S^3 Disease Surveillance for Structures and Systems

高价值工程资产成本的主要贡献者之一是维护成本。让一架飞机停止服役进行检查意味着损失收入。然而，考虑到成本和安全问题，另一种选择——允许损坏使飞机退出服役——是更不可取的。就海上风力发电场而言，为了更换一个75米长的叶片而计划外访问一个偏远地点的成本几乎无法想象。如果可以采用基于状态的维护方法，通过永久性传感器持续监测感兴趣的结构，并且数据处理算法在损坏发生时提醒所有者或用户，则可以在不牺牲安全性的情况下优化维护计划以降低成本。如果检测到早期损坏，维修而不是更换可能是一个可行的选择。不幸的是，现代结构的复杂性及其运行环境的挑战性使得开发能够检测和识别早期损伤的数据处理算法非常困难。与这些问题有关的学科——结构健康监测（SHM）——存在严重问题，阻碍了工业采用该技术。结构的复杂性使分析变得困难；然而，SHM的一种变体——基于数据的方法——在这方面显示出了希望。在这种情况下，人们直接从结构的数据中学习，使用模式识别技术来诊断不同程度的损伤。遗憾的是，基于数据的SHM有其自身的问题；首先，大多数模型识别方法都需要测量结构在所有可能的损伤状态下的数据。以飞机这样的结构为例——以A380为例——人们根本无法想象为了收集数据而损坏一架飞机，更不用说损坏多架了。幸运的是，如果一个人仅仅对是否存在损害感兴趣，这可以只使用健康状况的数据来完成。一个是构建结构健康状态的图像，然后监视与该状态的偏差。这就提出了基于数据的SHM的第二个主要问题；如果一个人正在监测结构的变化，他不希望因为环境或操作条件的良性变化而发出警报；这些被称为“混杂影响”。解决方案可能存在于医疗保健信息学社区中。过去20年来出现了一个称为“综合征监测”（SS）的领域，涉及通过监测人群快速发现疾病暴发。从非处方药销售到健康咨询网站的点击量，数据本身可能非常不同。这些数据被融合在一起并进行分析，以提供公共卫生的时空图景，并且类似于用于SHM的警报算法可以用来警告卫生保健专业人员流行病可能即将到来。这些想法甚至被嵌入到软件中，最典型的例子是ESSENCE II系统，它监视着美国的三个州。目前的提案旨在开发一种具有高价值资产故障快速检测和定位能力的工程结构SS系统。本文提出的基于人口的SHM方法有可能解决上面讨论的两个问题。如果对许多结构进行监测，结构之间的推断可以潜在地避免对单个结构的非常详细的知识的需要。当结构随着时间的推移而失效时，关于损伤状态的知识就会建立起来。对于第二个问题，SS系统一直在处理混杂影响，可以为基于数据的SHM的新算法提供灵感。如ESSENCE II的情况；该系统将嵌入到软件中，以便结构的多个操作人员可以从基于人口的系统的诊断能力中获得最大利益。

One of the main contributors towards the cost of high-value engineering assets is the cost of maintenance. Taking an aircraft out of service for inspection means loss of revenue. However, the alternative - allowing damage to remove the aircraft from service - is much more undesirable with cost and safety being issues. In terms of an offshore wind farm, the cost of an unscheduled visit to a remote site to potentially replace a 75m blade hardly bears thinking about. If one can adopt a condition-based approach to maintenance where the structure of interest is monitored constantly by permanent sensors and data processing algorithms alert the owner or user when damage is developing, one can optimise the maintenance program for cost without sacrificing safety. If incipient damage is detected, repair rather than replacement can be a viable option.Unfortunately, the complexity of modern structures together with the challenging environments in which they operate makes it very difficult to develop data-processing algorithms which can detect and identify incipient damage. The discipline concerned with these problems - structural health monitoring (SHM) - suffers from serious problems which have prevented uptake of the technology by industry. The structural complexity makes analysis difficult; however, one variant of SHM - the data-based approach - shows promise in this respect. In this case one learns directly from data from the structure using pattern recognition techniques to diagnose different levels of damage. Sadly, data-based SHM has its own problems; the first is that most pattern recognition approaches to SHM require one to measure data from the structure in all possible states of damage. In the case of a structure like an aircraft - consider the A380 - it is simply not conceivable that one should damage a single one for data collection purposes, let alone many. Fortunately, if one is only interested simply in whether damage is present or not, this can be accomplished using only data from the healthy condition. One builds a picture of the healthy state of the structure and then monitors for deviations from this state. This raises the second major issue with data-based SHM; if one is monitoring the structure for changes, one does not wish to raise an alarm because of a benign change in its environmental or operational conditions; these are termed 'confounding influences'.The solution may lie within the healthcare informatics community. A field called 'syndromic surveillance' (SS) has arisen over the last 20 years concerned with fast detection of disease outbreaks by monitoring human populations. The data themselves can be very different, from over-the-counter medicine sales to numbers of hits on health advice websites. The data are fused together and analysed to give a spatio-temporal picture of public health and alerting algorithms similar to the ones used for SHM can be used to warn healthcare professionals that an epidemic may be on the way. The ideas have even been embedded in software, the prime example being the ESSENCE II system which keeps a watchful eye over three US states.The current proposal aims to develop a SS system for engineering structures with the capability of fast detection and location for faults on high-value assets. The population-based approach to SHM proposed here has the potential to solve the two problems discussed above. If many structures are monitored, inferences between structures can potentially avoid the need for very detailed knowledge of individual structures. As structures fail with time, the knowledge of damage states will build. In terms of the second problem, SS systems have always dealt with confounding influences and can provide inspiration for new algorithms for data-based SHM. As in the case of ESSENCE II; the system will be embedded in software so that multiple operators of structures can derive maximum benefit from the diagnostic capability of the population-based system.