        BENCHMARK DATA SET FOR EVALUATION OF LINE BALANCING ALGORITHMS   Seamus M. McGovern (1) and Surendra M. Gupta (2)   (1) U.S. DOT National Transportation Systems Center 55 Broadway, Kendall Square Cambridge, Massachusetts 02142-1093 USA (2) Northeastern University Department of Mechanical and Industrial Engineering 334 Snell Engineering Center Boston, Massachusetts 02115-5000 USA     Abstract: A known-optimal solution benchmark data set consisting of varying-size instances has been formulated for use in the evaluation and analysis of heuristics used to solve the disassembly line balancing problem. These scalable instances are easily modified to allow for consideration of larger or smaller workstation cycle times (or bin sizes) and also have application to assembly line balancing problems and to the general class of bin-packing problems. Copyright © 2007 IFAC  Keywords: Data sets, heuristics, algorithms, manufacturing, combinatorial mathematics, benchmark examples.     1. INTRODUCTION The recent formulation of the Disassembly Line Balancing Problem and its proof as being NP- complete necessitates solution through the use of heuristics; however, heuristic solutions are typically sub-optimal, so it is of interest to evaluate and compare the performance of different heuristics on any new problem or when using newly developed heuristics. Unfortunately, a result of the contemporary nature of the Disassembly Line Balancing Problem is a dearth of available test instances. While similar in name to the Assembly Line Balancing Problem, there are enough differences and additional qualities to make the use of instances from the assembly area of study a partial solution at best. Due to these issues, a new benchmark has been proposed for use in evaluating various performance measures (including efficacy and time complexity, which is enabled through the scalable nature of the instance set). Because of the similarities to the Assembly Line Balancing Problem and the general class of Bin-Packing Problems, and since there is a reduction in the amount of evaluation criteria required by these problems, a corresponding reduction in this paper’s data set criteria may enable this benchmark to be used as an additional tool for heuristic evaluation in these established fields of study (i.e., assembly line balancing and bin packing). 2. BACKGROUND Disassembly has gained a great deal of attention in the recent literature due to its role in product recovery. A disassembly system faces many unique challenges; for example, it has significant inventory problems because of the disparity between the demands for certain parts or subassemblies and their yield from disassembly. The flow process is also different. As opposed to the normal “convergent” flow in regular assembly environment, in disassembly the flow process is “divergent” (a single product is broken down into many subassemblies and parts). There is also a high degree of uncertainty in the structure and the quality of the returned products. The condition of the products received is usually unknown and the reliability of the components is suspect. In addition, some parts of the product may cause pollution or may be hazardous. These parts may require special handling that can also influence the utilization of the disassembly workstations. For example, an automobile slated for disassembly contains a variety of parts that are dangerous to remove and/or present a hazard to the environment such as the battery, airbags, fuel, and oil. Various demand sources may also lead to complications in disassembly line balancing. The reusability of parts creates a demand for them, however the demands and availability of the reusable parts is significantly less predicable than what is found in the assembly