Agile Control System for Reconfigurable Manufacturing Systems

生产系统的配置影响着生产系统的性能，目前已有大量关于生产系统配置的研究，如装配线平衡、单元形成等。在特定的配置下，生产系统的性能受生产排序和缓冲区分配的影响。在多阶段生产库存系统中，订货系统确定每个阶段的订货时间和订货量，并分配缓冲区，而准时制订货系统则根据实际需求确定每个阶段的订货时间和订货量。在以往的文献中，大部分的JIT订货系统，稳定的需求变化被假定，和不稳定的需求变化的影响从来没有被分析。在此背景下，针对JIT订货系统提出了一些反应式系统。在反应式JIT订购系统中，检测需求的不稳定变化，并且调整缓冲区大小作为对检测到的变化的反应。然而，仅仅通过调整缓冲区大小，适应性是有限的。最近，可重构制造系统（RMS）已经被指出，并且可能实现能够响应于环境变化来调整配置的系统。因此，本报告介绍了由JIT订购系统控制的RMS中的动态生产线平衡和缓冲区分配。在所提出的系统中，不稳定的需求变化检测的指数加权移动平均（EWMA）图，和生产系统的配置，即，阶段和缓冲区大小的数量，改变响应于检测到的不稳定的变化，和工作元素分配到每个阶段的生产线平衡。此外，神经网络被应用于提高性能，以检测需求的不稳定变化。通过仿真实验，分析了该系统的有效性，并与已有的系统进行了比较。

The configuration of production system affects the performance of the production system, and there is a lot of research on configuring production systems, such as assembly line balancing and cell formation. Under the specified configuration, the performance of production system is affected by production ordering and buffer allocation. Ordering systems determine the timing and quantity in order release for each stage of multi-stage production inventory systems with allocated buffer, and just-in-time (JIT) ordering systems determine them on the basis of the actual demand. In most of the previous literature on JIT ordering systems, stable changes in demand have been assumed, and the influence of unstable changes in demand had never been analyzed. Based on the background, some reactive systems have been proposed for the JIT ordering system. In the reactive JIT ordering system, unstable changes in demand are detected, and the buffer size is adjusted as a reaction to the detected changes. However, only by adjusting the buffer size, the adaptability is limited. Recently, reconfigurable manufacturing systems (RMSs) have been pointed out, and systems which can adjust the configuration in response to environmental changes are likely to be realized. Therefore, this report introduces dynamic line balancing and buffer allocation in a RMS controlled by the JIT ordering system. In the proposed system, unstable changes in demand are detected by the exponentially weighted moving average (EWMA) charts, and the configuration of production system, that is, the number of stages and buffer size, are altered in response to the detected unstable changes, and work elements are allocated to each stage by line balancing. Also, a neural network is applied to improve the performance to detect unstable changes in demand. The effectiveness of the proposed system is analyzed and compared with that of the previous system by simulation experiments.