ITR/AP: Collaborative Research: Sampling Methods for Optimization and Control of Subsurface Contamination

该项目的目标是(1)同时为各种地下污染物修复问题设计优化方法以及地下水流和输送模拟器；(2)以一种利用优化和模拟的并行性的方式实施优化-模拟结合，包括基于网格的搜索-轮询类型方法的实现的一些想法；(3)将新算法应用于两个示范问题；(4)在这种多学科背景下培养新一代学生。其中两位PI，Dennis和Kelley，在广泛的优化技术方面拥有专业知识，包括用于平滑问题的基于梯度的方法，以及用于无法获得梯度的问题的采样方法。凯利和米勒在多相流模拟的数值方法方面有着长期的合作。PI将利用这一专业知识，同时设计模拟器和优化算法。软件领域将出现以IT为中心的活动。Focus和IFFCO这两个代码正在由主要调查人员编制。其中最成熟的IFFCO是一种隐式过滤优化代码，将从PVM移植到MPI，打包以便于安装和测试，记录为一本书，并以最终形式发布。PIS还将继续制定Focus，这是替代管理框架思想的实施。修复地下污染是世界范围内的一个重要问题。其目标是以最低的成本清理或使地下污染物不能移动，如地下储罐泄漏的燃料或泄漏的滚筒产生的工业废物。必须做出管理决策，而糟糕决策的潜在经济惩罚是巨大的。地下系统的数学模型经常被用来帮助做出这样的决策，这些模型越来越多地与优化方法联系在一起，以帮助设计和管理决策。地下补救战略涉及诸如使用恢复方法、选择安装井等设备的位置以及从这些井中抽取或注入这些井的速度等决定。在模拟中，与基本建模和数值方法的选择有关的公开科学问题需要作为优化补救方法的软件设计的一部分加以解决。这三位PI在多相流的模拟、建模和实验工作方面拥有专业知识，在优化方法方面也拥有广泛的专业知识。该项目的目标是同时设计模型、模拟器和优化方法，目标是基于更精确的模型进行更有效的优化设计。

The objectives of this project are (1) to simultaneously design optimization methods and subsurface flow and transport simulators for a variety of subsurface contaminant remediation problems, (2) to implement the optimization-simulation combination in a way that exploits parallelism in both the optimization and simulation, including some ideas for grid-based implementations of search-poll type methods, (3) to apply the new algorithms to two demonstration problems, (4) to train a generation of students in this multidisciplinary setting. Two of the PIs, Dennis and Kelley, have expertise in a wide range of optimization technology, both gradient-based methods for smooth problems and sampling methods for problems for which gradients are not available. Kelley and Miller have a long-standing collaboration in numerical methods for simulation of multi-phase flow. The PIs will exploit this expertise and design the simulators and optimization algorithms simultaneously. There will be IT-centered activity in software. Two codes, FOCUS and IFFCO, are under development by the principal investigators. The most mature of these, IFFCO an implicit filtering optimization code, will be ported to MPI from PVM, packaged for ease of installation and testing, documented as a book, and released in final form. The PIs will also continue to develop FOCUS, an implementation of the surrogate management framework ideas.Remediation of subsurface contamination is an important issue world-wise. The goal is to clean up, or render immobile, underground contaminants such as fuel spills from underground tanks or industrial waste from leaking drums, and do this at minimum cost. Management decisions must be made, and the potential financial penalty of bad decisions is great. Mathematical models of subsurface systems are often used to assist in making such decisions and these models are increasingly linked with optimization approaches to aid design and management decision making. Subsurface remediation strategies involve decisions such as the restoration approach to use, the selection of locations to install devices, such as wells, and the rates of extraction from or injection into these wells. There are open scientific problems in simulation related to fundamental modeling and choice of numerical methods that need to be addressed as part of the design of software to optimize the remediation methods. The three PIs have expertise in simulation, modeling, and experimental work in multi-phase flows as well as broad expertise in optimization methods. The objectives of this project are to design models, simulators, and optimization methods simultaneously with the goal of more efficient optimal design based on more accurate models.