High-disturbance-rejection LA-ALQR mode conversion closed-loop control method for TBCC engine

Abstract

The invention provides a TBCC engine high-disturbance-rejection LA-ALQR mode conversion closed-loop control method. A direct thrust control mode is adopted as a whole. A BP neural network is adopted to establish a thrust prediction model in the modal conversion process, and the deviation values between the total temperature, the total pressure and the Mach number of an inlet of an air inlet and a standard value are increased to serve as model input, so that the influence of environmental disturbance on thrust is accurately reflected. The invention further provides a design method of an augmented linear quadratic regulation controller (LA-ALQR controller) with a linear active-disturbance-rejection compensator, the tracking performance of the system in the control process is ensured by a linear quadratic optimal control theory, and the anti-interference capability of thrust in the conversion process is enhanced by the linear active-disturbance-rejection compensator.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of modeling and control of combined power engine mode conversion process, and particularly relates to a high-disturbance-resistance LA-ALQR mode conversion closed-loop control method for a TBCC engine. BACKGROUND

[0002] At present, a turbine-based combined cycle (TBCC) engine is a research hotspot of a hypersonic aircraft power system. The TBCC engine integrates a turbine engine and a ramjet engine (or a scramjet engine), and realizes wide flight envelope coverage from subsonic speed, supersonic speed to hypersonic speed. The TBCC engine has significant advantages such as horizontal take-off and landing capability, wide range and high economy. The core feature of the TBCC engine is the mode conversion process. The process refers to smooth transition of the engine from the “turbine mode” depending on the turbine engine to the “ram mode” depending on the ramjet engine. During the conversion, the control system operates the actuator to coordinate the closing of the turbine duct and the opening of the ram duct. At the same time, the air flow distribution at the inlet duct outlet is changed, and the main flow is gradually switched from the turbine duct to the ram duct. Through the precise mode conversion capability, the TBCC engine can maintain efficient power output and operation in the entire wide flight speed and height range, thereby supporting the complex task requirements of the hypersonic aircraft.

[0003] At present, the mode conversion control of the turbine-based combined cycle (TBCC) engine mainly adopts an open-loop mode. The core of the method is to directly apply a pre-set mode conversion control program to the actuator of the engine, and to realize the mode conversion through open-loop adjustment of the actuator. The open-loop control strategy has the advantages of simple control system structure and easy engineering implementation. However, the open-loop control mode has a significant disadvantage, that is, poor disturbance resistance. The mode conversion process occurs in a high flight Mach number range much higher than the supersonic cruise Mach number of the turbine engine. In this high Mach number environment, even slight environmental disturbances can cause large fluctuations in the engine thrust during the conversion process. When the thrust fluctuation is severe, it may even endanger flight safety.

[0004] To improve the anti-interference ability of the modal conversion process, some scholars have carried out research on modal conversion closed-loop control method. Nie Lingcong et al. of Beijing Power Machinery Research Institute based on extended Kalman filter ( EFK) to estimate the state parameters such as engine thrust in real time, on this basis, the modal conversion control method of series TBCC engine based on direct thrust form is studied, through the closed-loop control of thrust, flow, surge margin and other direct performance, the transient control error of engine thrust in modal conversion process is not more than 9%, and the steady-state fluctuation is not more than 2.1%[Nie Lingcong, Li Yan, Dai Donghong, et al. Multivariable control of turbo-ramjet combined engine during mode transition[J]. Propulsion technology, 2017, 38(05): 968-974.]. Yu Bingqiang et al. of Nanjing University of Aeronautics and Astronautics cooperated with the institute, put forward a modal conversion closed-loop control method based on neural network prediction feedback and inverse control, introduced a limit management strategy in the multivariable control architecture, which improved the safety of modal conversion[Yu Bingqiang, Zhang Yongliang, Nie Lingcong, et al. Multivariable limit management of TBCC engine based on neural network inverse control[J]. Propulsion technology, 2024, 45(12): 79-89.]. Modal conversion closed-loop control must meet two core requirements: accurate instruction tracking performance to avoid state fluctuation caused by control error; and strong anti-interference ability in high Mach number environment. However, the current closed-loop controller design often separates these two points, and cannot effectively unify tracking performance and anti-interference performance. SUMMARY

