Optimal control of a hyperbolic model of flow recirculation in a wind tunnel with dynamically constrained boundary controls

This thesis develops a new optimal control theory for a class of distributed-parameter systems governed by first-order quasilinear hyperbolic partial differential equations that arise in many physical applications such as fluid dynamics problems. These systems are controlled at their boundaries via boundary controls that are subject to dynamic constraints imposed by lumped-parameter systems governed by ordinary differential equations. A Mach number control problem for a closed-circuit wind tunnel is investigated. The flow is modeled using the Euler equations and is controlled by a compressor performance model defined as two-point boundary conditions. The boundary control inputs to the compressor are in turn controlled by two first-order lumped-parameter systems that represent dynamics of a drive motor system and an inlet guide vane system. Necessary conditions of optimality are developed by the minimum principle using the adjoint formulation of calculus of variations for a dual Hamiltonian system for the distributed and lumped-parameter systems. The theory is applied to analyze two problems of Mach number control in a wind tunnel: a nonlinear Mach number transition and a linear perturbation predictive feedforward optimal control in the presence of disturbance. Computational methods for general two-point boundary value problems involving coupled partial and ordinary differential equations are developed using a wave-splitting, finite difference upwind method with an explicit scheme for the state equations, and an implicit scheme and a quasi-steady state method for the adjoint equations. These computational methods are implemented to solve a two-point boundary value problem. Using a second-order gradient method, the optimal Mach number transition is computed. A linear-quadratic optimal control theory is developed for designing a Mach number control in the presence of a test model undergoing a continuous pitch motion. A feedback control is shown to not be able to control the Mach number to within a desired accuracy. Therefore, a predictive Mach number feedforward optimal control scheme is proposed and shown to be highly effective in maintaining the Mach number to within a specified tolerance band over a wide range of subsonic Mach number.