Data-driven decision support for in-service and post-hazard infrastructure management

针对在役和灾后基础设施管理的数据驱动决策支持

• 批准号: RGPIN-2016-06405
• 负责人: Goulet, JamesAlexandre
• 金额: $ 1.75万
• 依托单位: École Polytechnique de Montréal
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2019
• 资助国家: 加拿大
• 起止时间: 2019-01-01 至 2020-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=679617
• 关键词: Data; driven; decision; support; service

Around the world, public infrastructure maintenance has been underfunded for years. Canada is no exception, the aging of existing infrastructure has lead to an alarming number of bridges that are in critical condition. In addition, cities located on the west coast of Canada and along the Saint-Lawrence Basin face the threat of earthquakes which could have catastrophic impacts on our already vulnerable infrastructure. Public infrastructures are essential components of economic development. Yet, society cannot afford at short and medium term, to mitigate the risks of ageing infrastructure and hazard vulnerability by strengthening or replacing all deficient structures; we face the challenge of stretching even further the life of the deficient infrastructure. A path to mitigating these challenges is data-driven technologies for supporting management decisions regarding the condition and safety of structures during their service life and following extreme natural hazards such as earthquakes. Sensing technologies are now inexpensive and are commercially available for measuring structural behaviours such as displacement, strain, tilt, acceleration, and environmental effects. Comparatively, our capacity to interpret data is still in its infancy. Factors such as the complexity of the interactions between a structure and its operating environment, the errors caused by the operation in harsh conditions, and the sheer volume of data acquired are all sources of false alarms. Our capacity to autonomously filter out false alarms currently limits our capacity at detecting anomalies. ***The contributions envisaged are part of a long-term research program; its goal is to provide society with solutions for big-data analytics enabling the sustainable management of ageing infrastructure through massively-distributed structural health monitoring (SHM). The three objectives of this research proposal are (O1) create a generic framework for modelling the in-service behaviour of structures, detecting anomalies, and autonomously taking data-driven decisions; (O2) Create a data-driven decision support system for rapid post-earthquakes structural safety assessment; (O3) Develop an open-source computational platform enabling empirical studies, collaborative research, and technology transfer. Note that for Objective 1, “autonomous data-driven decisions” refers to (a) the capacity of the framework to learn by itself the baseline response of a structure, (b) when a potential anomaly is detected, to decide by itself if this deviation from the baseline response requires the attention of an engineer, and (c) adapt autonomously to changing conditions caused, for example, by degradation, climate changes, or maintenance activities. All three objectives share the same methodological background relying on specialized structural engineering knowledge, Bayesian methods, and Machine Learning.**

在世界各地，公共基础设施维护资金多年来一直不足。加拿大也不例外，现有基础设施的老化导致了数量惊人的桥梁处于危急状态。此外，位于加拿大西海岸和圣劳伦斯盆地沿线的城市面临着地震的威胁，这可能对我们本已脆弱的基础设施造成灾难性影响。公共基础设施是经济发展的重要组成部分。然而，在中短期内，社会负担不起通过加强或更换所有有缺陷的结构来缓解基础设施老化和危险脆弱性的风险；我们面临着进一步延长有缺陷的基础设施的寿命的挑战。缓解这些挑战的一个途径是数据驱动的技术，用于支持关于结构在其使用寿命期间和在地震等极端自然灾害之后的状况和安全的管理决策。传感技术现在价格低廉，可用于测量结构行为，如位移、应变、倾斜、加速度和环境影响。相比之下，我们解释数据的能力仍处于初级阶段。结构与其运行环境之间相互作用的复杂性、恶劣条件下的操作造成的错误以及获取的数据量等因素都是误报的来源。我们自主过滤错误警报的能力目前限制了我们检测异常的能力。*设想的贡献是长期研究计划的一部分；其目标是为社会提供大数据分析解决方案，通过大规模分布的结构健康监测(SHM)实现老化基础设施的可持续管理。这项研究建议的三个目标是(O1)创建一个通用框架，用于对结构在使用中的行为进行建模、检测异常和自主做出数据驱动的决策；(O2)创建一个数据驱动的决策支持系统，用于地震后结构安全的快速评估；(03)开发一个能够进行实证研究、协作研究和技术转让的开放源码计算平台。注意，就目标1而言，“自主数据驱动的决策”是指(A)框架自身学习结构基线响应的能力，(B)当检测到潜在异常时，自行决定这种偏离基线响应是否需要工程师注意，以及(C)自主适应由退化、气候变化或维护活动等引起的条件变化。所有这三个目标都有相同的方法学背景，依赖于专门的结构工程知识、贝叶斯方法和机器学习。