Deterministic Global Optimization in Systems Engineering

【摘要】pi: Christodoulos a . Floudas机构：普林斯顿大学，提案号：0002777本研究的主要目标是为各种化学工程设计和合成问题中的全局优化问题以及计算化学和生物学中出现的核磁共振结构精化问题发展理论、算法和计算技术。这些方法将补充化学过程系统的数学建模工作，并有助于分子、多肽和蛋白质的设计研究。PI计划研究(1)将他的全局优化框架aBB扩展到参数估计和控制中出现的微分代数模型；(2)针对不确定设计中出现的一般双层和多层非线性模型，以及过程综合、设计和操作的相互作用，建立了一种新的基于全局优化的方法；(3)过程综合中出现的凸和非凸混合整数非线性优化模型的新方法的发展；(4)核磁共振实验提供的具有稀疏约束的多肽结构预测；(5)开发全局优化工具，并研究上述(1)- (4)aBB全局优化方法的分布式计算框架。这些将应用于几个类别的化学过程合成，设计，控制和计算化学问题。本文的创新之处包括：(a)对具有微分和代数约束的模型采用新的全局优化方法；(b)双层和多层非线性模型的新的全局优化方法；(c)结合强切割面和分支定界原理求解混合整数非线性优化问题的新方法；(d)通过计算化学和生物学中出现的核磁共振约束来确定肽和蛋白质结构的新方法；(e)计划工作的所有任务的顺序和分布式计算工具。

ABSTRACTPI: Christodoulos A. Floudas Institution: Princeton UniversityProposal Number: 0002777The primary objective of this research is to develop theoretical, algorithmic and computational techniques for global optimization problems that arise in a variety of chemical engineering design and synthesis problems, as well as NMR structure refinement problems that arise in computational chemistry and biology. These methods will complement efforts in the mathematical modeling of chemical process systems and aid the search in the design of molecules, peptides, and proteins. The PI plans to investigate (1) the extension of his global optimization framework, aBB, to differential algebraic models that arise in parameter estimation and control; (2) the development of a new global optimization-based approach for general bilevel and multilevel nonlinear models that arise in the design under uncertainty, and the interaction of process synthesis, desing and operations; (3) the development of novel approaches for convex and nonconvex mixed integer nonlinear optimization models that arise in process synthesis; (4) the structure prediction of peptides with the presence of sparse restraints provided by NMR experiments; and (5) development of global optimization tools and the study of a distributed computing framework for the aBB global optimization approach for (1) - (4) above.These will be applied to several classes of chemical process synthesis, design, control and computational chemistry problems. The innovative features of this work consist of: (a) new global optimization methods for models with differential and algebraic constraints; (b) new global optimization approaches for bilevel and multilevel nonlinear models; (c) new approaches that combine strong cutting planes and branch and bound principles for mixed integer nonlinear optimization problems; (d) new methods for the determination of the structure of peptides and proteins through the NMR restraints that arise in computational chemistry and biology; and (e) sequential and distributed computational tools for all tasks of the planned work.