Linear quadratic servo control of a reusable rocket engine

Abstract

LINEAR QUADRATIC SERVO CONTROL OF A REUSABLE ROCKET ENGINEJeffrey L. MusgraveNational Aeronautics a.d Space AdministrationI,ewis Research CenterCleveland, Ohio 44135AbstractThis work develops a new design method for theServo Compensator in the frequency domain usingsingular values and applies the method to a reusablerocket engine. An Intelligent Control System forreusable rocket engines has been proposed whichincludes a diagnostic system, a control system and anintelligent coordinator which determines engine controlstrategies based on the identified failure modes. Themethod provides a means of generating various linearmultivariable controllers capable of meetingperformance and robustness specifications andaccommodating failure modes identified by thediagnostic system. Command following with set pointcontrol is necessary for engine operation. A Kalmanfilter reconstructs the state while Loop transfer recoveryrecovers the required degree of robustness whilemaintaining satisfactory rejection of sensor noise fromthe command error. The approach is applied to thedesign of a controller for a rocket engine satisfyingperformance constraints in the frequency domain.Simulation results demonstrate the performance of thelinear design on a nonlinear engine model over allpower levels during mainstage operation.IntroductionAn Intelligent Control System (ICS) for reusablerocket engines has been proposed t for the purpose ofwidening the range of operation, enhancing overallengine performance and reducing the amount of requiredmaintenance. Improvements in engine durabilitythrough control of critical temperatures and pressuresshould improve the usable engine life without sacrificingthe enormous thrust levels required for mission success.The ICS is composed of a real-time diagnostic system,a multivariable cont_oller and an intelligentcoordinator. The diagnostic system identifies enginefailure modes including actuator and sensor failures on-line and supplies this information to the coordinator.The coordinator selects the best possible control strategybased on the severity of the failure and provides thisdata to the reconfigurable controller. The controllerimplements the strategy based upon a priori designcriteria in order to balance the inherent tradeoff betweenengine performance and engine life. The servo-mechanism approach 2 is useful for solving such aproblem. Soft failures such as a drop in turbineefficiency may be accommodated by changing the setpoints of controlled engine variables. Hard failures suchas sensors out of range or sticking actuators (valves)will require reconfiguration of the controller based onthe particular component and its role in the closed loopsystem. Degradations in other engine components maybe accommodated by adding and]or removing variablesfrom the set of controlled quantifies and determiningnew set points. Command following is required becausetransitions between set points must not result in stresscycles which adversely affect the durability of theengine. This work develops a new design method in thefrequency domain for the Servo Compensator* whichmeets the requirements for an ICS controller.Using the linear quadratic regulator for the design ofa command following controller was first introduced byAthans 3. Davison and Goldenberg 4 use stateaugmentation in a similar manner to synthesize therobust Servo Compensator possessing certain degrees ofrobustness to variations in the plant model based on theavailability of outputs for measurement. The methodrelies primarily on augmenting the state vector of theplant with an internal model of the plant disturbancesand reference commands resulting in a multivariablecompensator with a stabilizing loop and a feod-forwardloop. Davison _ proposed a technique for constructingthe gains for the Servo Compensator for the case wherethe plant model is unknown. Wang and Munro _extendearlier results by demonstrating how the ServoCompensator can be used with linear quadratic regulatortheory to synthesize controller gains for both step andramp disturbances and input commands.Much emphasis has been placed on guaranteedrobustness (ability for a control system to maintainstability given uncertainty in the design