The method of least squares in the theory of Noetherian differential-algebraic boundary-value problems (Q2011283)

The author makes use of a generalized Moore-Penrose inverse to give pseudo-solutions of an undetermined differential-algebraic and linear boundary value problem and also systems with lumped delay, continuing his work on similar topics, published since 2008. As these systems generally have an infinite-dimensional space of solutions, the proposal is to find the one minimizing an associated least square constraint (including the squared error for the DAE and the boundary condition). Theorem 1 states the existence of this solution (under regularity conditions) for boundary value-constrained DAE systems like \[ A(t)z'(t) = B(t)z(t)+f(t), \,\, lz(.) = \alpha\in \mathbb R^k, \] while Theorem 2 addresses the problem \[ A(t)z'(t) = B(t)z(t)+C(t)z(t-\Delta)+f(t), \,\, \Delta \leq t \leq T=p\Delta \] with boundary condition \[ lz(.) = \alpha\in \mathbb R^k. \] In both cases the solution is developed explicitely, using a generalized Green operator for the not-delayed case. The construction for the second case is a continuous map \(C^1\) on the intervals \((k\Delta,(k+1)\Delta)\), \(k=0,\dots,(p-1)\). The paper is clearly written with fully covered examples, e.g. periodic DAEs with the boundary condition \[z(0)=z(2\pi).\] The reader may just expect an example from real modeling, for systems with delay.