A data-driven based adaptive multi-variable control method for a cycle engine

Abstract

The application discloses a kind of based on data-driven adaptive cycle engine multivariable sliding mode control method, belong to aero-engine control technical field, this method is based on a certain type of adaptive cycle engine, considering the influence of actuator, based on input, output data, the dynamic linearization data model of engine is established online;Three-variable model-free adaptive sliding mode controller based on sliding mode control algorithm is designed;Adaptive weight factor is introduced, the control conservative problem of controller when the controlled object parameter occurs large-scale change is solved, and the control effect is optimized;Finally, the algorithm designed is used to realize the multivariable steady control of adaptive cycle engine under different modes.The application can solve the problem that traditional sliding mode control of aero-engine depends on accurate mathematical model, ensure the control effect while having good real-time performance.

Description

Technical Field

[0001] This invention belongs to the field of aero-engine control, specifically relating to a data-driven adaptive cyclic engine multivariable control method. Background Technology

[0002] Aero-engines are complex nonlinear control systems with multiple inputs and multiple outputs, and uncertainties are unavoidable. On the one hand, with the development of aviation technology in various countries, the complexity of engine operating states has increased, the number of controllable variables has increased, and the coupling between internal parameters has increased. On the other hand, in actual control processes, the model parameters of aero-engines vary greatly throughout the entire flight envelope, and the operating environment is harsh, with uncertainties such as external disturbances and performance degradation. These uncertainties will have a significant impact on the stability and dynamic performance of aero-engine control systems. General feedback control methods and optimal control methods that rely on accurate models are difficult to guarantee control quality under these circumstances. Sliding Model Control (SMC), due to its variable structure characteristics, can cope with the uncertainties of model parameters and the influence of unknown external disturbances, but it still depends on the model of the controlled object. In order to design appropriate control methods when the mathematical descriptions of the controlled object and the environment are not completely determined, data-driven control has gradually been applied to the field of aero-engine control.

[0003] Currently, among data-driven methods, Model-Free Adaptive Control (MFAC) has gradually developed into a complete system, successfully applied in industrial manufacturing, urban expressway traffic control, structural vibration reduction, sheet metal forming, and marine robot control. However, traditional MFAC suffers from slow convergence speed and large overshoot due to inherent dynamic linearization errors. Sliding Mode Controller (SMC) designs the system's sliding surface based on the desired dynamic characteristics, offering advantages such as fast response, insensitivity to parameter changes and disturbances, and simple implementation. Combining MFAC and SMC to construct a Model-Free Adaptive Sliding Mode Controller (MFASMC) allows for the complementary advantages of both methods. Summary of the Invention

[0004] The technical problem to be solved by this invention is to overcome the shortcomings of existing sliding mode control which relies on the mathematical model of the controlled object, and to provide a data-driven adaptive cyclic engine multivariable sliding mode control method. This method combines the multivariable robust control characteristics of sliding mode control with the online learning capability of data-driven model-free adaptive control, and realizes the adaptive adaptation of controller parameters based on the input and output data of the adaptive cyclic engine. This method can cope with system uncertainties caused by environmental changes, performance degradation, etc., and maintain control quality.

[0005] This invention adopts the following technical solution: a data-driven adaptive cyclic engine multivariable sliding mode control method, comprising the following steps:

[0006] Step A: Considering the dynamic characteristics of the actuator, a compact data model is established online based on the output and input data using the dynamic linearization method;

[0007] Step B: Design a multivariable model-free adaptive sliding mode control algorithm and introduce an adaptive weighting factor to improve the control effect;

[0008] Step C: Implement multivariable control based on the input and output data of the adaptive cyclic engine using the control algorithm designed in step B.

[0009] Step A includes the following steps:

[0010] (1) Determining controller parameters

[0011] Considering the dynamic characteristics of the actuator, and based on the impact of the opening degree of the variable geometry components of the adaptive cycle engine on engine performance, the fuel W is selected. fb Tail nozzle throat area A8, rear ejector area A RVABI As a control variable. Based on the correlation analysis of various measurable parameters and thrust of the adaptive cycle engine, high-speed n... H Low-turn n L The pressure ratio EPR is used as the control target.

