A supervised machine learning approach to data-driven simulation of resilient supplier selection in digital manufacturing
A supervised machine learning approach to data-driven simulation of resilient supplier selection in digital manufacturing
复制标题
DOI：
10.1016/j.ijinfomgt.2019.03.004
复制
发表时间：
2019-12-01
影响因子：
21
通讯作者：
Ivanov, Dmitry
中科院分区：
文献类型：
--
作者：
Cavalcante, Ian M.;Frazzon, Enzo M.;Ivanov, Dmitry
近年来，人们对有弹性的供应商选择越来越感兴趣，其中很大一部分关注于预测中断概率。通过利用数字制造中的数据分析功能，我们概念化了一种完全不同的方法来分析不确定情况下供应商绩效的风险概况。数字化制造通过动态订单分配对供应商选择提出了独特的挑战，并为利用数字化数据来改进采购决策提供了新的机会。我们开发了一种结合模拟和机器学习的混合技术，并研究了其在弹性供应商选择中数据驱动决策支持方面的应用。我们认为准时交货是衡量供应商可靠性的一个指标，并探讨了围绕形成弹性供应绩效配置文件的条件。我们对供应商绩效的风险状况和弹性供应链绩效的概念进行了理论推导。我们表明，我们的方法可以有效地破译与弹性供应链绩效曲线的偏差与供应商绩效风险曲线之间的关联。结果表明，如果使用得当，有监督的机器学习和模拟的结合可以提高交付的可靠性。我们的方法在分析供应商基础和发现关键供应商或供应商组合时也是有价值的，这些供应商的中断会导致不利的业绩下降。这项研究的结果加深了我们对机器学习和模拟如何以及何时可以结合起来创建数字供应链双胞胎的理解，并通过这些双胞胎提高弹性。所提出的数据驱动的弹性供应商选择决策模型可以进一步用于供应链中断管理模型中风险缓解策略的设计、重新设计供应商基础或投资于最重要和最有风险的供应商。
There has been an increased interest in resilient supplier selection in recent years, much of it focusing on forecasting the disruption probabilities. We conceptualize an entirely different approach to analyzing the risk profiles of supplier performance under uncertainty by utilizing the data analytics capabilities in digital manufacturing. Digital manufacturing peculiarly challenge the supplier selection by the dynamic order allocations, and opens new opportunities to exploit the digital data to improve sourcing decisions. We develop a hybrid technique, combining simulation and machine learning and examine its applications to data-driven decision-making support in resilient supplier selection. We consider on-time delivery as an indicator for supplier reliability, and explore the conditions surrounding the formation of resilient supply performance profiles. We theorize the notions of risk profile of supplier performance and resilient supply chain performance. We show that the associations of the deviations from the resilient supply chain performance profile with the risk profiles of supplier performance can be efficiently deciphered by our approach. The results suggest that a combination of supervised machine learning and simulation, if utilized properly, improves the delivery reliability. Our approach can also be of value when analyzing the supplier base and uncovering the critical suppliers, or combinations of suppliers the disruption of which result in the adverse performance decreases. The results of this study advance our understanding about how and when machine learning and simulation can be combined to create digital supply chain twins, and through these twins improve resilience. The proposed data-driven decision-making model for resilient supplier selection can be further exploited for design of risk mitigation strategies in supply chain disruption management models, re-designing the supplier base or investing in most important and risky suppliers.