Neural Optimal Magnetohydrodynamic Control of Hypersonic Flows

MAGNETOHYDRODYNAMICS (MHD) as a flow control mechanism is being actively investigated for application in hypersonic systems. Several proposals suggest the use of MHD as an integral part of the flight systems [1-4]. These range from using MHD for the modification of the external flow field ahead of the flight vehicle to internal flow control for aircraft engine application. Our previous work developed an open-loop optimal control approach for an MHD channel at the inlet of an air-breathing hypersonic propulsion system [5]. That approach maximizes the performance of this MHD channel by specifying a profile for the electron beam current along the channel by assuming the knowledge of the flow variables at the lip of the channel. The results show that the flow profile can be successfully controlled for maximizing the net energy extracted from the system. The flow Mach number at the channel exit can be closely held to a prespecified value for a range of inlet conditions. This can make the scramjet combustor design much easier, as it does not have to account for a range of flow Mach numbers. In this work, we assume that the knowledge of system states is possible at a few positions in the channel through the use of sensors at these locations. We then propose a mixed predictive control and dynamic-programming-based control design that provides the optimal e-beam current settings along the particular portion of the channel based on the state information from the adjacent sensor.