[0005] Technical scheme: In order to solve the above technical problems, the present application takes improving the anti-interference ability of TBCC engine modal conversion process as the starting point, carries out the research on modal conversion closed-loop control method, the overall control structure adopts the form of direct thrust control, proposes and designs LA-ALQR modal conversion closed-loop control method, which can significantly improve the anti-interference ability of modal conversion process while ensuring good tracking performance of the controlled quantity, and specifically provides a TBCC engine high anti-interference ability LA-ALQR modal conversion closed-loop control method, which specifically includes designing LA-ALQR modal conversion closed-loop control method based on augmented linear quadratic regulator ALQR and linear auto-disturbance rejection compensator LADRC;

[0006] Wherein the LA-ALQR control structure is the closed-loop control of main combustion chamber fuel quantity W fb , tail nozzle throat area A8, and afterburner fuel quantity W fa , respectively. H , fan rotor speed n Land engine thrust F; the engine thrust F is obtained by a back propagation neural network (BP neural network) to establish a thrust prediction model of the modal conversion process, and the thrust prediction model is obtained by prediction calculation; the mode selection valve (MSV) and the front / rear variable area bypass injector (FVABI / RVABI) are open-loop controlled.

[0007] As an improvement, the BP neural network establishes a thrust prediction model of the modal conversion process, and the thrust prediction model of the modal conversion process is obtained by using a neural network offline training based on a small batch gradient descent method.

[0008] As an improvement, the prediction model includes an input layer, a hidden layer, and an output layer; the modal conversion process is equivalent to a second-order system, the input layer is a value parameter at a current time k and a historical time k-1, k-2, and the value parameter includes a deviation amount ΔT1, ΔP1, ΔMa of the total temperature, the total pressure, and the Mach number at the inlet of the inlet duct from a standard value, an engine closed-loop control variable fuel flow W fb , a nozzle throat area A8, and a thrust augmentation fuel flow W fa .

[0009] As an improvement, the hidden layer is 3 layers, and the number of nodes in the hidden layer is set to 25; the output layer is a predicted value of the engine thrust at the current time k The node functions of the hidden layer and the output layer are all set to Sigmoid functions.

[0010] As an improvement, the neural network output is a predicted thrust value at the current time The actual engine thrust F, and a threshold value of a relative error e F of the neural network prediction model thrust output, wherein the calculation formula of the relative error e F is as follows:

[0011]

[0012] As an improvement, the LA-ALQR control structure includes a baseline controller and a disturbance compensator, the baseline controller is a multivariable controller, including three control loops of a main combustion chamber fuel flow W fb , a closed-loop control compressor rotor speed n H (W fb →n H loop), a nozzle throat area A8 closed-loop control fan rotor speed n L (A8→n L loop), and a thrust augmentation combustion chamber fuel flow W fa closed-loop control engine thrust F (W fa →F loop); the disturbance compensator is a single variable controller, and is a baseline controller W faThe anti-interference compensator is designed based on the F control loop, and the anti-interference compensator takes the difference ΔF between the actual thrust of the engine and the current thrust set value as input, and outputs the afterburning fuel correction amount ΔW fa The actual afterburning fuel amount of the LA-ALQR controller is calculated as follows:

[0013] W fa = W fa,ALQR +△W fa (2)

[0014] In the formula, W fa,ALQR is the afterburning fuel amount calculated by the ALQR controller, and ΔW fa is the afterburning fuel compensation amount calculated by the compensator.

[0015] As an improvement, the baseline controller is a control structure based on the ALQR control theory, and the structure specifically includes setting the engine as a state variable model of a controlled object during operation as follows:

[0016]

[0017] In the formula, u is a system input variable, x is a state variable, y is a system output variable, A, B, C, and D are state variable model coefficients; the control target is to make the state variable model output variable y track the given signal r(t) = a·I(t), wherein a is a constant, I(t) is a unit step function, the control error e = r-y, and e is an error signal; the state variable model is differentiated to obtain,

[0018]

[0019] The augmented state vector is The design object is represented as,

[0020]

[0021] In the formula,

[0022] For the LQR state regulator represented by formula (6), let That is, the state variables are all zero; at the same time of eliminating the steady-state error e = 0, the parameter variables are unchanged, and then the control requirement is met. It is assumed that the performance functional J of formula (4) is,

[0023]

[0024] In the formula, Q = Q T ≥ 0, R = R T > 0, Q ∈ R 9×9 , and R ∈ R 3×3are the appropriate dimensional matrices, then the control law where and P satisfies the Riccati equation,

[0025]

[0026] Let According to and e is expressed as a block matrix then,

[0027]

[0028] Taking Laplace transform of equation (9), the system control law is

[0029]

[0030] where s is a complex frequency.

