CN118188176A - Predictive control method and device for mode switching process of multi-mode turbine engine - Google Patents

Abstract

The invention discloses a predictive control method for a mode switching process of a multi-mode turbine engine. The prediction control method divides the input variable geometry directional regulator into two stages of flameout of a main combustion chamber and directional regulation of a variable geometry regulating mechanism, and uses a feedback correction loop formed by an airborne self-adaptive composite model and a multi-mode prediction controller to perform prediction control on the multi-mode turbine engine in the mode switching process. The invention also discloses a predictive control device for the mode switching process of the multi-mode turbine engine. Compared with the prior art, the method can realize the accurate control of the thrust and the flow in the mode switching process of the multi-mode turbine engine under the condition of the degradation of the engine performance.

Description

Predictive control method and device for mode switching process of multi-mode turbine engine

Technical Field

The invention belongs to the technical field of aviation aerospace propulsion theory and engineering, and particularly relates to a predictive control method and device for a mode switching process of a multi-mode turbine engine.

Background

The development of the wide airspace and wide speed range reusable aircraft pulls the technical innovation of the air suction type power, in order to improve the range and economy of the aircraft, the air suction type power device needs to simultaneously consider the economy under low Mach number and the high thrust performance under high Mach number, so that a new generation of aeroengines represented by XA-100 self-adaptive cycle engines and hypersonic speed power represented by turbine punching combined engines are generated, the two engines work in a turbofan mode at low flight speed, the oil consumption of the engine is reduced, the economy is improved, and the thrust performance of the engine under high Mach number is improved as much as possible when the engines work in a turbojet/punching mode at high flight speed. The multi-working mode is an important characteristic of the new generation of reusable aerospace power, and the mode switching process control is an important difficulty that the design of the engine control system must pay important attention to.

The mode switching process is often accompanied by flow passage adjustment of variable geometry mechanisms such as a mode selection valve, a bypass ejector and the like, transient switching of the bypass function can cause abrupt change of engine thermodynamic cycle, the abrupt change is reflected to the great fluctuation of engine thrust and required flow at the engine performance output end, and because the mode switching process is often under the condition of high flying speed, the transient thrust and flow abrupt change is extremely easy to cause instability of an aircraft, meanwhile, the mode switching failure and the like can be possibly caused, so that the mode switching process control method is a research hot spot of the current variable cycle engine and turbine-based combined power.

For the research of a mode switching process control method of a variable cycle engine, hao et al propose a mode switching process adjustable parameter optimization method aiming at direct thrust control, and obtain a corresponding mode switching control rule [ A NEW DESIGN Method for Mode Transition Control Law of Variable CYCLE ENGINE [ R ]. Zhang et al propose a general design method [ GENERAL DESIGN method of control law for ADAPTIVE CYCLE ENGINE mode transition [ J ] ] of a mode switching process based on a particle swarm optimization method and a sensitivity calculation method, and Zheng et al propose an adjustable variable adjustment strategy such as an intermediate state staged mode switching control law design method and a mode switching rotating speed so as to realize continuous engine thrust in the mode switching process [ adaptive cycle engine mode switching transition state control law design method research [ J ] ]. Chen et al propose a constant flow mode switching method, by controlling the fan static pressure [ Flow control of double bypass variable CYCLE ENGINE IN modal transition [ J ] ] during mode switching, the engine flow is ensured to be continuous during mode switching. Wang et al propose a mode switching process control scheduling design method [ Game-distance-Based Mode Switch Control Schedule Design for Variable CYCLE ENGINE [ J ] ] based on Game optimization idea in order to realize fast mode switching. Yu et al build linear switching models under different modes, design and plan corresponding smooth switching control strategies based on fuzzy control and intelligent algorithms [Active disturbance rejection control for uncertain nonlinear systems subject to magnitude and rate saturation:Application to aeroengine[J]].

For the research of the combined power mode switching process control method, lv et al have conducted a great deal of research work. The TBCC structure proposal [Mode transition analysis of a turbine-based combined-cycle considering ammonia injection pre-compressor cooling and variable-geometry ram-combustor[J]], based on liquid ammonia jet precooling [Thermodynamic modeling and analysis of ammonia injection pre-compressor cooling cycle:A novel scheme for high Mach number turbine engines[J]] and a variable geometry stamping combustion chamber is provided to improve the thrust performance of a turbofan mode engine in the mode switching process, a steady-state switching path [Mode transition path optimization for turbine-based combined-cycle ramjet stage under uncertainty propagation of integrated airframe-propulsion system[J]], which meets the condition that the thrust is unchanged and the oil consumption is lower in the mode switching process is obtained through a multi-objective optimization method, the accurate control [Intelligent ammonia precooling control for TBCC mode transition based on neural network improved equilibrium manifold expansion model[J]]. of the liquid ammonia jet precooling temperature in the mode switching process is realized by introducing an improved neural network balance manifold model and an extended state observer, in order to ensure the safe and stable operation of an air inlet channel and a turbine engine in the mode switching process, a mode switching process control law design method [Mode transition control law analysis of ammonia MIPCC aeroengine considering inlet–compressor safety matching[J]].Xi and the like which consider the non-starting margin of the air inlet channel and the outlet temperature of the combustion chamber are provided, the mode switching process thrust increasing method [Design of thrust augmentation control scheduleduring mode transition for turbo-ramjet engine[J]],Zheng based on the improved control plan is provided, and the like are provided to change the thrust resistance characteristic of an aircraft by optimizing the flight track of the aircraft, and the problem of discontinuous thrust in the mode switching process is solved [ Trajectory optimization for a TBCC-powered supersonic VEHICLE WITH transition thrust pinch [ J ].

