This study introduces an innovative scheduling tool for dynamic identical parallel machine environments, leveraging a non-preemptive overlapping load adjustment methodology. Central to this tool is a dual-phase proactive-reactive dynamic scheduling mechanism. Initially, in the proactive phase, the tool utilizes an earliest loading strategy to map out a range of viable scheduling alternatives, capitalizing on job slack times without cementing any schedules prematurely. This is followed by a reactive phase, wherein the tool dynamically adjusts to real-time disturbances via a user-controlled module, thereby crafting adaptable schedules while ensuring continuous feasibility. Enhanced by real-time graphical displays, the tool offers end-users a holistic view of the scheduling process and the implications of unexpected disruptions, aiding in informed decision-making. Extensive computational experiments encompassing 600 instances of varying sizes - small, medium, and large - underscore the tool’s efficiency. These experiments reveal the generation of a wide spectrum of scheduling alternatives, ranging on average from 1676 for small-sized instances to as many as 298,210 for large-sized ones. Notably, the average maximum completion time (Cmax) closely mirrors the results obtained from the Cplex optimizer, with an average deviation of merely 1.07%, 0.48%, and 12.23% for small, medium, and large instances, respectively. Moreover, despite the creation of up to 300,000 potential scheduling solutions for larger instances, the tool consistently yields a low average number of tardy jobs; 1.28, 1.87, and 2.07 for the respective problem sizes, which clearly highlights its effectiveness. The outcomes conclusively demonstrate the tool’s flexibility and efficiency, offering end-users a multitude of superior scheduling options in the presence of unforeseen disruptions that characterize today’s operating environment.
Read full abstract