[0012] (2) Establish a tight-format dynamic linearized data model

[0013] The adaptive cyclic engine can be represented as a multi-input multi-output nonlinear discrete-time system as follows:

[0014] y(k+1)=f(y(k),…,y(kn y ),u(k),…,u(kn u (9)

[0015] Where u(k)∈R m y(k)∈R m These represent the system input and system output at time k, respectively; n y n u It is the unknown order of the system; It is an unknown nonlinear function.

[0016] A three-variable controller is employed, and a compact form dynamic linearization (CFDL) data model of the controlled object at the current operating point is established online based on the output and input data using a dynamic linearization method.

[0017] Δy(k+1)=Φ c (k)Δu(k) (10)

[0018] Where, Δu=[ΔW fb ΔA8 ΔA RVABI ] T To control the quantities: changes in fuel volume, nozzle throat area, and rear ejector area, Δy = [Δn] H ΔEPR Δn L ] T The controlled variables are: changes in high-voltage rotor speed, overall machine pressure ratio, and low-voltage rotor speed. Let φ be the pseudo-Jacobian matrix (PJM) of the system, where φ ij (k), i = 1, 2, 3, j = 1, 2, 3 are elements of the PJM matrix, satisfying |φ ij (k)|≤b1,b2≤|φ ii (k)|≤αb2, α≥1, b2>b1(2α+1)(m-1), i=1,…,m, j=1,…,m, i≠j, and Φ c The sign of all elements in (k) remains unchanged at any time k.

[0019] Using the CFDL data model (10) of a Multiple Input Multiple Output (MIMO) system, the following parameter estimation criterion function is given.

[0020]

[0021] Where μ > 0 is a weighting factor used to penalize excessive changes in the PJM estimate. Minimizing the criterion function (11), the improved projection algorithm is obtained as follows:

[0022]

[0023] Where η∈(0,2] is the step size factor; It is PJMΦ c The estimated value of (k).

[0024] Step B includes the following steps:

[0025] (1) Sliding mode controller design

[0026] Introducing sliding surfaces

[0027] s(k)=c T E(k) (13)

[0028] Where c T = [1 c0], where c0 is a constant greater than 0, E(k) = [e(k) e(k-1)] T e(k) is the tracking error, defined as (14).

[0029] e(k) = y r (k)-y(k) (14)

[0030] Where y r (k) represents the desired tracking trajectory.

[0031] Using sliding mode reaching law

[0032] s(k+1)=s(k)-q·s(k)-εsgn(s(k)) (15)

[0033] The sliding mode controller parameters ε and q satisfy 0 < q < 1 and 0 < ε. Substituting the sliding surface (13) and the sliding mode reaching law (15) into the dynamic linearization model (10), and introducing an adjustable weight factor λ > 0, the control law can be obtained as equation (16).

[0034]

[0035] (2) Design adaptive weighting factors

[0036] Considering an appropriate weighting factor λ can enable the controller to achieve good control performance. However, when the internal parameters of the controlled object change within a large range, using a fixed λ may cause overshoot and oscillation in the output. Therefore, introducing an adaptive weighting factor λ is necessary. v To solve this problem, equation (16) can be rewritten as:

[0037]

[0038] Where λ v As shown in equation (18):

[0039]

[0040] in and Weighting factor λ v Based on the pseudopartial derivative estimate The threshold for switching based on the amplitude, λ max , λ std , λ min Preset values ​​for segmented weighting factors in response.

[0041] Step C includes establishing a compact format data model of the current operating point online based on step A, calculating the multivariable sliding mode control parameters designed in step B in real time, establishing a data-driven adaptive cycle engine multivariable sliding mode control system, and performing multivariable control simulations of a certain type of adaptive cycle engine in different modes at typical operating points.

