Optimal weighted L2 model reduction of delay systems (Q4955115)

Consider a linear time-invariant system with delay \[ G(s): \exp\{- sT_d\}G_1(s), \] where \(G_1(s)\) is a stable linear finite-dimensional system with minimal realization \((A,b,c)\) \[ G_1(s)= c(sI- A)^{-1}b. \] The optimal weighted \(L_2\) modal reduction problem is to find an \(m\)th order stable reduced-order system \(\widetilde G(s)\) and delay time \(\tau_d\geq 0\) \[ \widetilde G(s)= \exp\{-s\tau_d\}\widetilde G_1(s),\quad \widetilde G_1(s)= \widetilde c(sI- \widetilde A)^{-1}\widetilde b, \] such that the weighted \(L_2\) error \[ E= \|W(s)(\exp\{- sT_d\} G_1(s)- \exp\{- s\tau_d\}\widetilde G_1(s))\|_2 \] is minimized, where \[ W(s)= \left[\begin{matrix} A_w & b_w\\ C_w & d_w\end{matrix}\right]. \] The weighted \(L_2\) modal reduction error between \(G(s)\) and \(\widetilde G(s)\) is given from the solutions to special Lyapunov's equations.