Integrating Reconfiguration Cost Into the Design of Multi-Period Scalable Reconfigurable Manufacturing Systems
Integrating Reconfiguration Cost Into the Design of Multi-Period Scalable Reconfigurable Manufacturing Systems
复制标题
DOI：
10.1115/1.2383196
复制
发表时间：
2007-02
影响因子：
4
通讯作者：
P. Spicer;H. J. Carlo
中科院分区：
文献类型：
--
作者：
P. Spicer;H. J. Carlo
可重构制造系统（RMS）是专门为通过系统重构来适应生产能力变化而设计的，称为可扩展-RMS。可扩展 RMS 随着时间的推移而变化的系统配置集称为其配置路径。本文研究了如何确定可扩展 RMS 的最佳配置路径，从而在已知需求的有限范围内最大限度地减少投资和重新配置成本。首先，提出了一个实用的成本模型来计算两个可扩展 RMS 配置之间的重新配置成本。该模型涵盖了劳动力成本、容量损失成本以及由于系统重新配置和升级而导致的投资/抢救成本。其次，本文提出了使用动态规划（DP）的多周期可伸缩-RMS 的最优解模型。第三，提出了组合的整数规划/动态规划（IP-DP）启发法，允许用户控制DP考虑的系统配置的数量，以减少求解时间，同时仍然提供合理的解决方案。涉及两级和三级可扩展 RMS 的数值问题可以使用 DP 和 IP-DP 方法来解决。实验结果表明，DP 方法虽然是最优的，但对于大型问题来说计算效率不高。然而，组合的 IP-DP 方法以更少的计算量提供了合理的结果。
A reconfigurable manufacturing system (RMS) that is designed specifically to adapt to changes in production capacity, through system reconfiguration, is called a scalable-RMS. The set of system configurations that a scalable-RMS assumes as it changes over time is called its configuration path. This paper investigates how to determine the optimal configuration path of a scalable-RMS that minimizes investment and reconfiguration costs over a finite horizon with known demand. First, a practical cost model is presented to compute the reconfiguration cost between two scalable-RMS configurations. This model comprehends labor costs, lost capacity costs, and investment/salvage costs due to system reconfiguration and ramp up. Second, the paper presents an optimal solution model for the multiperiod scalable-RMS using dynamic programming (DP). Third, a combined integer programming/dynamic programming (IP-DP) heuristic is presented that allows the user to control the number of system configurations considered by the DP in order to reduce the solution time while still providing a reasonable solution. Numerical problems involving a two-stage and a three-stage scalable-RMS are solved using the DP and IP-DP methodologies. Experimental results suggest that the DP approach, although it is optimal, is not computationally efficient for large problem sizes. However, the combined IP-DP approach offers reasonable results with much less computational effort.