[0031] As an improvement, the anti-disturbance compensator is designed based on the LADRC control theory to complete anti-interference, and specifically includes that the controlled object combined with the cycle engine can be regarded as a second-order controlled object, as shown in the following formula,

[0032]

[0033] In the formula, x, x', x" are system state variables and their first and second derivatives, respectively, and f(x, x', ω) is a generalized disturbance of the system, let x1=y, x2=y', assume f'=h, and expand the state variable x3=f, then the expanded state equation of the controlled object is,

[0034]

[0035] The linear extended state observer LESO is established as,

[0036]

[0037] In the formula, β1, β2, β3 are adjustable parameters in the LESO, and z1, z2, z3 are estimated values of state variables x1, x2, x3, respectively;

[0038] From the above formula, the characteristic equation of the LESO is,

[0039] λ(s) = s 3 + β1s 2 + β2s + β3 (14)

[0040] According to the 3ω parameter configuration method, the ideal characteristic equation λ(s) = (s + ω0) 3 is selected, and β1 = 3ω0, β2 = 3ω0 2 , and β3 = ω0 3ω 0, namely the observer bandwidth;

[0041] The control law part of the LADRC adopts proportional-derivative PD control, and the PD control law is designed as,

[0042]

[0043] In the formula, k p , k d , and b0 are control parameters of the PD link, which correct the system after disturbance compensation to meet the requirements of various indexes of the system; if the disturbance quantity of the extended observer is basically accurate, namely z3≈f, then,

[0044]

[0045] Further, the closed-loop dynamic equation and the transfer function of the entire LADRC compensator are,

[0046]

[0047] Similarly, according to the 3ω parameter configuration method, the ideal characteristic equation λ(s) of the closed-loop system is selected as λ(s)=(s+ω c ) 3 , k p =ω c 2 , k d =2ω c , and ω c is the controller bandwidth, and by selecting appropriate ω0, ω c , and b0 parameters, a satisfactory control effect can be obtained.

[0048] Beneficial effects: the application proposes a linear active disturbance rejection compensation LA-ALQR mode conversion closed-loop control method, which can ensure good tracking performance of the controlled quantity during the mode conversion process and significantly improve the anti-disturbance ability of the mode conversion process. Compared with the prior art, the thrust fluctuation amplitude under the mode conversion open-loop control of the application is 0.74% at most; the engine thrust fluctuation amplitude is reduced to 0.36% under the ALQR closed-loop control, and the maximum thrust fluctuation amplitude can be further reduced to 0.19% under the LA-ALQR controller, and even at the beginning and end of the mode conversion, the thrust fluctuation amplitude under the LA-ALQR controller is not greater than 0.3%, which proves the effectiveness of the mode conversion closed-loop control proposed by the application. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 The application is an engine mode conversion closed-loop control structure diagram.

[0050] Figure 2The thrust dynamic prediction model of the present application.

[0051] Figure 3 The relative error diagram of the thrust prediction model of the present application, wherein (a) is the training error, and (b) is the test error.

[0052] Figure 4 The modal conversion control system of the present application with an interference compensator.

[0053] Figure 5 The LADRC control structure of the present application.

[0054] Figure 6 The LA-ALQR thrust control effect simulation of the present application, wherein (a) is the high-pressure speed response curve, (b) is the low-pressure speed response curve, (c) is the thrust response curve, and (d) is the afterburner fuel response curve.

[0055] Figure 7 The afterburner fuel flow variation comparison of the present application, wherein (a) is the afterburner fuel flow curve, (b) is the enlarged view of region A, and (c) is the enlarged view of region B.

[0056] Figure 8 The controller anti-interference effect under environmental disturbance of the present application.

[0057] Figure 9 The variation of other key parameters in the modal conversion process under environmental disturbance of the present application, wherein (a) is the variation of the normal shock position, (b) is the high and low pressure rotor speed, (c) is the surge margin, (d) is the air flow, (e) is the afterburner outlet temperature, and (f) is the main combustion chamber outlet temperature. DETAILED DESCRIPTION

[0058] The specific embodiments of the present application will be further described in detail below in conjunction with the examples. The following examples are used to illustrate the present application, but are not used to limit the scope of the present application.

[0059] The present application specifically solves the above technical problems by using the following method:

[0060] 1. The overall direct thrust control form is adopted, the BP neural network is used to establish the thrust prediction model of the modal conversion process, and the deviation amount of the total temperature, total pressure, and Mach number at the inlet of the inlet duct from the standard value is added as the model input, so as to accurately reflect the influence of environmental disturbance on the thrust.

