Design of manufacturing systems for product variety using cluster and statistical analysis

该研究计划的范围是制造系统的优化，以支持产品的多样性和可持续性。 为了在市场上保持竞争力，原始制造商（OM）需要满足客户更多的产品选择，并满足环境法规和社会期望。 在此背景下，虽然形式优化模型已广泛用于系统设计中，但其通常面临模型和计算复杂性的挑战（例如，大量整数变量和收敛问题）。 而不是采用优化方法，本研究将调查和开发集群和统计技术的制造系统的设计。 新颖之处在于开发了一种自下而上的方法来系统地解决复杂的结构决策，其目标是在解的收敛性和计算时间上具有更好的性能，以及与元启发式算法具有可比的最优性能。 拟议的研究计划分为两个主题。 第一个主题是单元制造系统（CMS）的聚类算法的发展。 一个研究目标是提高能力的层次聚类（HC）搜索最佳的结构决策CMS的设计。 方法学方法包括耦合分布和可逆HC的统计分析，以避免过早组的锁定。 此外，我们还将关注鲁棒性和模块化问题。 在鲁棒性方面，应用聚类方法来连接相关更改并有目的地部署资源以减少不必要的传播影响。 在模块化方面，本研究的研究理念是利用集群概念来匹配产品结构和制造系统的模块化，以增强系统对市场变化的灵活性。 第二个主题是设计可持续发展的闭环制造系统。 总体目标是支持OM参与产品的寿命终止管理。 考虑三个方面：装配、拆卸和重新装配。 “组装”指的是正向生产，而“拆卸”和“重新组装”指的是考虑报废选择和产品回收的反向循环。 在这个应用程序中，聚类方法被用来配置这个网络系统，以提高可持续性的措施。 从研究到实践，我们将使开发的方法和知识以实用的专门知识的形式进入加拿大制造业（例如，如何最小化变更传播和设计健壮的系统）。 对于零部件供应商，他们可以学习如何提高制造灵活性，以获得更多的合同机会。 对于OM，他们可以通过设计一个闭环网络来管理首次使用后的材料流动，从而促进可持续性。

The scope of this research program is the optimization of manufacturing systems to support product variety and sustainability. To stay competitive in the market, original manufacturers (OM) need to satisfy customers with more product options and meet the environmental regulations and societal expectations. In this context, while formal optimization models have been widely used in system design, they often face the challenges of model and computational complexity (e.g., high number of integer variables and convergence issues). Instead of employing optimization methods, this research will investigate and develop clustering and statistical techniques for the design of manufacturing systems. The novelty is to develop a bottom-up approach to address complex structural decisions systematically with the target of better performance on solution convergence and computing time, as well as comparable optimality performance with metaheuristic algorithms. The proposed research program is organized into two themes. The first theme is the development of clustering algorithms for cellular manufacturing systems (CMS). One research goal is to improve the capability of hierarchical clustering (HC) to search for optimal structural decisions in the design of CMS. The methodological approach includes statistical analysis of coupling distributions and reversible HC to avoid the lock-in of premature groups. Also, we will focus on the issues of robustness and modularity. Regarding robustness, the clustering approach is applied to concatenate relevant changes and deploy resources purposely to reduce unnecessary propagation effects. Regarding modularity, the research idea is to match the modularity of product architecture and manufacturing systems using clustering concepts to enhance system flexibility for market changes. The second theme is the design of a closed-loop manufacturing system for sustainability. The general goal is to support OM to participate in the product's end-of-life management. Three aspects are considered: assembly, disassembly, and reassembly. While “assembly” is referred to the forward production, “disassembly” and “reassembly” are referred to the backward loop that considers the end-of-life options and product recovery. In this application, the clustering approach is used to configure this network system to improve sustainability measures. From research to practice, we will make the developed methodology and knowledge accessible to the Canadian manufacturing sector in a form of practical know-how (e.g., how to minimize change propagation and design robust systems). For component suppliers, they can learn how to enhance their manufacturing flexibility for more contract opportunities. For OM, they can promote sustainability by designing a close-loop network that manages the flow of materials after their first usage.