Exponential stability of a nonlinear distributed parameter system (Q1978084)

This paper studies the spatial one-dimensional model of a fixed-bed bioreactor consisting of a pair of nonlinear partial differential equations: \[ \begin{cases} {\partial u_1\over\partial t}= -k_d u_1+ \mu(u_1, u_2)u_1\\ {\partial u_2\over\partial t}= D{\partial^2u_2\over\partial x^2}- c(t){\partial u_2\over\partial x}- k_1\mu(u_1, u_2)u_1,\end{cases}\tag{1} \] with the boundary conditions \[ \begin{cases} {\partial u_2\over\partial x} (0, t)= {c(t)\over D} (u_2(0, t)- s(t)),\\ {\partial u_2\over\partial x} (1,t)= 0\end{cases}\quad (t\in \mathbb{R}_+)\tag{2} \] and the initial condition \[ \begin{cases} u_1(x, 0)= u_{10}(x),\\ u_2(x, 0)= u_{20}(x).\end{cases}\tag{3} \] The states \(u_1= u_1(x,t)\) and \(u_2= u_2(x,t)\) are the concentrations of the biomass of the microorganisms and the substrate, respectively, the specific growth rate of the microorganisms is \(\mu(u_1, u_2)= \mu_m u_2/(k_2 u_1+ u_2)\), the input flow \(c= c(t)\) is the control variable, the input substrate concentration \(s= s(t)\) is a disturbance variable and the output function \(y(t)= u_2(1,t)\) denotes the substrate concentration at the end of the reactor. The problem (1)--(3) is put into abstract form: \[ \dot v+ A(t)v= F(t,v),\quad v(0)= v_0, \] where \(A(t)\) is a linear unbounded operator having the form \[ A(t)= \begin{pmatrix} A_{11} & A_{12}\\ A_{21} & A_{22}(t)\end{pmatrix} \] and \(F(t,v)\) is a nonlinear function and quadratically bounded by \(v\). For this problem the authors investigate the exponential stability around the steady state \(\overline u=(\overline u_1,\overline u_2)\) which is solution to the steady state equations: \[ 0= -k_d\overline u_1+ \mu(\overline u_1,\overline u_2)\overline u_1, \] \[ 0= D{\partial^2\overline u_2\over\partial x^2}-\overline c {\partial\overline u_2\over\partial x}- k_1\mu(\overline u_1, \overline u_2)\overline u_1. \] The main result states that there exists a constant \(\delta>0\) such that when \([c]_{C^{\rho-}(\overline\mathbb{R}_+)}< \delta\), \(0<\rho\leq 1\), then the operator \(A(t)\) has an evolution operator \(U(t,\tau): C(\overline G)\times L_p(G)\to C(\overline G)\times L_p(G)\), \(G= (0,1)\), and there exist \(\alpha\in({1\over 2},1]\) and \(\beta> 0\) such that for \(w\in C(\overline G)\times W^{\alpha,p}(G)\) and for \(0< \tau< t\), \[ \|U(t,\tau)w\|_{C(\overline G)\times W^{\alpha, p}(G)}\leq Me^{-\beta(t-\tau)}\|w\|_{C(\overline G)\times W^{\alpha, p}(G)}, \] \[ \|U(t,\tau) w\|_{C(\overline G)\times W^{\alpha, p}(G)}\leq M(t- \tau)^{-(\alpha- {1\over 2})} e^{-\beta(t- \tau)}\|w\|_{C(\overline G)\times L_p(G)}. \] The authors show also a result about the long-time behaviour for the solutions to problem (1)--(3), namely: if \(\max\{{1\over p},{1\over 2}\}< \alpha\leq 1\), then there exists a constant \(\delta> 0\) such that \[ \|e^{\beta_1(\cdot)}(c- \overline c)\|_\infty< \delta,\quad\|e^{\beta_1(\cdot)}\dot c\|_\infty< \delta, \] \[ \|e^{\beta_1(\cdot)} (s-\overline s)\|_\infty< \delta,\quad\|e^{\beta_1(\cdot)}\dot s\|_\infty< \delta \] and if \(\|u_0-\overline u\|_{C(\overline G)\times W^{\alpha, p}(G)}< \delta\), then: \[ \lim_{t\to\infty} \|u(\cdot,t)-\overline u\|_{C(\overline G)\times W^{\alpha, p}(G)}= 0, \] \(\overline c\) and \(\overline s\) being independent of time in the steady state.