[0061] 2. The design method of the ALQR controller (LA-ALQR controller) with LADRC compensator is proposed, the linear quadratic optimal control theory is used to ensure the tracking performance of the system in the control process, and the linear active disturbance rejection (LADRC) compensator is used to enhance the anti-interference ability of the thrust in the conversion process.

[0062] SeeFigure 1 The diagram shown is a structural diagram of the engine mode conversion closed-loop control of the present invention, specifically including...

[0063] (1) Obtain the mode transition control plan under T2 through H, Ma, and PLA;

[0064] (2) The mode conversion closed-loop control system of the ALQR controller with LADRC self-interference compensator performs open-loop and closed-loop control on the state parameters. The parameters for open-loop control are the mode selection valve MSV and the front / rear variable area duct ejector FVABI / RVABI; the parameters for closed-loop control include the fuel quantity W in the main combustion chamber. fb The exhaust nozzle throat area is A8, and the afterburner fuel quantity is W. fa Controlling the compressor rotor speed n H Fan rotor speed n L The engine thrust F is obtained by predicting and calculating the thrust through a thrust prediction model of the mode transition process established by a BP neural network.

[0065] Furthermore, the BP neural network establishes a thrust prediction model for the mode transition process. This model is obtained through offline training of the neural network based on mini-batch gradient descent. The mode transition process is equated to a second-order system, and the values ​​of the above input quantities at the current time (k) and historical time (k-1, k-2) are collected to establish the neural network thrust prediction model.

[0066] As a specific embodiment of the present invention, the prediction model includes an input layer, a hidden layer, and an output layer: the input layer consists of the current time k and historical time k-1, k-2 value parameters, including the intake manifold inlet total temperature, total pressure, Mach number deviations from standard values ​​ΔT1, ΔP1, ΔMa, and the engine closed-loop control variable fuel flow rate W. fb The exhaust nozzle throat area A8 and the afterburner fuel flow rate W fa The hidden layer has three layers; the output layer is the predicted thrust value.

[0067] Preferably, see Figure 2 As shown, the thrust prediction model contains 18 inputs, hidden layers, and the neural network output is the predicted thrust value at the current moment. The actual thrust F of the engine, and the relative error e of the thrust output of the neural network prediction model. F The threshold, where the relative error e F The calculation formula is as follows:

[0068]

[0069] In the present application, the node functions of the output layer are all set to Sigmoid functions, and the number of nodes of the hidden layer is set to 25. According to the typical working state of the mode transition process, the rated mode transition point (Ma=2.8, H=19.16km) is selected to train the neural network model, and 1684592 simulation data are obtained under the condition of fully exciting the model. After the data samples are randomly shuffled, 90% of them are selected as training samples, and 10% of them are selected as test samples. The relative error of the thrust output predicted by the neural network prediction model is as shown in Figure 3 .

[0070] As shown in Figure 3 , the training relative error of most data points is not greater than 0.1%, and the error of individual points is larger but still not greater than 0.15%. The test error is slightly larger than the training error, but the overall error is still within 0.2%. The calculation accuracy of the whole model is high, which meets the accuracy requirement of the thrust control system.

[0071] As shown in Figure 4 , the thrust controller with a compensator of the present application is an ALQR controller with a self-disturbance compensator LADRC. The characteristics of the compensator are that the non-zero disturbance input and the zero reference input r(t)=0. When the engine produces small-range fluctuations around the set value due to some disturbance during operation, the compensator takes the difference ΔF between the actual thrust of the engine and the current thrust set value as the input, and outputs the afterburning fuel correction amount ΔW fa , so as to realize the control target r(t)=0, thereby reducing the thrust fluctuation of the original control system under the condition of disturbance and improving the anti-interference ability of the system.