The research results provide rich theoretical support for the multi-mode air suction type power mode switching process control, but the multi-mode air suction type power application scene is a reusable aircraft, and the problem of component performance degradation is inevitably caused in the long-term service process. It is well known that suction power is a strong nonlinear system, and engine component performance degradation can result in a change in the mapping of engine tuning variables to thrust, and the mode switching process further exacerbates this nonlinear mapping uncertainty due to changes in thermodynamic cycle characteristics. For the control law of the mode switching process designed based on the optimization method, the control target that the expected engine thrust and flow are continuous cannot be ensured by performing the mode switching under the performance degradation condition, the mode switching control method has the defect of poor applicability, and the optimization method has poor instantaneity, and cannot finish the optimization of the adjustment variable instruction within the control step length at the current hardware level of the aeroengine controller, so that the optimization method cannot be applied to the real-time optimization control of the mode switching process, and the mode switching can only be realized in an off-line control law mode. The control method based on the airborne self-adaptive model can directly take the non-testability performance parameters such as engine thrust, flow and the like as control targets, automatically compensate the influence of engine performance degradation on a control instruction, and has better application prospect in the mode switching process control. However, the mode switching process is often not performed at only one working point, a certain switching envelope exists, and the complexity and difficulty of the airborne modeling of the mode switching process are aggravated by the characteristics of multiple working modes and multiple adjusting variables, so that the application of the control method based on the airborne model in the control field of the mode switching process is restricted.

Disclosure of Invention

The invention aims to solve the technical problems of overcoming the defects of the prior art and providing a predictive control method for a mode switching process of a multi-mode turbine engine, which can realize the accurate control of thrust and flow in the mode switching process of the multi-mode turbine engine under the condition of engine performance degradation.

The technical scheme provided by the invention is as follows:

the predictive control method of the mode switching process of the multi-mode turbine engine divides the input variable geometry directional regulator into two stages of flameout of the main combustion chamber and directional regulation of the variable geometry regulating mechanism; in the mode switching process, a feedback correction loop formed by an airborne self-adaptive composite model and a multi-mode predictive controller is used for predictive control of the multi-mode turbine engine; the airborne self-adaptive composite model is used for estimating the thrust and the flow of the engine in real time; the multi-mode predictive controller comprises a turbofan mode predictive controller and a turbojet mode predictive controller which are respectively used for carrying out predictive control on the multi-mode turbine engine in a main combustion chamber flameout stage and a variable geometry adjusting mechanism directional adjusting stage, the turbofan mode predictive controller and the turbojet mode predictive controller both take the airborne self-adaptive composite model as a predictive model, and take the minimum engine thrust and flow change amplitude as performance targets to carry out online rolling optimization so as to generate engine control parameters in the mode switching process, thereby realizing the smooth transition of the engine thrust and flow in the mode switching process.

Preferably, the airborne self-adaptive composite model comprises a steady-state baseline model, a health parameter estimation module and a propulsion system matrix module, wherein the steady-state baseline model, the health parameter estimation module and the propulsion system matrix module respectively comprise a turbofan mode taking the fuel flow of a main combustion chamber as a scheduling variable and a turbojet mode taking a mode selection valve angle as a scheduling variable; the steady-state baseline model takes the flying height, mach number and engine adjustable variables under different working modes as input, takes the engine steady-state performance parameters as output, is obtained through training by a deep neural network method and is used for estimating the steady-state performance of the engine in real time; the health parameter estimation module is used for carrying out online real-time estimation on the performance degradation condition of the engine; the propulsion system matrix module is stored with a propulsion system matrix which is built offline in advance and corresponds to working points of different baseline models and reflects the influence condition of component performance degradation parameters on engine performance parameters, and the propulsion system matrix module is used for compensating and correcting the engine steady-state performance parameters output by the steady-state baseline model according to the engine performance degradation condition estimated by the health parameter estimation module.

Further preferably, the health parameter estimation module is implemented by off-line designing a Kalman filter by augmenting an engine component performance degradation parameter into an engine state space model state variable.

Preferably, the variation range of the engine control parameters for building the onboard adaptive composite model is obtained by offline optimization solving the following optimization objective functions in advance:

Wherein F _,obj、ma_2,obj represents the engine thrust and flow before the mode switching, F [ k ] and ma ₂ [ k ] are the engine thrust and flow at the kth step in the mode switching process, u [ k ] is the control amount at the kth step, and ω ₁、ω₂ is the weight coefficient.

Preferably, the online rolling optimization is achieved by optimizing the solution of the following objective function:

Wherein r is an engine control instruction and is a fixed value in the mode switching process; n _p、N_u is the prediction time domain and the control time domain respectively; Estimating values of instruction variables for the prediction model; Δu is the difference between the front and rear timings of the engine control amount, i.e., Δu (k+i) =u (k+i) -u (k+i-1); q and R are semi-positive definite matrices.

