Examining the Performance of Heuristics for the Disassemble-to-Order Problem under Rolling Planning using Actual Product Structures

Product recovery management, where firms take products back from customers at the end of their use, has received increasing attention due to the push towards sustainability in operations management. The firm has several options as to what can be done with the returned products, one of which is remanufacturing. In remanufacturing, the product is disassembled, the parts inspected and reworked into "good as new'' quality, and the product reassembled and sold with the same warranty as a newly produced product. With a certain amount of remanufactured products demanded each with a certain amount of parts defined by the product structure, a disassemble to order problem results tasked with determining the optimal amount of returned products to disassemble to fulfill this demand. The problem can be formulated as a mixed integer linear program, but this has been shown to be NP hard, so practitioners will not be able to rely on exact solution methods for industrial sized problems. This contribution examines the performance of both exact as well as heuristic solution methods for the disassemble-to-order problem under a rolling schedule planning environment. The study utilized random variants of real product structures obtained from practice and literature. The results indicate that the performance of the heuristics can surpass that of the exact solution in a rolling horizon environment. This is particularly the case for the more advanced of the two heuristics tested, where in an overwhelming majority of the instances generated from real product structures, did the heuristic outperform the exact solution method