Research On Multi-model Predictive Control Of A Turbofan Engine

Aero-engines have the characteristics of wide operating range,strong nonlinearity,and many variables and constraints,which pose great challenges to the design of their control systems.Multi-model predictive control combines multi-model with predictive control,which not only has the advantages of predictive control in dealing with constrained multivariable control problems,but also can accurately approximate the nonlinear characteristics of the whole working condition by using multi-model,so it can be well applied to aero-engine control problems.In this paper,the research of multi-model predictive control for a turbofan engine is carried out,including engine operating range division,multi-model predictive controller design,fault-tolerant control and hardware-in-the-loop simulation system.The main work of this paper is as follows:(1)Aiming at the difficulty of characterizing nonlinear systems with linear model sets in a wide range of aero-engine operating conditions,a model set selection algorithm based on gap metric is used.The gap metric can be used to characterize the dynamic difference between the two linear systems,and the effective neighborhood range of the linear model at a certain equilibrium point can be estimated to approximate the original nonlinear system,and then the entire operating range of the engine is divided.Finally,a multi-model set consisting of a set of linear models is obtained,which can more accurately characterize the nonlinear characteristics of the aero-engine in the entire operating range.(2)Aiming at the problem that the control variable output fluctuates greatly when the aeroengine controller switches between different operating conditions,a multi-model predictive control method based on controller weighting is proposed.Firstly,the integral of error is introduced to augment the prediction model to eliminate the steady-state error of the system.Secondly,based on the multi-parameter quadratic programming method,the online calculation problem of model predictive control is converted into an offline solution.Then,a multi-model predictive controller based on convex combination weighting is designed.The simulation results show that the output response of this control method changes more gently when the region is switched compared with the direct switching method.Finally,a multi-model predictive controller based on gap metric weighting is designed.The simulation results show that the control method has faster output response time and less oscillation than the method based on convex combination weighting,which verifies the effectiveness of the designed controller.(3)Aiming at the problem of hard faults of aero-engine sensors and soft faults of actuators,a multi-model predictive fault-tolerant control method is proposed,a fault-tolerant predictive model library is established,and predictive controllers are designed for each fault type.Then,aiming at the hard fault of the complete failure of the high-voltage rotor speed sensor,a multi-model predictive fault-tolerant control method based on convex combination weighting is proposed,and a differential-integral link is introduced.Moreover,there is no overshoot and steady-state error,and it has a good fault-tolerant control effect.Finally,aiming at the soft fault of the actuator,a multi-model predictive fault-tolerant controller based on gap metric weighting is designed.The simulation results show that this method can ensure that the output response has no oscillation and overshoot,and has a good fault-tolerant control effect,which verifies the effectiveness of the designed controller.(4)The effectiveness of the designed control method is verified by building a hardwarein-the-loop experimental platform.On the basis of analyzing the system design requirements,the design process of the hardware-in-the-loop system controller module,engine model module,communication module and host computer monitoring software is introduced.Finally,the hardware-in-the-loop simulation experiment of the multi-model predictive controller designed in this paper is carried out,which verifies the real-time performance and engineering feasibility of the control method designed in this paper.