Scalable Parallel Multilevel Algorithms for the Solution and Optimization of Partial Differential Equations

9973276该建议侧重于几个基本的计算问题，涉及并行隐式解决偏微分方程所产生的最优控制和变分制定的物理问题。 研究的目标包括但不限于以下方面： (1)阐述和分析可证明有效的，鲁棒的和可扩展的多级算法，用于偏微分方程定义的系统的优化和求解。 (2)包含上述算法的软件的开发及其在制造、工程和科学界的传播。拟议的软件将扩展由UC San Diego研究者开发的现有代码PLTMG、MC和SNOPT的范围和有效性。 (3)吸引高级研究生进入这一研究领域，通常涉及科学和工程中出现的系统的分析和计算研究。研究人员结合了优化，偏微分方程数值方法和并行计算的专业知识。 该研究计划旨在为研究生提供广泛的教育经验，在这种环境中，科学计算的这些不同方面被整合。 该研究的一个关键目标是将新成果转移到制造业，工程和科学界。 数学软件是最有效的技术转移手段之一。 相反，有效算法开发的一个关键组成部分是与面临现实问题的工程师和科学家密切联系。 研究人员将继续并扩大与应用科学和技术其他领域的研究人员的合作，包括制造，环境建模，电力优化，生物分子建模，计算机图形学，固体力学，航空航天和金融。 所有这些应用程序都受益于现有的算法和软件，并为之做出了贡献。

9973276This proposal focuses on several fundamental computational issues involved in the parallel implicit solution of partial differential equations arising from optimal control and the variational formulation of physical problems. The goals of the research include, but are not limited to, the following: (1) The formulation and analysis of provably efficient, robust and scalable multilevel algorithms for the optimization and solution of systems defined by partial differential equations. (2) The development of software embodying the above algorithms and its dissemination within the manufacturing, engineering and scientific community. The proposed software will extend the scope and effectiveness of the existing codes PLTMG, MC, and SNOPT that have been developed by the investigators at UC San Diego. (3) The attraction of advanced graduate students into this research area, which generally concerns the analysis and computational study of systems arising in the sciences and engineering.The investigators have combined expertise in optimization, numerical methods for partial differential equations, and parallel computation. This research program is intended to provide graduate students with a broad educational experience in an environment in which these different aspects of scientific computation are integrated. A key aim of the research is the transfer of new results to the manufacturing, engineering and scientific community. Mathematical software is one of the most effective means of technology transfer. Conversely, a crucial component of effective algorithm development is close association with engineers and scientists facing real-world problems. The investigators will continue and broaden their collaboration with researchers in other areas of applied science and technology, including manufacturing, environmental modeling, electrical power optimization, biomolecular modeling, computer graphics, solid mechanics, aerospace and finance. All these applications have benefited from and contributed to existing algorithms and software.