Based on the same invention idea, the following technical scheme can be obtained:

The predictive control device of the mode switching process of the multi-mode turbine engine comprises a feedback correction loop formed by an airborne self-adaptive composite model and a multi-mode predictive controller, and is used for performing predictive control on the multi-mode turbine engine in the mode switching process; the variable geometry directional adjustment plan input into the predictive control device is divided into two stages of flameout of the main combustion chamber and directional adjustment of the variable geometry adjustment mechanism; the airborne self-adaptive composite model is used for estimating the thrust and the flow of the engine in real time; the multi-mode predictive controller comprises a turbofan mode predictive controller and a turbojet mode predictive controller which are respectively used for carrying out predictive control on the multi-mode turbine engine in a main combustion chamber flameout stage and a variable geometry adjusting mechanism directional adjusting stage, the turbofan mode predictive controller and the turbojet mode predictive controller both take the airborne self-adaptive composite model as a predictive model, and take the minimum engine thrust and flow change amplitude as performance targets to carry out online rolling optimization so as to generate engine control parameters in the mode switching process, thereby realizing the smooth transition of the engine thrust and flow in the mode switching process.

Preferably, the airborne self-adaptive composite model comprises a steady-state baseline model, a health parameter estimation module and a propulsion system matrix module, wherein the steady-state baseline model, the health parameter estimation module and the propulsion system matrix module respectively comprise a turbofan mode taking the fuel flow of a main combustion chamber as a scheduling variable and a turbojet mode taking a mode selection valve angle as a scheduling variable; the steady-state baseline model takes the flying height, mach number and engine adjustable variables under different working modes as input, takes the engine steady-state performance parameters as output, is obtained through training by a deep neural network method and is used for estimating the steady-state performance of the engine in real time; the health parameter estimation module is used for carrying out online real-time estimation on the performance degradation condition of the engine; the propulsion system matrix module is stored with a propulsion system matrix which is built offline in advance and corresponds to working points of different baseline models and reflects the influence condition of component performance degradation parameters on engine performance parameters, and the propulsion system matrix module is used for compensating and correcting the engine steady-state performance parameters output by the steady-state baseline model according to the engine performance degradation condition estimated by the health parameter estimation module.

Further preferably, the health parameter estimation module is implemented by off-line designing a Kalman filter by augmenting an engine component performance degradation parameter into an engine state space model state variable.

Preferably, the variation range of the engine control parameters for building the onboard adaptive composite model is obtained by offline optimization solving the following optimization objective functions in advance:

Wherein F _,obj、ma_2,obj represents the engine thrust and flow before the mode switching, F [ k ] and ma ₂ [ k ] are the engine thrust and flow at the kth step in the mode switching process, u [ k ] is the control amount at the kth step, and ω ₁、ω₂ is the weight coefficient.

Preferably, the online rolling optimization is achieved by optimizing the solution of the following objective function:

Wherein r is an engine control instruction and is a fixed value in the mode switching process; n _p、N_u is the prediction time domain and the control time domain respectively; Estimating values of instruction variables for the prediction model; Δu is the difference between the front and rear timings of the engine control amount, i.e., Δu (k+i) =u (k+i) -u (k+i-1); q and R are semi-positive definite matrices.

Compared with the prior art, the technical scheme of the invention has the following beneficial effects:

(1) The invention still maintains accurate control of the thrust and flow of the engine under the condition of degradation of the engine performance, the maximum fluctuation of the thrust is less than 0.66%, and wider applicability is shown;

(2) The method has the advantages that the control effect is equivalent to that of the traditional SQP optimization method, the single-step calculation average time consumption is only 2.8% of that of the SQP optimization method, the thrust force, the flow control precision and the calculation instantaneity of the mode switching process are considered, and the method has better engineering application prospects in the design field of the mode switching control system of the multi-mode turbine engine.

Drawings

FIG. 1 is a schematic diagram of a design flow of a mode switching process prediction control method according to the present invention;

FIG. 2 is a multi-mode airborne adaptive composite model structure;

FIG. 3 is a multi-mode turbine engine mode switch predictive control architecture;

FIG. 4 (a) is a graph of baseline model engine thrust test errors (turbofan mode);

FIG. 4 (b) is a baseline model engine flow test error map (turbofan mode);

FIG. 5 (a) is a comparison of engine thrust accuracy for an on-board model with degraded component performance (turbojet mode);

FIG. 5 (b) is a comparison of flow accuracy of an on-board model engine with degraded component performance (turbojet mode);

FIG. 6 (a) is an adaptive estimation module estimation accuracy for a single component performance degradation case;

FIG. 6 (b) is an adaptive estimation module estimation accuracy for a multi-component performance degradation case;

FIG. 7 (a) shows the engine thrust variation during mode switching;

FIG. 7 (b) is a change in engine flow during a mode switch;

FIG. 8 is a single step time-consuming comparison of different control methods for the mode switching process;

FIG. 9 (a) shows engine thrust variation during mode switching with degraded component performance;

FIG. 9 (b) shows engine flow variation during mode switching with component performance degradation

Detailed Description

Aiming at the defects of the prior art, the invention divides the input variable geometry directional regulator into two stages of flameout of a main combustion chamber and directional regulation of a variable geometry regulating mechanism, and a feedback correction loop formed by an onboard self-adaptive composite model and a multi-mode predictive controller is used for predictive control of the multi-mode turbine engine in the mode switching process so as to realize accurate control of thrust and flow in the mode switching process of the multi-mode turbine engine under the condition of degradation of engine performance.

