new mixed integer programming formulations
Eur. J. Oper. Res.
(2016) - et al.
Models and matheuristics for the unrelated parallel machine scheduling problem with additional resources
Eur. J. Oper. Res.
(2017) - et al.
Iterated greedy local search methods for unrelated parallel machine scheduling
Eur. J. Oper. Res.
(2010) - et al.
Size-reduction heuristics for the unrelated parallel machines scheduling problem
Comput. Oper. Res.
(2011) - et al.
Scheduling unrelated parallel machines with optional machines and jobs selection
Comput. Oper. Res.
(2012) - et al.
Reformulations and an exact algorithm for unrelated parallel machine scheduling problems with setup times
Comput. Oper. Res.
(2019) - et al.
Algorithms for the unrelated parallel machine scheduling problem with a resource constraint
Eur. J. Oper. Res.
(2018) Job selection in a heavily loaded shop
Comput. Oper. Res.
(1997)