Generalized Derivatives of Differential–Algebraic Equations

详细信息    查看全文

• 作者：Peter G. Stechlinski ; Paul I. Barton
• 关键词：Generalized Jacobians ; Sensitivity analysis ; Nonsmooth analysis ; Optimization ; Parametric uncertainty
• 刊名：Journal of Optimization Theory and Applications
• 出版年：2016
• 出版时间：October 2016
• 年：2016
• 卷：171
• 期：1
• 页码：1-26
• 全文大小：593 KB
• 刊物主题：Calculus of Variations and Optimal Control; Optimization; Optimization; Theory of Computation; Applications of Mathematics; Engineering, general; Operations Research/Decision Theory;
• 出版者：Springer US
• ISSN：1573-2878
• 卷排序：171

文摘

Nonsmooth equation-solving and optimization algorithms which require local sensitivity information are extended to systems with nonsmooth parametric differential–algebraic equations embedded. Nonsmooth differential–algebraic equations refers here to semi-explicit differential–algebraic equations with algebraic equations satisfying local Lipschitz continuity and differential right-hand side functions satisfying Carathéodory-like conditions. Using lexicographic differentiation, an auxiliary nonsmooth differential–algebraic equation system is obtained whose unique solution furnishes the desired parametric sensitivities. More specifically, lexicographic derivatives of solutions of nonsmooth parametric differential–algebraic equations are obtained. Lexicographic derivatives have been shown to be elements of the plenary hull of the Clarke (generalized) Jacobian and thus computationally relevant in the aforementioned algorithms. To accomplish this goal, the lexicographic smoothness of an extended implicit function is proved. Moreover, these generalized derivative elements can be calculated in tractable ways thanks to recent advancements in nonsmooth analysis. Forward sensitivity functions for nonsmooth parametric differential–algebraic equations are therefore characterized, extending the classical sensitivity results for smooth parametric differential–algebraic equations.