For the convenience of public understanding, the following describes the technical scheme of the present invention in detail by taking a certain triaxial, multi-combustor configuration with higher nonlinearity as an example, and using a turbine engine with a turbojet/turbofan multi-working mode as an example, and referring to the accompanying drawings:

The design flow of the mode switching process prediction control method in this embodiment is shown in fig. 1, firstly, determining, by an SQP optimization method, that the mode switching process meets the variation range of engine control parameters (main combustion chamber fuel flow, inter-stage combustion chamber fuel flow, afterburner fuel flow, nozzle throat area, etc.) with constant engine thrust and flow, and reducing the input parameter modeling interval of an airborne model in the mode switching process; then developing a modeling method of a multi-mode turbine engine airborne self-adaptive composite model oriented to a mode switching process, outputting performance parameters such as engine thrust, flow, compression component stability margin and the like on line by establishing the multi-mode turbine engine airborne self-adaptive model, estimating the performance degradation condition of the engine in real time by utilizing a Kalman filter, and compensating and correcting the output of the airborne model by a propulsion system matrix module; and finally, the airborne self-adaptive composite model is used as a prediction model to design a mode switching process prediction controller so as to realize smooth transition of the thrust and flow of the engine in the mode switching process.

In order to obtain the engine control parameter variation range which meets the condition that the engine thrust and the flow are unchanged and is used for establishing an airborne self-adaptive composite model, the invention converts the mode switching process control parameter variation problem into a single-target multi-constraint dynamic programming problem, and the invention determines the optimization objective function as shown in the following formula:

Wherein F _,obj、ma_2,obj represents the engine thrust and flow before the mode switching, F [ k ] and ma ₂ [ k ] are the engine thrust and flow at the kth step in the mode switching process, u [ k ] is the control amount at the kth step, and ω ₁、ω₂ is the weight coefficient. By this objective function, the engine thrust and flow rate change can be minimized in each switching step.

And (3) carrying out off-line optimization solving on the problem by adopting an SQP method, thereby obtaining the parameter variation ranges of the regulating variables such as the main combustion chamber, the interstage combustion chamber, the afterburner fuel flow, the nozzle throat area and the like which meet the conditions of constant thrust and flow of the engine in the mode switching process.

After the parameter variation range is determined, an onboard self-adaptive composite model of the multi-mode turbine engine is further established. The invention provides a multi-mode airborne self-adaptive composite model shown in figure 2, which mainly comprises three modules, namely a steady-state baseline model, a propulsion system matrix module and a health parameter estimation module, wherein each module comprises a turbofan mode and a turbojet mode.

The steady-state baseline model is obtained by taking the fly height, mach number and engine adjustable variables under different working modes as input, taking the engine performance parameters as output and training and establishing by a deep neural network method, and is used for estimating the steady-state performance of the engine in real time. The health parameter estimation module establishes state space models in different modes offline, amplifies engine degradation parameters in different modes and carries out Kalman filter design, so that online real-time estimation of engine performance degradation conditions is realized, and an airborne model self-adaption function is realized. And estimating the performance degradation condition of the engine by a health parameter estimation module, transmitting the engine performance degradation condition to a propulsion system matrix, and correcting the influence of the performance degradation of the engine part on the steady-state baseline model output of the airborne model by the propulsion system matrix to realize the high-precision tracking output of the performance parameter of the airborne model under the condition of the engine performance degradation. The modeling process of each part will be specifically described below.

For the switching problem of the multi-mode airborne composite model, the invention respectively establishes a turbofan mode and a turbojet mode airborne composite model in a mode switching interval, and in the flameout stage of a main combustion chamber in the mode switching process, the invention selects the turbofan mode model by taking the fuel flow of the main combustion chamber as a dispatching variable, when the main combustion chamber works normally, the airborne composite model is switched to the turbojet mode airborne composite model when the fuel flow of the main combustion chamber is smaller than the flameout boundary of the main combustion chamber, and the output value of the engine performance under different inputs is calculated by taking the angle of a mode selection valve as the dispatching variable.

According to the baseline model modeling method based on the BP neural network, according to different mode working characteristics, a turbofan mode baseline model and a turbojet mode baseline model are respectively established through offline training. The input parameters u= [ H, ma and m _fb,m_fbin,A₈,m_fa]^T of the turbofan mode baseline model are flight altitude, flight Mach number, main combustion chamber fuel flow, inter-stage combustion chamber fuel flow, nozzle throat area and afterburner fuel flow in sequence; the output parameters y＝[F,m_a2,N_f,N_CDFS,N_c,S_mf,S_mCDFS,S_mc,T_t48]^T are, in order, engine thrust, engine flow, three rotor speeds, compression component surge margin, and total interstage combustor outlet temperature. The input parameters u= [ H, ma and m _fbin,A₈,m_fa,α_MSV]^T of the vortex-jet mode baseline model are flight height, flight Mach number, inter-stage combustion chamber fuel flow, nozzle throat area, afterburner fuel flow and mode selection valve angle in sequence; the output parameter y= [ F, m _a2,S_mf,T_t48,N_f]^T is in turn the engine thrust, engine flow, fan surge margin, total temperature of the interstage combustor outlet, low pressure rotor speed.

