Various types of risks exist in a supply chain, and disruptions could lead to economic loss or even breakdown of a supply chain without an effective mitigation strategy. The ability to detect disruptions early can help improve the resilience of the supply chain. In this paper, the application of principal component analysis (PCA) and dynamic PCA (DPCA) in fault detection and diagnosis of a supply chain system is investigated. In order to monitor the supply chain, data such as inventory levels, market demands, and amount of products in transit are collected. PCA and DPCA are used to model the normal operating conditions (NOC). Two monitoring statistics, the Hotelling’s T² and the squared prediction error (SPE), are used to detect abnormal operation of the supply chain. The confidence limits of these two statistics are estimated from the training data based on the χ²-distributions. The contribution plots are used to identify the variables with abnormal behavior when at least one statistic exceeds its limit. Two case studies are presented—a multi-echelon supply chain for a single product that includes a manufacturing process and a gas bottling supply chain with multiple products. In order to validate the proposed method, supply chain simulation models are developed using the programming language Python 3.7, and simulated data is collected for analysis. PCA and DPCA are applied to the data using the scikit-learn machine learning library for Python. The results show that abnormal operation due to transportation delay, supply shortage, and poor manufacturing yield can be detected. The contribution plots are useful for interpreting and identifying the abnormality.

中文翻译：

---

使用主成分分析的供应链监控

供应链中存在各种类型的风险，如果没有有效的缓解策略，中断可能导致经济损失，甚至导致供应链崩溃。尽早发现中断的能力可以帮助提高供应链的弹性。本文研究了主成分分析（PCA）和动态PCA（DPCA）在供应链系统故障检测和诊断中的应用。为了监控供应链，收集了诸如库存水平，市场需求和运输中的产品数量之类的数据。PCA和DPCA用于模拟正常运行条件（NOC）。两项监测统计数据，Hotelling的T ²和平方预测误差（SPE）用于检测供应链的异常运行。这两个统计数据的置信度。基于该χ训练数据估计²-分布。当至少一个统计量超过其极限时，使用贡献图来识别具有异常行为的变量。提出了两个案例研究-包含制造过程的单一产品的多级供应链和具有多种产品的气体瓶装供应链。为了验证所提出的方法，使用编程语言Python 3.7开发了供应链仿真模型，并收集了仿真数据进行分析。使用适用于Python的scikit-learn机器学习库将PCA和DPCA应用于数据。结果表明，可以检测到由于运输延误，供应短缺和不良的制造良率而导致的异常操作。贡献图对于解释和识别异常很有用。