Learning Short and Long Term Failure Patterns from Massive Network Failure Data 从大量网络故障数据中学习短期和长期故障模式

光华讲坛——社会名流与企业家论坛第6541期

主 题：Learning Short and Long Term Failure Patterns from Massive Network Failure Data 从大量网络故障数据中学习短期和长期故障模式

主讲人：新加坡国立大学 叶志盛教授

主持人：管理科学与工程学院 肖辉教授

时间：6月4日 9:30-11:30

举办地点：柳林校区通博楼D301会议室

主办单位：管理科学与工程学院 科研处

主讲人简介：

叶博士获得清华大学材料科学与工程和经济学联合学士学位(2008)。获新加坡国立大学博士学位。他目前是新加坡国立大学工业系统工程与管理系教授兼系主任。他的研究领域包括可靠性建模、估计和优化、基于状态的维护和数据驱动的决策。他的研究成果发表在可靠性、统计学和运营管理领域的旗舰期刊上，包括Technometrics、JQT、IISE Trans、ITR、RESS、JRSS-B、JASA、Biometrika、JMLR、MSOM和POMS。

内容简介：

Many lifeline infrastructure systems consist of thousands of components configured in a complex directed network. Disruption of the infrastructure constitutes a recurrent failure process over a directed network. Statistical inference for such network recurrence data is challenging because of the large number of nodes with irregular connections among them. In this talk, we focus on both short term cascading failures and long term ageing failures. Repair of a pipe might generate shocks to neighbouring pipes and cause short term cascading failures. Understanding the short-term cascading failure is important for the utility to allocate additional resources to monitor the neighbouring pipes after a repair. On the other hand, understanding long-term failures is helpful in risk analysis of the whole pipe network and prioritizing replacements of old pipes. Statistical modelling of the two failure modes are extremely challenging because of the large pipe network and the huge failure data set. We develop novel statistical methods that are computationally tractable to fit the data. Applying the methods to a large data set from the Scottish Water network, we demonstrate the usefulness of our models in aiding operation management and risk assessment of the water utility.

许多生命线基础设施系统由数千个组件组成，这些组件配置在一个复杂的定向网络中。基础设施的中断构成了有向网络上反复出现的故障过程。由于大量节点之间具有不规则连接，因此对此类网络复发数据的统计推断具有挑战性。在本次演讲中，我们将重点关注短期级联故障和长期老化故障。管道的修复可能会对邻近的管道产生冲击，并导致短期级联故障。了解短期级联故障对于公用事业公司在维修后分配额外资源来监测邻近管道非常重要。另一方面，了解长期故障有助于对整个管网进行风险分析，并优先更换旧管道。由于管网庞大，故障数据集庞大，两种失效模式的统计建模极具挑战性。我们开发了新的统计方法，这些方法在计算上易于处理以拟合数据。将这些方法应用于苏格兰水网的大型数据集，我们证明了我们的模型在帮助水务公司的运营管理和风险评估方面的