Because the research object of the embodiment is a three-rotor structure, the rotating parts are more, and the problem of performance degradation parameter selection of the rotating parts is solved, the invention carries out correlation analysis according to documents, and finally determines the degradation of turbofan mode fan, core driving fan, compressor flow, high, medium and low pressure turbine efficiency, and the degradation of turbojet mode fan flow and low pressure turbine efficiency are health estimation parameters. The functional relationship between control quantity and output quantity can be expressed as:

ΔY＝PΔu (2)

Wherein, the turbofan mode ΔY＝[ΔF,Δm_a2,ΔN_f,ΔN_CDFS,ΔN_c,ΔS_mf,ΔS_mCDFS,ΔS_mc,ΔT_t48]^T( turbojet mode deltay= [ deltaf, deltam _a2,ΔN_f,ΔS_mf,ΔT_t48]^T) is an m-dimensional vector, and deltay represents the deviation of the engine output parameter; Δu= [ η _F,η_CDFS,η_HC,η_HT,η_MT,η_LT]^T (vortex spray pattern Δu= [ η _F,η_LT]^T) is an n-dimensional vector, Δu represents an engine performance degradation parameter, and P is an mxn-dimensional propulsion system matrix element.

The performance degradation variable and the output variable selected according to the present embodiment, the propulsion system matrix of the turbofan mode thereof may be expressed as:

the corresponding turbojet mode propulsion system matrix is:

Through offline calculation, component performance degradation parameter propulsion system matrixes corresponding to different modes and different base line model working points are established, a propulsion system matrix database is formed, and the component performance degradation parameter propulsion system matrixes are integrated into an airborne model, so that real-time compensation of performance degradation conditions is realized, and the purpose of self-adaption of the airborne model is achieved.

According to the invention, a Kalman filter method is adopted for estimating the performance degradation condition of the engine under different modes, and the basic thought is to amplify the performance degradation parameters of the engine parts into state variables of an engine state space model, design the Kalman filter offline, establish a health parameter estimation module, integrate the health parameter estimation module into an airborne model, take the deviation of measurable output parameters of the engine as the input of the filter, further estimate the performance degradation condition of each part of the engine online, and realize real-time tracking of the airborne model in the real state of the engine.

The mathematical expression of the engine state space model taking into account the component performance degradation Δη and the ambient noise is as follows:

In the formula, x represents the state quantity of the engine, and the state space model is considered to be also applied to a prediction model in a model prediction control method, so that the control quantity u= [ m _fb,m_fbin,A₈,m_fa]^T ] of the turbofan mode x＝[F,m_a2,N_f,N_CDFS,N_c,S_mf,S_mCDFS,S_mc,T_t48]^T,, and the output quantity consisting of measurable parameters is y= [ N _f,N_CDFS,N_c,T_t6,P_t3,P_t6]^T; the vortex spray pattern x= [ F, m _a2,N_f,S_mf,T_t48]^T, the control quantity u= [ m _fbin,A₈,m_fa]^T), the output quantity composed of measurable parameters is y= [ N _f,T_t6,P_t6]^T, A, B, C, D, L, M is a fitting matrix, the vortex fan pattern Δη= [ Δη _F,Δη_CDFS,Δη_HC,Δη_HT,Δη_MT,Δη_LT]^T (vortex spray pattern Δη= [ Δη _F,Δη_LT]^T), w and v are system noise and measurement noise, respectively. .

Since Δη cannot be directly obtained, the form of a full-dimensional observer based on a Kalman filter can be obtained by taking the performance degradation amount as an amplified state quantity based on the formula (5):

In the middle of B_k＝[B 0]^T,C_k＝[C M]，/>For the estimation, it can be obtained from an on-board model, and the Kalman filter gain matrix K is determined by equation (7).

Wherein P is a solution of Riccati equation, Q, R is covariance matrix of white noise matrix w and v, and I represents identity matrix.

A_kP+PA_k ^T-PC_k ^TR^-1C_kP+Q＝0 (8)

After the self-adaptive airborne composite model is established, a mode switching process thrust and flow smooth transition control research based on a model predictive control method is carried out on the basis. Aiming at the problem that the mode switching process involves the switching of a prediction model, the invention provides the multi-mode prediction controller shown in fig. 3 to realize the mode switching process control.