[0042] The present invention has the following beneficial effects:

[0043] (1) This invention is based on the online dynamic linearization of the input and output parameters of the adaptive cyclic engine, and establishes a tight-form dynamic linearization data model online to perform multivariable sliding mode control. Compared with the linear model method, it solves the problem that the relatively mature sliding mode control developed in aero engines relies on the mathematical model of the controlled object and is difficult to guarantee control quality when faced with uncertain parameters such as aero engine performance degradation. The control system combines the online learning capability of model-free adaptive control and the robust control characteristics of sliding mode control, which can fully exploit the nonlinear characteristics of the engine and ensure the control quality of sliding mode control.

[0044] (2) The present invention introduces an adaptive weighting factor, which can effectively suppress overshoot and oscillation when the parameters of the controlled object change over a wide range, thereby improving the control effect. Attached Figure Description

[0045] Figure 1 This is a diagram showing the working cross-section of an adaptive cycle engine.

[0046] Figure 2 This is a structural diagram of the data-driven adaptive cyclic engine multivariable sliding mode control method designed in this invention;

[0047] Figure 3a The controlled variable n is the control method designed in this invention applied to the three-function mode of the adaptive cycle engine. H Simulation results;

[0048] Figure 3b The results are simulations of the controlled variable EPR of the control method designed in this invention under the three-function mode of the adaptive cycle engine.

[0049] Figure 3c The controlled variable n is the control method designed in this invention applied to the three-function mode of the adaptive cycle engine. L Simulation results;

[0050] Figure 3d The control quantity W is the control quantity of the control method designed in this invention applied in the three-function mode of the adaptive cycle engine. fb Simulation results;

[0051] Figure 3eThe control quantity A8 simulation result is the control method designed in this invention applied to the three-function mode of the adaptive cycle engine.

[0052] Figure 3f The control quantity A is the control quantity A of the control method designed in this invention applied in the three-function mode of the adaptive cycle engine. RVABI Simulation results;

[0053] Figure 4a The controlled variable n is the control method designed in this invention applied to the dual-function mode of an adaptive cyclic engine. H Simulation results;

[0054] Figure 4b The simulation results of the controlled variable EPR of the control method designed in this invention are applied in the dual-function mode of the adaptive cyclic engine.

[0055] Figure 4c The controlled variable n is the control method designed in this invention applied to the dual-function mode of an adaptive cyclic engine. L Simulation results;

[0056] Figure 4d The control quantity W is the control quantity designed by this invention in the dual-function mode of an adaptive cycle engine. fb Simulation results;

[0057] Figure 4e The simulation results of the control quantity A8 are the control quantities of the control method designed in this invention applied to the dual-function mode of the adaptive cycle engine.

[0058] Figure 4f The control quantity A is the control quantity A of the control method designed in this invention applied in the dual-function mode of the adaptive cycle engine. RVABI Simulation results. Detailed Implementation

[0059] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings:

[0060] The working cross-sectional diagram of a certain type of adaptive cycle engine used in this invention is shown below. Figure 1 As shown, the main components include the intake duct, fan, FLADE (Fan-on-blade) fan, core driven fan stage (CDFS), high-pressure compressor, combustion chamber, high-pressure turbine, low-pressure turbine, mixing chamber, afterburner, and exhaust nozzle. The nonlinear mathematical model of the engine is obtained by using the component method based on C language, and then encapsulated into a dynamic link library for digital simulation verification in the MALTAB environment.

[0061] The specific implementation of this invention takes the steady-state control of a certain type of adaptive cycle engine as an example, such as... Figure 2The diagram shows the structure of a data-driven multivariable sliding mode control method for an adaptive cyclic engine. The adaptive cyclic engine is represented by a component-level model. A compact-format dynamic linearized data model of the adaptive cyclic engine at the current operating point is established using process data generated during engine operation. A sliding mode controller is designed based on this model to achieve data-driven multivariable control. By introducing an adaptive weighting factor, overshoot and oscillation are suppressed under large-range changes in the parameters of the adaptive cyclic engine.

