A flow-on-manifold formulation of differential-algebraic equations (Q1692090): Difference between revisions

This technical paper presents the thesis of the author. It establishes a flow formulation for implicit real DAEs \[ F(t,x,\dot{x}) = 0 \] and its application to the dynamics of DAEs. Roughly speaking, once the manifold of constraints has been obtained, here the derivative array method (Brenan, Campbell, Petzold 1996) is used, an ODE on a constraint manifold \[ \hat{F}(t,x,\dot{x}) = 0 \] is locally proved to be equivalent to the original system. This is theoretically known since e.g. the work of \textit{P. J. Rabier} and \textit{W. C. Rheinboldt} [Differ. Integral Equ. 4, No. 3, 563--582 (1991; Zbl 0722.34004); J. Differ. Equations 109, No. 1, 110--146 (1994; Zbl 0804.34004)]. However, the author gives an explicit representation of the flow, using the Moore-Penrose inverse on one hand and the concept of time-varying submanifold on the other hand, to take into account the non-autonomous DAEs. The central theorem (Theorem 3) of this paper gives the local equivalence of the original system \(F(t,x,\dot{x}) = 0\) to a decomposed one: \[ \begin{cases} \dot{x_d} = h_{MP}(t, x_d), \\ x_a = g_{MP}(t, x_d), \\ x_d(t_0) = P_{MP}(t_0, x_0, \dot{x}(t_0))x_0. \end{cases} \] The Moore-Penrose inverse, locally defined using the derivative \(\partial_{\dot{x}}\hat{F}\), induces a projector \(P_{MP}\) which allows to separate the algebraic and differential components of the unknown \(x\): \[ x_a = P^{\bot}_{MP}(t,x,\dot{x})x, \;\; x_d =P_{MP}(t,x,\dot{x})x. \] This choice of the dependent/independent coordinates \(x_a/x_d\) is shown to be locally constant in a time interval \([t_0, t'_0)\) around a generic initial condition in Lemma 7, which allows an effective approach. In a specific section devoted to linear DAEs \[ E(t)\dot{x}-A(t)x-b(t)=0, \] the flow is explicited through an integral formula, generalizing Duhamel's formula (Theorem 6). The author shows how to apply the Moore-Penrose inverse to the linearization of the flow near a solution, in order to study the stability and the positivity property. This theory gives a global understanding on how to apply practically the methods developed for dynamical systems to nonlinear DAEs. It could be used to improve software packages on solving implicit DAEs.