The multi-mode turbine engine has two working states of a turbofan mode and a turbojet mode, the turbofan mode can not meet the thrust requirement under the working condition of high Mach number, and the turbojet mode needs to be switched to further increase the engine thrust. The engine needs to perform main combustion chamber flameout and variable geometry diverter ring and mode selector valve adjustment to achieve turbofan to turbojet operating mode switching. The invention divides the variable geometry directional adjustment plan in fig. 3 into two sequential stages of main combustion chamber flameout (main combustion chamber fuel flow m _fb linearly drops to 0) and variable geometry adjustment mechanism directional adjustment. The directional adjustment of the variable geometry adjustment mechanism refers to the synchronous adjustment of the mode selection valve (alpha _MSV is adjusted from 0 deg. to 90 deg.) and the variable geometry diverter ring (alpha _splitter is adjusted from 0 deg. to-90 deg.). The multi-mode predictive controller in the figure comprises a turbofan mode predictive controller and a turbojet mode predictive controller. The control targets of both controllers are the same, and the engine thrust and flow do not become the control targets before and after mode switching (i.e., the reference trajectories in the multi-mode predictive controller of fig. 3). The invention proposes to call different predictive controllers according to a variable geometry directional adjustment plan: in the flameout stage of the main combustion chamber, a turbofan mode prediction controller is called, and in the adjusting process of the variable geometry adjusting mechanism, the turbofan mode prediction controller is called. Because the engine thrust and the flow are all non-measurable parameters, the invention establishes a multi-mode airborne self-adaptive composite model, and the engine thrust and the flow are estimated by using the airborne self-adaptive composite model, so that a feedback correction loop is formed in the predictive controller in FIG. 3. Meanwhile, the multi-mode airborne self-adaptive composite model is also a prediction model used by a rolling optimization module in prediction control, rolling optimization is performed by using the minimum engine thrust and flow change amplitude as performance targets through a multi-mode prediction controller, three control parameters of inter-stage combustion chamber fuel flow (m _fbin), afterburner fuel flow (m _fa) and nozzle throat area (A ₈) are calculated, and the three control parameters are input to the engine through an executing mechanism, so that a complete loop of prediction control of smooth transition of the engine thrust and flow in the mode switching process is realized.

As shown in fig. 3, the control structure mainly comprises two core parts of a prediction model and an online rolling optimization module, wherein the prediction model (multi-mode airborne self-adaptive composite model) is described in detail in the foregoing, and the online rolling optimization part is described below.

In the mode switching process, the expected control target is that the thrust and the flow of the engine are kept continuous, and the thrust and the flow are kept unchanged as far as possible in the whole mode switching process, so that the invention constructs the following objective function:

Wherein r is an engine control instruction and is a fixed value in the mode switching process; n _p is the prediction time domain, N _u is the control time domain, Estimating values of instruction variables for the prediction model; Δu is the difference between the front and rear timings of the engine control amount, i.e., Δu (k+i) =u (k+i) -u (k+i-1); the first term of the objective function is used for quickly tracking the command value, and the second term of the objective function is used for ensuring that the control quantity is as stable as possible after the engine tracks the command; q and R are semi-positive definite matrices.

After determining the objective function, converting the online prediction optimization problem of the mode switching process into a quadratic programming problem, and solving the optimization problem by adopting an effective set method.

In order to verify the effect of the predictive control device, relevant simulation and analysis are carried out.

Firstly, the accuracy of the built airborne self-adaptive composite model is verified in a simulation mode. According to the optimization of the mode switching process, for the turbofan mode, the modeling interval of the invention is H=20-24 km, the Ma=3.75-4.2, the variation range of the main combustion chamber fuel flow m _fb is 0.4-0 kg/s, the variation range of the inter-stage combustion chamber fuel flow m _fbin is 1.2-1.8 kg/s, the variation range of the afterburner fuel flow m _fa is 1.6-2.4 kg/s, and the variation range of the nozzle throat area A ₈ is 0.55-0.61 m ². For the turbojet mode, the variable range of the variable geometry splitter alpha _MSV is 0-90 degrees, and the variable range of the inter-stage combustion chamber fuel flow, the afterburner fuel flow and the nozzle throat area is the same as that of the turbofan mode. As can be seen from fig. 4 (a) and 4 (b), the baseline modeling accuracy of the engine thrust and flow rate is high, the maximum error is less than 1%, and the average error is less than 0.035%.

In order to further verify the accuracy of the built airborne self-adaptive composite model, engine fan flow degradation and low-pressure turbine efficiency degradation simulation are carried out in a vortex-spray mode, the performance degradation ranges are all 0-3%, and the comparison result of the airborne model output and the engine part level data accuracy is shown in fig. 5 (a) and 5 (b).

In fig. 5 (a) and fig. 5 (b), the left column shows the accurate condition of different parameters under the condition of the performance degradation of the engine component by adopting the existing machine-mounted model based on the neural network, and the right column shows the accurate condition of different parameters under the condition of the performance degradation of the engine component by adopting the machine-mounted self-adaptive composite model provided by the invention. According to the graph, the PSM matrix module is used for correcting the performance degradation condition of the engine, so that the accuracy of an airborne model is effectively improved, and the modeling accuracy of the thrust and the flow of the engine is remarkably improved. Under the condition of performance degradation of engine components in the vortex-spray mode, the calculation precision of each parameter of the airborne self-adaptive composite model is better than 0.337%, and a solid foundation is provided for accurately estimating performance parameters such as engine thrust, flow and the like in model predictive control.