[0062] This invention adopts the following technical solution: a data-driven adaptive cyclic engine multivariable sliding mode control method, comprising the following steps:

[0063] Step A: Considering the dynamic characteristics of the actuator, a compact data model is established online based on the output and input data using the dynamic linearization method;

[0064] Step B: Design a multivariable model-free adaptive sliding mode control algorithm and introduce an adaptive weighting factor to improve the control effect;

[0065] Step C: Implement multivariable control based on the input and output data of the adaptive cyclic engine using the control algorithm designed in step B.

[0066] Step A includes the following steps:

[0067] (1) Determining controller parameters

[0068] Considering the dynamic characteristics of the actuator, the performance parameters selected are: FLADE bypass ratio, intermediate bypass ratio, bypass ratio, high-pressure speed, low-pressure speed, FLADE surge margin, fan surge margin, CDFS surge margin, thrust, and fuel consumption rate. Based on the influence of the opening of variable geometry components on engine performance in the adaptive cycle engine, fuel consumption, nozzle throat area, and rear ejector area are selected as control variables. Based on the correlation analysis between the measurable parameters of the adaptive cycle engine and thrust, high and low speeds and pressure ratio are used as control targets.

[0069] (2) Establish a tight-format dynamic linearized data model

[0070] The adaptive cyclic engine can be represented as a multi-input multi-output nonlinear discrete-time system as follows:

[0071] y(k+1)=f(y(k),…,y(kn y ),u(k),…,u(kn u (19)

[0072] Where u(k)∈R m y(k)∈R m These represent the system input and system output at time k, respectively; ny n u It is the unknown order of the system; It is an unknown nonlinear function. The adaptive cyclic engine can be considered a general nonlinear system, satisfying the following assumptions.

[0073] Assumption 1: f i (…), i = 1,…,m, regarding the (n)th... y Each component of the +2) variables has a continuous partial derivative.

[0074] Assumption 2: The system satisfies the generalized Lipschitz condition, that is, for any k1≠k2, k1,k2≥0 and u(k1)≠u(k2), ||y(k1+1)-y(k2+1)||≤b||u(k1)-u(k2)||, where y(k1)≠u(k2)||. i +1)=f(y(k i ),L,y(k i -n y ),u(k i ),L,u(k i -n u )), i=1,2; b>0 is a constant, that is, a bounded change in input energy should produce a bounded change in output energy within the system.

[0075] This invention employs a three-variable controller and uses a dynamic linearization method to establish a compact form dynamic linearization (CFDL) data model of the controlled object at the current operating point based on the output and input data.

[0076] Δy(k+1)=Φ c (k)Δu(k) (20)

[0077] Where, Δu=[ΔW fb ΔA8 ΔA RVABI ] T For the change in the control quantity, Δy = [Δn] H ΔEPR Δn L ] T The change value of the controlled variable. This is the pseudo-Jacobian matrix (PJM) of the system.

[0078] For the sake of rigor in the stability analysis, the following assumptions are made:

[0079] Assumption 3: The system's PJMΦ c(k) is a diagonally dominant matrix that satisfies the following condition (the modulus of each element in the main diagonal is greater than the sum of the moduli of the elements in its row), i.e., |φ ij (k)|≤b1,b2≤|φ ii (k)|≤αb2, α≥1, b2>b1(2α+1)(m-1), i=1,…,m, j=1,…,m, i≠j, and Φ c The signs of all elements in (k) remain unchanged at any time k. The numerical values ​​chosen in this invention are:

[0080]

[0081] This assumption concerns the input-output relationship of closed-loop data. For MIMO nonlinear systems, since the object model is unknown and only the I / O data up to the current time is known, the diagonal dominance condition of the system's input-output data relationship may be the only feasible choice to describe the coupling between the system's variables. For most adaptive control methods, it is a reasonable assumption that the control gain is a constant with a known sign.

