A scalable uncertainty quantification and data assimilation framework for tracking stochastic fluid interfaces: application to civil and environmental engineering

用于跟踪随机流体界面的可扩展不确定性量化和数据同化框架：在土木和环境工程中的应用

• 批准号: 312528-2010
• 负责人: Sarkar, Abhijit
• 金额: $ 1.6万
• 依托单位: Carleton University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2013
• 资助国家: 加拿大
• 起止时间: 2013-01-01 至 2014-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=526559
• 关键词: scalable; uncertainty; quantification; data; assimilation

Many engineering problems involve complex dynamics of evolving flow front such as contaminant spills, floodwater and oil recovery operation. Therefore the ability to predict the evolving front is of great practical concern. For example, a continuously updated computer prediction with measurement data (airborne images and land-based sensor data) on how contaminant front or floodwater will propagate in few hours, days or weeks can help plan evacuations in order to save lives and properties. In this proposal, a computational model will be developed to continuously update prediction by injecting observational data into a running computer simulation for more accurate early warning on the motion of the moving flow front. To achieve this goal, the proposed research covers five key areas: (1) uncertainty quantification and sequential data assimilation techniques will be used to continuously update flow simulation models with measurement data to improve the confidence in their predictions; (2) the requisite high performance simulation models will exploit a domain decomposition method; (3) dynamic deformation of fluid interfaces will be tracked using the level set method; (4) a scalable integrated software suite will be developed to implement the above algorithms to take advantage of emerging high performance computing hardware; (5) this approach will be applied to contaminant tracking, flood forecasting and oil recovery. The proposed research is of benefit to water resource decision-makers (e.g. irrigation), ecosystem and environmental decision-makers (e.g. related to water and air quality), provincial and local emergency management (e.g. due to toxic chemical release) and oil recovery effort in Canada.

许多工程问题涉及复杂的流动前沿演化动力学，如污染物泄漏、注水和采油作业。因此，预测不断演变的前沿的能力是非常现实的问题。例如，使用测量数据(机载图像和陆基传感器数据)不断更新的计算机预测，关于污染物锋面或洪水在几小时、几天或几周内将如何传播，可以帮助计划疏散，以拯救生命和财产。在这项提议中，将开发一个计算模型，通过将观测数据注入运行的计算机模拟中来不断更新预测，以更准确地对移动的流动前沿的运动进行预警。为了实现这一目标，拟议的研究涵盖五个关键领域：(1)不确定性量化和顺序数据同化技术将用于不断更新流动模拟模型的测量数据，以提高其预测的可信度；(2)所需的高性能模拟模型将采用区域分解方法；(3)将使用Level Set方法跟踪流体界面的动态变形；(4)将开发一个可扩展的集成软件套件来实现上述算法，以利用新兴的高性能计算硬件；(5)该方法将被应用于污染物跟踪、洪水预报和石油开采。拟议的研究对水资源决策者(例如灌溉)、生态系统和环境决策者(例如与水和空气质量有关)、省和地方应急管理(例如由于有毒化学品释放)以及加拿大的石油开采工作都是有益的。