In order to verify the accurate tracking capability of the airborne self-adaptive composite model to the performance degradation of the engine, single-component performance degradation and multi-component performance simultaneous degradation simulation are respectively carried out under the flight working conditions of H=24 km and Ma=3.75, and the results are shown in fig. 6 (a) and 6 (b). Fig. 6 (a) is a single component performance degradation simulation result, where fan component flow degradation is set to 2% at t=0s and low pressure turbine efficiency degradation is set to 2% at t=40s; fig. 6 (b) is a simulation result of multiple component performance simultaneous progressive degradation, where at t=0s, fan flow progressive degradation 1%, compressor flow progressive degradation 2%, intermediate pressure turbine efficiency progressive degradation 1.5%, and at t=40s, core drive fan flow progressive degradation 1.5%, high pressure turbine efficiency progressive degradation 1%, low pressure turbine efficiency progressive degradation 2% are set. As shown by simulation results, the airborne self-adaptive composite model established by the invention has better performance degradation tracking capability, no steady-state estimation error and good self-adaptive capability for single-component performance abrupt degradation and multi-component simultaneous performance gradual degradation.

Fig. 7 (a) and 7 (b) show the engine performance parameter response results when mode switching is performed under the flight condition of h=24 km, ma=3.75. MPC in the legend represents the method proposed by the invention, SQP represents the SQP optimization method, and PID represents the traditional PID control method. As can be seen from the engine thrust and flow response conditions in fig. 7 (a) and 7 (b), the SQP optimization method and the MPC control method effectively ensure that the engine thrust and flow remain unchanged basically in the mode switching process, and the maximum thrust fluctuation amounts of the PID control method, the SQP optimization method and the MPC control method are respectively: 2.8%, 0.03% and 0.73%, and the maximum fluctuation amounts of the engine flow in the three methods are respectively as follows: 3.4%, 0.52% and 0.31%.

FIG. 8 shows the comparison of the single step calculation time consumption for different control methods, with the simulated processor being Intel Core i5-6500U 3.20GHz, memory 8G. The SQP optimizing method has the best thrust control effect, but the calculation time consumption is the largest, the calculation processing capacity of the existing aeroengine controller cannot be optimized in real time within a single control step length of 25ms, the average single-step calculation time consumption of the MPC control method is 0.204ms, the average single-step calculation time consumption is approximately 50 times higher than that of the PID control method, the average single-step calculation time consumption is obviously smaller than that of the SQP optimizing method, the maximum single-step calculation time consumption is also at a lower level, and the method is hopefully applied to actual multi-mode turbine engine engineering control.

In order to further verify the applicability of the mode switching process prediction control method provided by the invention, 2% of fan flow degradation and 2% of low-pressure turbine efficiency degradation are simultaneously set in the mode switching process under the same flight conditions as the simulation, and the mode switching process control effects of different control methods under the condition of engine performance degradation are shown in fig. 9 (a) and 9 (b). As can be seen from fig. 9 (a) and fig. 9 (b), the MPC control method provided by the present invention can still ensure that the engine thrust and the flow rate in the mode switching process remain substantially unchanged under the condition of component performance degradation, the maximum variation ranges are respectively 0.66% and 1.71%, the PID control and SQP optimization method respectively has the maximum variation ranges of 2.0% and 0.28% in the mode switching process under the condition of component performance degradation, and the maximum variation ranges of the engine flow rate are respectively 2.8% and 1.71%. It is worth noting that for both the SQP optimization and MPC control methods, the maximum flow variation occurs at the mode switch start point due to degradation of the setup component performance. For the whole switching process, steady-state errors exist in the PID control and SQP optimization methods, accurate tracking of the control instruction in the mode switching process cannot be realized, and the MPC control method has no steady-state error, so that the adaptive capacity of the mode switching prediction control method provided by the invention for the performance degradation of engine parts is good.

Claims (10)

1. The predictive control method for the mode switching process of the multi-mode turbine engine is characterized in that an input variable geometry directional regulator is divided into two stages of flameout of a main combustion chamber and directional regulation of a variable geometry regulating mechanism; in the mode switching process, a feedback correction loop formed by an airborne self-adaptive composite model and a multi-mode predictive controller is used for predictive control of the multi-mode turbine engine; the airborne self-adaptive composite model is used for estimating the thrust and the flow of the engine in real time; the multi-mode predictive controller comprises a turbofan mode predictive controller and a turbojet mode predictive controller which are respectively used for carrying out predictive control on the multi-mode turbine engine in a main combustion chamber flameout stage and a variable geometry adjusting mechanism directional adjusting stage, the turbofan mode predictive controller and the turbojet mode predictive controller both take the airborne self-adaptive composite model as a predictive model, and take the minimum engine thrust and flow change amplitude as performance targets to carry out online rolling optimization so as to generate engine control parameters in the mode switching process, thereby realizing the smooth transition of the engine thrust and flow in the mode switching process.