[0082] PJM estimation algorithm: Using the CFDL data model (20) of the MIMO system, the following parameter estimation criterion function is given.

[0083]

[0084] Where Δy(k)=y(k)-y(k-1), Δu(k-1)=u(k-1)-u(k-2), μ>0 is a weighting factor used to penalize excessive changes in the PJM estimate. By minimizing the criterion function (22), the improved projection algorithm can be obtained as follows:

[0085]

[0086] Where η∈(0,2] is the step size factor; It is PJMΦ c The estimated value of (k).

[0087] To enhance the estimation algorithm's ability to track time-varying parameters, an algorithm reset mechanism is introduced. If... or

[0088]

[0089] if or i,j = 1, 2, 3, i ≠ j

[0090]

[0091] in for The initial values ​​are i = 1, 2, 3, j = 1, 2, 3; λ > 0, μ > 0, ρ ∈ (0, 1], η ∈ (0, 2).

[0092] Step B includes the following steps:

[0093] (1) Sliding mode controller design

[0094] Introducing sliding surfaces

[0095] s(k)=c T E(k) (26)

[0096] Where c T = [1 c0], where c0 is a constant greater than 0, E(k) = [e(k) e(k-1)] T e(k) is the tracking error, defined as (27).

[0097] e(k) = y r (k)-y(k) (27)

[0098] Where y r (k) represents the desired tracking trajectory.

[0099] Using the sliding mode reaching law:

[0100] s(k+1)=s(k)-q·s(k)-εsgn(s(k)) (28)

[0101] The sliding mode controller parameters ε and q satisfy 0 < q < 1 and 0 < ε. Substituting the sliding surface (13) and the sliding mode reaching law (15) into the dynamic linearization model (10), and introducing an adjustable weight factor λ > 0, the control law can be obtained as equation (16).

[0102]

[0103] Considering the need to satisfy the stability condition of equation (26), the controller parameter c0 needs to satisfy the following condition:

[0104]

[0105] (2) Design adaptive weighting factors

[0106] Considering an appropriate weighting factor λ can enable the controller to achieve good control performance. However, when the internal parameters of the controlled object change within a large range, using a fixed λ may cause overshoot and oscillation in the output. Therefore, introducing an adaptive weighting factor λ is necessary. v To solve this problem, equation (16) can be rewritten as:

[0107]

[0108] Where λ v As shown in equation (18):

[0109]

[0110] in and Weighting factor λ v Based on the pseudopartial derivative estimate The threshold for switching based on the amplitude, λ max , λ std , λ min Preset values ​​for segmented weighting factors in response.

[0111] Step C includes establishing a compact format data model of the current operating point online based on step A, calculating the multivariable sliding mode control parameters designed in step B in real time, establishing a model-free adaptive sliding mode control (MFASMC) system for the adaptive cycle engine, and performing multivariable control simulations on different modes of a certain type of adaptive cycle engine at typical operating points. The typical operating modes of the adaptive cycle engine are shown in Table 1, where MSV stands for Medium Bypass Mode Selection Valve (MSV).

[0112] Table 1 Typical operating modes of the adaptive cycle engine

[0113]

[0114] In the adaptive cycle engine, H=0km, Ma=0, in the intermediate state of the three-function mode, i.e., after normalization W... fb =0.885, A8=0.997, A RVABI =2.42, the initial value of the CFDL data model is taken as The relevant parameters selected are shown in Table 2.

[0115] Table 2 Parameters of the Intermediate State Control System for the Three-Culvert Mode at Ground Points

[0116]

[0117] Figure 3 shows the step response curve of the adaptive cycle engine at the intermediate state of the three-function mode at the ground point, and Table 3 shows the step response performance index of the controller. Based on the simulation results, it can be concluded that applying the MFASMC method to the adaptive cycle engine can effectively enable the three controlled variables to track the control command, maintain a settling time of 5 seconds with no steady-state error, and exhibit only small fluctuations in steady state.

