Cognitive Robotic Disassembly Sequencing For Electromechanical End-Of-Life Products Via Decision-Maker-Centered Heuristic Optimization Algorithm
No Thumbnail Available
Authors
Alshibli, Mohammad
Issue Date
2018-07-30
Type
Thesis
Language
en_US
Keywords
Decision support techniques , Electronics disassembly , End-of-life management , Robotics
Alternative Title
Abstract
End-of-life (EOL) disassembly has developed into a major research area within the sustainability paradigm, resulting in the emergence of several algorithms and models to solve related problems. End-of-life disassembly focuses on regaining the value added into products which are considered to have completed their useful lives due to a variety of reasons such as lack of technical functionality and/or lack of demand. Disassembly is known to possess unique characteristics due to possible changes in the EOL product structure and hence, cannot be considered as the reverse of assembly operations. With the same logic, obtaining a near-optimal/optimal disassembly sequence requires intelligent decision making during the disassembly when the sequence need to be regenerated to accommodate these unforeseeable changes. That is, if one or more components which were included in the original bill-of-material (BOM) of the product is missing and/or if one or more joint types are different than the ones that are listed in the original BOM, the sequencer needs to be able to adapt and generate a new and accurate alternative for disassembly. These considerations require disassembly sequencing to be solved by highly adaptive methodologies justifying the utilization of image detection technologies for online real-time disassembly. These methodologies should also be capable of handling efficient search techniques which would provide equally reliable but faster solutions compared to their exhaustive search counterparts. Therefore, EOL disassembly sequencing literature offers a variety of heuristics techniques such as Genetic Algorithm (GA), Tabu Search (TS), Ant Colony Optimization (ACO), Simulated Annealing (SA) and Neural Networks (NN). As with any data driven technique, the performance of the proposed methodologies is heavily reliant on the accuracy and the flexibility of the algorithms and their abilities to accommodate several special considerations such as preserving the precedence relationships during disassembly while obtaining near-optimal or optimal solutions. This research proposes three approaches to the EOL disassembly sequencing problem. The first approach builds on previous disassembly sequencing research and proposes a Tabu Search based methodology to solve the problem. The objectives of this proposed algorithm are to minimize: (1) the traveled distance by the robotic arm, (2) the number of disassembly method changes, and (3) the number of robotic arm travels by combining the identical-material components together and hence eliminating unnecessary disassembly operations. In addition to improving the quality of optimum sequence generation, a comprehensive statistical analysis comparing the results of the previous Genetic Algorithm with the proposed Tabu Search Algorithm is also included. Following this, the disassembly sequencing problem is further investigated by introducing an automated disassembly framework for end-of-life electronic products. This proposed model is able to incorporate decision makers’ (DMs’) preferences into the problem environment for efficient material and component recovery. The proposed disassembly sequencing approach is composed of two steps. The first step involves the detection of objects and deals with the identification of precedence relationships among components. This stage utilizes the BOMs of the EOL products as the primary data source. The second step identifies the most appropriate disassembly operation alternative for each component. This is often a challenging task requiring expert opinion since the decision is based on several factors such as the purpose of disassembly, the disassembly method to be used, and the component availability in the product. Given that there are several factors to be considered, the problem is modeled using a multi-criteria decision making (MCDM) method. In this regard, an Analytic Hierarchy Process (AHP) model is created to incorporate DMs’ verbal expressions into the decision problem while validating the consistency of findings. These results are then fed into a metaheuristic algorithm to obtain the optimum or near-optimum disassembly sequence. In this step, a metaheuristic technique, Simulated Annealing (SA) algorithm, is used. In order to test the robustness of the proposed Simulated Annealing algorithm an experiment is designed using an Orthogonal Array (OA) and a comparison with an exhaustive search is conducted. In addition to testing the robustness of SA, a third approach is simultaneously proposed to include multiple stations using task allocation. Task allocation is utilized to find the optimum or near-optimum solution to distribute the tasks over all the available stations using SA. The research concludes with proposing a serverless architecture to solve the resource allocation problem. The architecture also supports non-conventional solutions and machine learning which aligns with the problems investigated in this research. Numerical examples are provided to demonstrate the functionality of the proposed approaches.
Description
Citation
M. Alshibli, "Cognitive Robotic Disassembly Sequencing For Electromechanical End-Of-Life Products Via Decision-Maker-Centered Heuristic Optimization Algorithm", Ph.D. dissertation, Dept. of Computer Science and Engineering, Univ. of Bridgeport, Bridgeport, CT, 2018.