2. The predictive control method for a multi-mode turbine engine mode switching process of claim 1, wherein the onboard adaptive composite model comprises a steady-state baseline model, a health parameter estimation module and a propulsion system matrix module, each of which comprises a turbofan mode with main combustor fuel flow as a scheduling variable and a turbojet mode with a mode selector valve angle as a scheduling variable; the steady-state baseline model takes the flying height, mach number and engine adjustable variables under different working modes as input, takes the engine steady-state performance parameters as output, is obtained through training by a deep neural network method and is used for estimating the steady-state performance of the engine in real time; the health parameter estimation module is used for carrying out online real-time estimation on the performance degradation condition of the engine; the propulsion system matrix module is stored with a propulsion system matrix which is built offline in advance and corresponds to working points of different baseline models and reflects the influence condition of component performance degradation parameters on engine performance parameters, and the propulsion system matrix module is used for compensating and correcting the engine steady-state performance parameters output by the steady-state baseline model according to the engine performance degradation condition estimated by the health parameter estimation module.

3. The predictive control method for a mode switching process of a multi-mode turbine engine of claim 2, wherein said health parameter estimation module is implemented by off-line designing a kalman filter by augmenting engine component performance degradation parameters into engine state space model state variables.

4. The predictive control method for a multi-mode turbine engine mode switching process of claim 1, wherein the range of variation of the engine control parameters for building an on-board adaptive composite model is obtained by offline optimization solving in advance the following optimization objective functions:

Wherein F _,obj、ma_2,obj represents the engine thrust and flow before the mode switching, F [ k ] and ma ₂ [ k ] are the engine thrust and flow at the kth step in the mode switching process, u [ k ] is the control amount at the kth step, and ω ₁、ω₂ is the weight coefficient.

5. The predictive control method for a multi-mode turbine engine mode switching process of claim 1, wherein said online rolling optimization is achieved by optimizing solutions to the following objective functions:

Wherein r is an engine control instruction and is a fixed value in the mode switching process; n _p、N_u is the prediction time domain and the control time domain respectively; Estimating values of instruction variables for the prediction model; Δu is the difference between the front and rear timings of the engine control amount, i.e., Δu (k+i) =u (k+i) -u (k+i-1); q and R are semi-positive definite matrices.

6. The prediction control device of the mode switching process of the multimode turbine engine is characterized by comprising a feedback correction loop formed by an airborne self-adaptive composite model and a multimode prediction controller, and the feedback correction loop is used for performing prediction control on the multimode turbine engine in the mode switching process; the variable geometry directional adjustment plan input into the predictive control device is divided into two stages of flameout of the main combustion chamber and directional adjustment of the variable geometry adjustment mechanism; the airborne self-adaptive composite model is used for estimating the thrust and the flow of the engine in real time; the multi-mode predictive controller comprises a turbofan mode predictive controller and a turbojet mode predictive controller which are respectively used for carrying out predictive control on the multi-mode turbine engine in a main combustion chamber flameout stage and a variable geometry adjusting mechanism directional adjusting stage, the turbofan mode predictive controller and the turbojet mode predictive controller both take the airborne self-adaptive composite model as a predictive model, and take the minimum engine thrust and flow change amplitude as performance targets to carry out online rolling optimization so as to generate engine control parameters in the mode switching process, thereby realizing the smooth transition of the engine thrust and flow in the mode switching process.

7. The predictive control apparatus of a multi-mode turbine engine mode switching process of claim 6, wherein said on-board adaptive composite model includes a steady state baseline model, a health parameter estimation module, a propulsion system matrix module, each including two states, a turbofan mode with main combustor fuel flow as a schedule variable and a turbojet mode with a mode selector valve angle as a schedule variable; the steady-state baseline model takes the flying height, mach number and engine adjustable variables under different working modes as input, takes the engine steady-state performance parameters as output, is obtained through training by a deep neural network method and is used for estimating the steady-state performance of the engine in real time; the health parameter estimation module is used for carrying out online real-time estimation on the performance degradation condition of the engine; the propulsion system matrix module is stored with a propulsion system matrix which is built offline in advance and corresponds to working points of different baseline models and reflects the influence condition of component performance degradation parameters on engine performance parameters, and the propulsion system matrix module is used for compensating and correcting the engine steady-state performance parameters output by the steady-state baseline model according to the engine performance degradation condition estimated by the health parameter estimation module.

8. The predictive control apparatus for a multi-mode turbine engine mode switching process of claim 7, wherein said health parameter estimation module is implemented by off-line designing a kalman filter by augmenting engine component performance degradation parameters into engine state space model state variables.

9. The predictive control apparatus for a multi-mode turbine engine mode switching process as set forth in claim 6, wherein a range of variation of said engine control parameters for building an on-board adaptive composite model is obtained by off-line optimization solving in advance the following optimization objective function:

Wherein F _,obj、ma_2,obj represents the engine thrust and flow before the mode switching, F [ k ] and ma ₂ [ k ] are the engine thrust and flow at the kth step in the mode switching process, u [ k ] is the control amount at the kth step, and ω ₁、ω₂ is the weight coefficient.

10. The predictive control apparatus for a multi-mode turbine engine mode switching process of claim 6, wherein said online rolling optimization is achieved by optimizing solutions to the following objective functions:

Wherein r is an engine control instruction and is a fixed value in the mode switching process; n _p、N_u is the prediction time domain and the control time domain respectively; Estimating values of instruction variables for the prediction model; Δu is the difference between the front and rear timings of the engine control amount, i.e., Δu (k+i) =u (k+i) -u (k+i-1); q and R are semi-positive definite matrices.