[0118] Table 3 Performance Indicators of Step Response Controller in Intermediate State of Ground Point Tri-mode

[0119]

[0120] In the adaptive cycle engine with H=0km, Ma=0, in the intermediate state of the dual-function mode, i.e., after normalization W... fb =0.82, A8=0.97, A RVABI =2.4, the initial value of the CFDL data model is taken as The relevant parameters selected are shown in Table 4.

[0121] Table 4 Parameters of the intermediate state control system for the ground point dual-path mode

[0122]

[0123] Figure 4 shows the step response curve of the adaptive cycle engine at the intermediate state of the dual-function mode at the ground point, and Table 5 shows the step response performance indicators of the controller. The MFASMC control system designed in this invention can effectively track the control command for the three controlled variables, maintain a settling time of 5 seconds without steady-state error, and exhibit only small fluctuations in steady state.

[0124] Table 5 Performance Indicators of Step Response of Ground Point Dual-Hydraulic Mode Intermediate State Controller

[0125]

[0126] In summary, the data shows that the MFASMC control system designed in this invention only requires about 8 seconds of CPU time to run during a 25-second digital simulation, demonstrating good real-time performance. After the operating mode changes, it can achieve good control results without repeatedly adjusting the initial values ​​of the control system parameters, and can effectively handle the control tasks of the adaptive cycle engine in different modes.

[0127] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0128] This invention utilizes system input and output data to establish a dynamic linearized data model online, overcoming the dependence of the relatively mature sliding mode control in the field of aero-engine control on precise mathematical models. It can realize data-driven multivariable control while ensuring control effect and real-time performance.

Claims

1. A data-driven based adaptive cycle engine multivariable sliding mode control method, characterized in that: Includes the following steps: Step A: Considering the dynamic characteristics of the actuator, a compact data model is established online based on the output and input data using the dynamic linearization method; Step B: Design a multivariable model-free adaptive sliding mode control algorithm and introduce an adaptive weighting factor to improve the control effect; Step C: Implement multivariable control based on the input and output data of the adaptive cyclic engine using the control algorithm designed in Step B; Step A includes the following steps: (1) Select control parameters based on the influence of the opening degree of the variable geometry components of the adaptive cycle engine on engine performance and the correlation analysis between each measurable parameter and thrust; (2) Establish a compact dynamic linearized data model A three-variable controller is employed, and a compact-form dynamic linearized data model of the controlled object at the current operating point is established online based on the output and input data using a dynamic linearization method. ； wherein, are control variables: the variation of the fuel, the nozzle throat area, the rear ejector area, are controlled variables: the variation of the high-pressure rotor speed, the overall pressure ratio, the low-pressure rotor speed, is the pseudo Jacobian matrix of the system, PJM, wherein , , are elements of the PJM matrix; The PJM estimation algorithm is as follows: ； wherein, is a step factor; is an estimate of the PJM of the PJM. Step B includes the following steps: (1) Sliding mode controller design Introducing sliding surfaces ； in , A constant greater than 0 , The tracking error is defined as follows: , ； in To track the desired trajectory; Using sliding mode reaching law ； Among the sliding mode controller parameters , satisfy , ; The control law is formula , ； (2) Design adaptive weighting factors When the internal parameters of the controlled object change significantly, a fixed method is used. This can cause overshoot and oscillation in the output, so an adaptive weighting factor is introduced. The control law is optimized into a formula. , ； in For formula As shown: ； in and Weighting factors Based on the pseudopartial derivative estimate The threshold for switching based on the amplitude. , , Preset values ​​for segmented weighting factors in response.

2. The data-driven adaptive cyclic engine multivariable sliding mode control method as described in claim 1, characterized in that: Step C includes establishing a compact format data model of the current operating point online based on step A, calculating the multivariable sliding mode control parameters designed in step B in real time, establishing a data-driven adaptive cycle engine multivariable sliding mode control system, and performing multivariable control simulations of a certain type of adaptive cycle engine in different modes at typical operating points.

Images