Disassembly

Disassembly planning is a key step in EoL treatment and has a crucial linking function between the product EoL and the recycling alterative in product recovery [Duflou et al., 2008]. The literature review indicates that a large number of researches in this field have been dedicated to the study of the disassembly cost, revenue and component clustering. A product can be usually disassembled through various ways, known as “sequence of disassembly unit operations” which has to be determined prior to the physical operation [Lambert, 2007, Gungor and Gupta, 1998]. That being said, an extensive body of research has been created in the past by focusing on disassembly sequencing as well as finding optimal or near-optimal disassembly sequence plan (DSP) [Gungor and Gupta, 1997, Wan and Krishna Gonnuru, 2013, Smith et al., 2012, Kara et al., 2006, Kaebernick et al., 2000].

A DSP is a sequence of disassembly which starts by processing a given product resulting in subassembly(ies) through different methods (e.g., connection graph, direct graph, AND/OR graph, etc.). In this regard, using CAD data, as seen in researches by Mani et al, and Arai and Iwata, is amongst the most classic research topics to evaluate the disassembly process during the design iteration phase [Mani et al., 2001, Arai and Iwata, 1993]. According to Güngör and Gupta, disassembly sequencing of a product can be either a partial or a complete operation [Güngör and Gupta, 2002]. The disassembly precedence tree has been formed fully or partially using geometrical relationship to optimally prioritize the disassembly process in several researches [Zhang and Kuo, 1997, Kuo, 2000, Kuo, 2006a, Tang et al., 2002]. Later on, attempting to seek the highest net revenue, finding the optimal disassembly depth and sequence have been also stressed particularly. Mathematical Programming (MP), heuristic, metaheuristics and artificial intelligence techniques are amongst the most common approaches in this field of research [Willems et al., 2006, Lambert, 2007, Go et al., 2012, Hui et al., 2008, Kalaycılar et al., 2016].

Achilas et al. proposed a decision support tool to determine the optimal depth of product disassembly [Achillas et al., 2013]. The developed model is a mathematical formulation based on cost benefit analysis concept in order to determine the depth of disassembly considering both environmental and economic concerns. This included the minimum recycling, reuse rate, personnel cost and recovered material prices. Seven discrete scenarios have been considered through altering these parameters in order to examine the effectiveness of the proposed approach. Despite the

optimization values that this method may offer, it is still not generic and can face difficulties to be used for other products EoL streams. Furthermore, they have not accommodated the disassembly intrinsic factors such as, type of disassembly actions needed to reclaim parts and tool (used into the methodology channel) which may result in partial effectiveness of this approach.

The heuristic methods are still being used widely by researchers to reach promising solutions in a shorter time as compared to other available methods. However, they do not necessarily result in the most optimal solutions. Consequently, their applications are often limited to collect all the good- enough solutions and then let the Mathematical Programming (MP) take steps. Literature is fairly rich on the heuristic applications [Güngör and Gupta, 2002, Langella, 2007, Inderfurth and Langella, 2006]. Güngör and Gupta implemented this method to modify the disassembly line balancing for an intricate product or for a large quantity of products in order to maximize the productivity by optimizing the line balancing [Güngör and Gupta, 2002].

MP applications are broad due to their capacity to find the optimum value when combined with heuristic or metaheuristics methods. Basically, a model containing connection diagram and a set of precedence relationships are needed. Mainly this information is described using AND/OR representation which contains all of the disassembly sequences in a product. Suzuki et al. conducted a research using binary integer linear programming to model the assembly process [Suzuki et al., 1993]. AND/OR graphs are a set of graphical presentation of the subassembly precedence. It is a useful tool when the number of elements in a product is not relatively high.

Various researches have been dedicated to the Artificial Intelligence (AI) applications in disassembly planning and line balancing problem (DLBP) [Avikal et al., 2014, Kalayci et al., 2015, Luo et al., 2016, Go et al., 2012]. Seo et al. developed a heuristic algorithm based on Genetic Algorithm (GA) to solve the disassembly sequence problem with an emphasis on the environmental and economic criteria [Seo et al., 2001]. The GA dynamically explores the disassembly nodes to find the optimal sequence. Hui et al. also implemented GA tool to solve a disassembly sequence plan through finding the optimal sequence based on a feasibility information graph (DFIG) [Hui et al., 2008]. Nonetheless, running a genetically optimized model may become more difficult and time-consuming when the number of connections and mated parts are high. Moreover, selecting a good fitness function and defining the solution space before genetic search space starts are also amongst the most prevalent issues.

As Smith et al. indicated in a research work, some of the noteworthy shortcomings of these models are: increased search time, low model quality and high complexity [Smith et al., 2012]. Besides, the study of relevant literature indicates that an efficient and feasible disassembly sequence can only be obtained if the disassembly operation itself is optimized, planned properly with aims to address the disassembly economy, coordination with the environment and technical feasibility [Go et al., 2012, Yan et al., 2006]. Nonetheless, the overall volume of literature dedicated to the disassembly physical operation is low and there is still much to learn about the subject. The present work in this thesis is intended to help filling this specific gap.