A data-driven modeling approach for integrated disassembly planning and scheduling

Abstract

Over the past three decades, practitioners and researchers in engineering and operations sciences have focused on disassembly planning as a way to increase profitability of re-manufacturing, recycling and disposal processes for end-of-life products. An important task in disassembly planning is the representation of feasible operations sequences for the products. Precedence constraints can be derived from geometrical and technical relations among a product’s parts and joints and used to narrow down the set of possible sequences. There are several studies which address the problem of finding minimal process trees or AND/OR graphs as a prerequisite to disassembly sequence planning. However, since most of the existing approaches focus on specific product examples or industrial case studies there is still a lack of generic data sets for the academic purpose of model testing and evaluation. In this study, a systematic approach is presented to establish feasible AND/OR graphs from scratch based on general product design assumptions. Artificial process data as generated using the proposed methodology can be applied to various problems in disassembly decision making. In this study, it is used to analyze the combined problem of operations sequence planning and machine scheduling for the disassembly of multiple heterogeneous products. For this matter, disassembly sequences for each product and the order of operations at the stations have to be determined simultaneously with the objective of minimizing makespan. In contrast to existing problem formulations, the presented model explicitly considers divergence of the product structure during disassembly by allowing for parallel processing of separate sub-assemblies that have been extracted from the same product. A mixed-integer-program is developed based on disassembly process graphs which are derived for each product to represent alternative and parallel operations. Model performance is evaluated using 360 random instances created by the proposed process data generator. In addition, an industrial case study is presented to demonstrate the application of the proposed MIP model in a real-world disassembly context.