Multiple objective decision support framework for configuring, loading and reconfiguring manufacturing cells

The potential advantages of Cellular Manufacturing Systems (CMS) are very well known in industry. However it is also shown that their performance is very sensitive to changing production requirements. The detrimental effects of changing production requirements on the performance of CMS can be alleviated by "implementing better manufacturing cell designs", "employing effective part loading strategies" and "reconfiguration". This thesis proposes a decision support framework that provides solution strategies for manufacturing cell design, cell loading and reconfiguration problems. There are three main modules in the proposed framework, named as cell formation, loading and reconfiguration. Each module can handle multiple objectives and integrates several planning and design functions, by considering the capabilities of manufacturing resources. Reconfiguration decisions are made explicitly in the proposed framework by answering the questions "when to reconfigure?" and "how to reconfigure?”. In order to answer these questions, the modules of the proposed framework are interconnected. The cell formation module creates the initial set of cells. The loading module makes the 'part to cell assignment' and the scheduling in each production period. The reconfiguration module regenerates manufacturing cells, if the loading module can not find a satisfactory solution. The cell formation module solves the part-machine cell formation problem by simultaneously considering multiple objectives and constraints. Overlapping machine capabilities and generic part process plans are taken into account in the model formulation. A new approach for the evaluation of machine capacities is also presented. Results of the comparative study show that the proposed cell formation method gives better results than several other cell-formation procedures. The manufacturing cells are formed with improved capacity utilisation levels and reduced extra machine requirements. The method is also more likely to produce independent manufacturing cells with higher flexibility. The loading module solves the 'part to cell assignment' and 'cell scheduling' problems simultaneously for cellular manufacturing applications. Alternative parts to cell and machine assignments are considered by making use of generic part process plans in the model formulation. A parametric simulation model is developed to determine cell schedules for a given part assignment scenario. The proposed loading system can assess performance of the CMS in each production period. Therefore a decision can be made about its reconfiguration. It is also shown that the efficiency of CMSs facing changing production requirements can be improved and/or sustained by using the proposed loading strategy. The reconfiguration module takes the existing cell configuration as the current solution and generates a new solution from it, to enhance its performance. The model is objective driven and considers multiple objectives and constraints within a goal programming framework. The virtual cell concept is applied as the reconfiguration strategy. In the virtual cell approach the physical locations of machines are not changed, only cell memberships of machines are updated after reconfiguration. The results of the test studies showed that it is possible to improve the performance of CMS by reconfiguring it using virtual cells. The cell formation, loading and reconfiguration problems issues discussed in this thesis are combinatorially complex multiple objective optimisation problems. Additionally simulation is used to evaluate several of the objective functions used in the modelling of loading and reconfiguration problems. Classical optimisation algorithms have various limitations in solving such problems. Therefore Tabu Search (TS) based multiple objective optimisation algorithms are developed. The proposed TS algorithms are general-purpose and can also be used to solve other multiple objective optimisation problems. The results obtained from several test problems show the proposed TS algorithms to be very effective in solving multiple objective optimisation problems. More than 500/0 improvement in solution quality is obtained in some test problems.