Application Of Sliding Mode Control And Model Predictive Control To Limit Management For Aero-engines

In the aero-engine control,limit management is critical to prevent engine into abnormal conditions,such as overspeed,overtemperature,stall/surge,etc.Recently,research results have demonstrated the conservativeness of the traditional linear regulators based Min-Max structure in dealing with limits of aero-engines.Further,other control variables except fuel flow are not fully utilized in the closed-loop control.Therefore,advanced control algorithms such as sliding mode control and model predictive control are proposed due to their own merits.Traditional limit management methods are thus improved and the proposed strategies are extended to multi-input case,resulting in enhanced control performance of aero-engines.The main contents are as follows:1.In order to remove the conservativeness of the traditional linear regulators based Min-Max structure,nonlinear sliding mode control is proposed to substitute all linear regulators,including the main regulator and limit regulators,due to the unilateral convergence characteristics of sliding mode control.Based on this improved strategy,a multivariable main sliding mode regulator,a predictor,new selection logic and smooth switching logic are designed and connected in a hybrid way,to construct the multi-input controller for handling output limits.Further,the regulators that are activated at steady-state and limit protection invariance are analyzed theoretically.The proposed approaches are tested with a high-fidelity engine model known as C-MAPSS40 k.The results show that both multi-input and single-input sliding mode controllers are able to make full use of the limits to accelerate dynamic response.Especially,the multivariate strategy is superior to the single-input method.On one hand,speed tracking can be achieved with tight limits,and on the other hand,faster response can be obtained.2.Based on the idea that model predictive control can realize power management and limit management simultaneously,linear model predictive control algorithm(LMPC)is applied to the aero-engine,thus reducing the complexity of the auxiliary structure(such as Min-Max)used for limit management.According to the requirements of the aero-engines control,baseline LMPC algorithms are improved.Single-input and multi-input LMPC controller are then designed respectively.The influence of key parameters and the limits of different kinds are analyzed.Moreover,after analyzing the robustness of LMPC controller,an adaptive LMPC controller using a multi-parameter scheduling scheme is proposed.Nonlinear simulation results show that LMPC controllers can track speed setpoints fast and limited-outputs work close to their limits during the transient state.3.Nonlinear model predictive controller(NLMPC)is designed to regulate the acceleration/deceleration transient state with input and output limits.Piecewise linear model is utilized as the predictive model,and penalty function is proposed to be incorporated in the sequential quadratic programming(SQP)algorithm for limit management.Meanwhile,penalty function can be regarded as a terminal inequality constraint item,thus ensuring the stability of NLMPC controller.Further,key parameters tuning,as well as the linearization method,are investigated to reduce NLMPC computation.The results indicate that the penalty function can manage the output limits effectively without increasing computation burden;and the NLMPC controller has simpler structure,better dynamic effects,but a larger computation amount,compared with the adaptive LMPC controller.4.Based on the LMPC controller of aero-engine,an active fault-tolerant control scheme is proposed that can judge faults in a hidden way and adjust the control law accordingly.The predictive model library is established first,including engine linear models of normal modes and linear models of known faults modes.At each sampling time,current engine state is compared with the dynamic models in the library,and the best matching model is selected as the predictive model.Moreover,smooth switching logic is designed between sub-controllers.Finally,fault-tolerant control simulations of two kinds of forethought component-level faults,as well as two unknown faults,are carried out.The results show the effectiveness of the proposed strategy.