| With the growing trend towards great product variety and fluctuating market demands, today's manufacturing companies are looking for a reconfigurable manufacturing system capable of producing a family of parts with different volumes, which is referred to as a Multiple Parts Manufacturing System (MPMS). A MPMS should be capable of adapting to volume changes and introduction of new parts within the same family. This thesis provides a MPMS design methodology with a given part family and a reconfiguration planning methodology when new requirements are present.; To design a MPMS, a mixed integer nonlinear model is proposed to pursue a comprehensive optimal solution, which concurrently takes several issues into account: multiple-parts task allocation, line balancing, equipment selection, and throughput analysis. Since it is a NP hard problem, conventional mathematical optimization techniques can hardly solve it in polynomial time. The thesis thus presents a Genetic Algorithm based method to find a near-optimal solution in a shorter time span.; Given a family of part models, it is not always beneficial to produce all of them in a MPMS. Sometimes a part model is better to be produced in a Single Part Manufacturing System (SPMS). The thesis identifies some selection principles that guide the system designers to decide which part model should be manufactured individually in a SPMS rather than a MPMS.; To response a changing market, the requirements for the MPMS must change as well during the operational phase. Therefore, it is necessary to reconfigure an existing MPMS to satisfy the new requirements. A generic reconfigurable object model is constructed to capture the reconfigurability of all levels of objects in a reconfigurable MPMS, either on the system level like the MPMS itself, or on the component level like reconfigurable machine tools and reconfigurable fixtures. Based on this model, an AI-based Computer Aided Reconfiguration Planning (CARP) framework is developed to assist the human planners to generate a reconfiguration plan for a specific reconfigurable object. The CARP framework either employs the arbitration rules defined by the human experts to create a viable plan or relies on an A* searching method to find the optimal plan. |
