Artificial intelligence search methods for multi-machine two-stage scheduling with due date penalty, inventory, and machining costs

Abstract

This paper evaluates artificial intelligence search methods for multi-machine two-stage scheduling problems with due date penalty, inventory, and machining costs. We compare four search methods: tabu search, simulated annealing, genetic algorithm, and neighborhood search. Computational results show that the tabu search performs best in terms of solution quality. The tabu search also requires much less computational time than the genetic algorithm and simulated annealing. As expected, the neighborhood search needs the smallest computational time, but gives the worst solution quality. To further improve the solution quality and computational time, this paper proposes a two-phase tabu search. The two-phase tabu search sequentially addresses two aspects of sequencing for the same problem, order- and component-based sequencing. The order-based tabu search identifies a sequence for customers’ orders. Starting from the sequence identified for customers’ orders, the component-based tabu search fine-tunes the sequence for components produced at the fabrication stage. The results show that the two-phase tabu search is better in solution quality and computational time than the one-phase tabu search. The difference in solution quality is more pronounced at the early stage of the search. Scope and purpose Most manufacturing firms have some form of separate fabrication and assembly stages. Raw materials are transformed into components at the fabrication stage and the components are then assembled into finished products at the assembly stage. The components and assembly items are typically routed in batch quantities through several machines/work centers in a predetermined order before the finished products are delivered to customers. In this study, we model fabrication and assembly work centers as multi-machine two-stage manufacturing systems where a given machine is used to assemble/produce at least one component/product. The scheduling problem considered in this study involves a scheduling decision that achieves three objectives concurrently: (1) meeting customers’ due dates, (2) minimizing inventory cost, and (3) minimizing machining cost. Each order is an indivisible scheduling element that needs to be delivered to customers on the due date. Each order triggers successive production events from upstream to downstream according to the bill-of-material structure between components and end products. The objective of this paper are three-fold: (1) to present a solution representation for the multi-machine two-stage scheduling problem, (2) to identify the best artificial intelligence search method for this problem based on extensive computational experiments, and (3) to propose a modified tabu search method to further improve the solution quality and computational time.