Greedy randomized adaptive search procedure for simultaneous scheduling of production and preventive maintenance activities in dynamic flexible job shops

Abstract

In the present study, we proposed a greedy randomized adaptive search procedure (GRASP) for integrated scheduling of dynamic flexible job shops with a novel preventive maintenance policy. In most of the real-life scheduling practices, unexpected and unknown events occur frequently, which necessitates solving operations and maintenance scheduling problems dynamically. Dynamic events like new order arrival, machine breakdown, changes in due date, order cancellation, and urgent order are considered in this study. Moreover, order acceptance/rejection decisions and an order review release mechanism are also taken into account in order to enhance the overall performance by adjusting capacity regarding to customer due date requirements. Four objectives namely mean tardiness, schedule instability, makespan, and mean flow time are considered within a lexicographic programming logic. Random test instances are generated for the stated dynamic scheduling problem. In order to confirm the applicability of the proposed GRASP-based algorithm, extensive experiments were also conducted on well-known job shop scheduling benchmark instances and flexible job shop scheduling benchmark instances with preventive maintenance activities. Computational experiments conducted under various experimental settings such as flexibility level and due date tightness in addition to different preventive maintenance policies. To the best of our knowledge, the present study presents the first attempt through GRASP for simultaneous dynamic scheduling of operations and preventive maintenance activities in flexible job shops. Results of the extensive computational experiments demonstrate that simultaneous scheduling of manufacturing operations and preventive maintenance activities is a viable and effective approach for performance improvement in dynamic flexible job shop scheduling environments.

Appendices

Appendix A

In Tables 13, 14 and 15, Event type “1” depicts new order arrival. For instance, a new order J1 arrives at 50th minutes of the schedule. Event type “2” indicates machine breakdown. For example, machine M2 breakdowns at 240th minutes of the schedule and it takes 42 min to fix the machine M2. Event type “3” indicates order cancellation. For instance, eleventh order is cancelled at 550th minutes of the schedule. Event type “4” demonstrates rush order. For instance, the ninth order arrives at 70th minutes of the schedule as rush order. Event type “5” indicates change of due date. For instance, the due date of the eighth order is renewed at 100th minutes of the schedule and the new due date is updated to 2500 min. Setup times are uniformly distributed (8,15). Processing times, which are given in Tables 13, 14 and 15 are uniformly distributed. Capable machine sets of the operations are assigned randomly. Each shift is considered 480 min. Due date of orders are determined by using total work content methodology. The PM operation time and fixed time interval of the PM operation in Tables 13, 14 and 15 are used in the PM policy 1. In the proposed PM policy 2, the earliest beginning time of PM operations for each machine is equal to the fixed time interval value of PM operations and the latest time of PM operations for all machines is considered as 480 min.

Table 13 Data for 5 × 5 DFJSP

Table 14 Data for 10 × 10 DFJSP

Table 15 Data for 15 × 15 DFJSP

Appendix B

See Tables 16 and 17, 18, 19, 20 and 21.

Table 16 Computational results under PM policy 1 for 5 × 5 problems

Table 17 Computational results under PM policy 2 for 5 × 5 problems

Table 18 Computational results under PM policy 1 for 10 × 10 problems

Table 19 Computational results under under PM policy 2 for 10 × 10 problems

Table 20 Computational results under PM policy 1 for 15 × 15 problems

Table 21 Computational results under PM policy 2 for 15 × 15 problems