Research On The Modeling And Solution Methods Of Disassembly Sequence Planning Problem

Disassembly Sequence Planning(DSP) method is an optimization method to find optimized disassembly sequences of products. It helps to improve the product disassembly efficiency, shorten the disassembly period and reduce the disassembly cost. Its application in e-waste will help to reduce the environmental hazards and improve the recovery value of e-wastes. DSP problem is a NP-hard combinatorial optimization problem. Various methods have been proposed to solve the problem in the past ten years. However, it is difficult to solve the DSP problems effectively with different conditions and objectives even using the best proposed method.This dissertation reviews the research works on disassembly modeling, disassembly evaluation, DSP methods, etc. Based on the literature reviews, this dissertation conducts some systematic and in-depth studies on the DSP problem, multi-objective DSP problem and selective DSP problem. This dissertation proposes a new disassembly modeling method, systematic disassembly evaluation indicators, and some effective and efficient DSP methods. Moreover, a prototype DSP system is designed and developed according to the research achievements. The proposed DSP methods are applied in an engineering case study. This dissertation mainly includes the following research contents:(1) The main differences between the DSP problem for and the assembly sequence planning problem are studied. Based on it, the dissertation presents a disassembly modeling method that distinguishes the fasteners and parts, and systematic disassembly evaluation indicators for e-waste disassembly. The work lays the foundation for the following of this dissertation.(2) The DSP problem and its extension are studied deeply. A DSP method based on a Modified TLBO(MTLBO) algorithm is proposed, which is a discrete and papulation based evolutionary algorithm. It inherits the core idea of teaching-learning-based evolutionary mechanism from the TLBO algorithm, and is modified according to the characteristics of the DSP problem. The MTLBO algorithm mainly contains three parts: the Feasible Solution Generator(FSG) for generating feasible disassembly sequences, the Teaching Phase Operator(TPO) and Learning Phase Operator(LPO) for updating the population meanwhile keeping the feasibilities of solutions in the population. The effectiveness and superiority of the proposed MTLBO algorithm for solving DSP problem are verified by experiments.(3) The multi-objective DSP problem is studied deeply. A multi-objective DSP method based on a multi-objective MTLBO algorithm is proposed, which is a discrete and papulation based multi-objective evolutionary algorithm. It inherits the evolutionary mechanism from the MTLBO algorithm and borrows the selection mechanism from the Non-dominated Sorting Genetic Algorithm II. Meanwhile, it is modified according to the characteristics of the multi-objective DSP problem. The effectiveness and superiority of the proposed multi-objective MTLBO algorithm for solving multi-objective DSP problem are verified by experiments.(4) The Selective DSP problem is studied deeply. The method of Disassembly Petri Net(DPN) is used to model the disassembly process. A Q-Learning algorithm based SDSP method is proposed. The method of representing the states and actions based on the DPN model, the method of selecting actions and the method of reinforcement learning are described in detail. The effectiveness and superiority of the proposed Q-Learning algorithm based SDSP method are verified by experiments.(5) The application background of the DSP system is analyzed. A prototype DSP system is developed according to the research achievements of this dissertation. The proposed DSP methods are applied in an engineering case study.Finaly, the main contributions and innovations of this dissertation are summarized, and the future works are drawn.