Stochastic stability analysis of fault-tolerant control systems in the presence of noise (Q4540252)

The authors consider the general fault-tolerant control system with Markovian parameters (FTCSMP) represented by the multidimensional stochastic differential equation of the form \[ dx(t)=f\big(x(t),\eta(t),u(x(t),\Psi(t),t),t\big) dt +g\big(x(t),\eta(t),u(x(t),\Psi(t),t),t\big) dW(t). \tag{1} \] Here \(x(t)\in \mathbb{R}^n\) is interpreted as the system state, \(u(x(t),\Psi(t),t)\in \mathbb{R}^m\) is the input, \(\eta(t)\) and \(\Psi(t)\) are finite-state Markov processes representing an actuator failure process and a fault detection and isolation (FDI) processes, respectively; \(f\) and \(g\) are matrix-valued functions, and \(W(t)\) is a multidimensional Wiener process. Special attention is paid to FTCSMP of the linear form \[ \begin{multlined} dx(t)=\big(A(t)x(t)+B(\eta(t))u(t)\big)dt + D\big(x(t),\eta(t),t\big) dW_1(t) +\\ +E\big(u(t),\eta(t),t\big) dW_2(t) +F\big(\eta(t)\big) dW_3(t)\end{multlined}\tag{2} \] with \[ D\big( x(t),\eta(t),t\big) =\sum_i D_i\big( \eta(t) \big)x_i(t),\quad E\big( u(t),\eta(t),t\big) =\sum_j E_j\big(\eta(t)\big)u_j(t), \] \[ u(t)=u\big(x(t),\;\Psi(t),t\big) =-K\big(\Psi(t)\big)x(t). \] For FTCSMP (1), the authors obtain necessary conditions for the exponential stability in mean square. The result is used to obtain necessary conditions for the exponential stability in mean square of FTCSMP~(2) and sufficient conditions for its almost sure asymptotic stability in probability. Since these conditions are difficult to test and verify, \textit{testable} necessary and sufficient conditions for the exponential stability in mean square of FTCSMP~(2) are also derived. An illustrating numerical example is presented.