[0072] That is, the LA-ALQR control structure of the present application includes a baseline controller and an anti-interference compensator. The baseline controller is a multivariable controller, which includes three control loops of the main combustion chamber fuel amount W fb , the closed-loop control compressor rotor speed n H (W fb →n H loop), the tail nozzle throat area A8 closed-loop control fan rotor speed n L (A8→n L loop), and the afterburning chamber fuel amount W fa closed-loop control engine thrust F(W fa →F loop). The anti-interference compensator is a single-variable controller, which is designed on the basis of the W fa →F control loop of the baseline controller to enhance the anti-interference ability of the thrust in the transition process. The anti-interference compensator takes the difference ΔF between the actual thrust of the engine and the current thrust set value as the input, and outputs the afterburning fuel correction amount ΔW fato achieve the control objective r(t) = 0; the actual afterburner fuel flow of the LA-ALQR controller is calculated as follows

[0073] W fa = W fa,ALQR +△W fa (2)

[0074] wherein W fa,ALQR is the afterburner fuel flow calculated by the ALQR controller, and△W fa is the afterburner fuel flow compensation calculated by the compensator.

[0075] Further, the baseline controller is a control structure based on the ALQR control theory, which specifically includes setting the engine as a state variable model of the controlled object when working as follows:

[0076]

[0077] wherein u is the system input variable, x is the state variable, y is the system output variable, A, B, C, and D are the state variable model coefficients; the control objective is to make the state variable model output variable y track the given signal r(t) = a·I(t), wherein a is a constant, I(t) is the unit step function, the control error e = r-y, and e is the error signal; the state variable model is differentiated to obtain,

[0078]

[0079] The augmented state vector is The design object is represented as,

[0080]

[0081] wherein,

[0082] For the LQR state regulator represented by equation (6), let that is, all state variables are zero; at the same time when eliminating the steady-state error e = 0, each parameter variable remains unchanged, then the control requirement is met, and the performance functional J of equation (4) is set as,

[0083]

[0084] wherein Q = Q T ≥ 0, R = R T > 0, Q ∈ R 9×9 , and R ∈ R 3×3 are appropriate dimension matrices, then the control law wherein and P satisfies the Riccati equation as follows,

[0085]

[0086] will be described in detail below according to the present application and e are expressed as block matrices then,

[0087]

[0088] Taking Laplace transform to equation (9), the system control law is

[0089]

[0090] where s is complex frequency.

[0091] For the modal transformation closed-loop control system, the engine state variable is selected as x = [W fb A8 W fa n H n L F], the engine output variable is y = [n H n L F], and the control variable is u = [W fb A8 W fa ].

[0092] The anti-interference compensator in the present application is designed based on the LADRC control theory to complete anti-interference, as shown in Figure 5 The LADRC control structure diagram in the present application is shown in the figure, and the controlled object combination circulation engine can be regarded as a second-order controlled object, as shown in the following formula,

[0093]

[0094] In the formula, x, x', x" are system state variables and their first and second derivatives, respectively, and f(x, x', ω) is the generalized disturbance of the system, let x1=y, x2=y', and assume f'=h, and the extended state variable x3=f, then the extended state equation of the controlled object is,

[0095]

[0096] The linear extended state observer LESO is established as,

[0097]

[0098] In the formula, β1, β2, β3 are adjustable parameters in the LESO, and z1, z2, z3 are estimated values of state variables x1, x2, x3, respectively;

[0099] From the above formula, the characteristic equation of the LESO is,

[0100] λ(s) = s3 +β1s 2 +β2s+β3 (14)

[0101] According to the 3ω parameter configuration method, an ideal characteristic equation λ(s) = (s + ω0) 3 is selected, and β1 = 3ω0, β2 = 3ω0 2 , and β3 = ω0 3 are obtained, wherein ω0 is an observer bandwidth;

[0102] The control law part of the LADRC adopts proportional-derivative PD control, and the PD control law is designed as,

[0103]

[0104] In the formula, k p , k d , and b0 are control parameters of the PD link, which corrects the system after disturbance compensation to meet the requirements of various indexes of the system; if the disturbance quantity of the extended observer is basically accurate, that is, z3 ≈ f, then,

[0105]

[0106] Further, the closed-loop dynamic equation and the transfer function of the entire LADRC compensator are,

[0107]

[0108] According to the 3ω parameter configuration method, an ideal characteristic equation λ(s) = (s + ω c ) 3 is selected, and k p = ω c 2 , k d = 2ω c , and ω c is a controller bandwidth, and by selecting appropriate ω0, ω c , and b0 parameters, a satisfactory control effect is obtained.

[0109] Embodiment 1

[0110] Next, taking the throttle lever step simulation under the mode transition flight condition as an example, the LA-ALQR controller design process and working process of the present application are introduced and described through specific numerical values.

[0111] In order to avoid the physical actual values of the above variables being greatly different, the variables are normalized before the state variable model is solved. The engine state variable model at the equilibrium point is obtained by least square identification as follows:

[0112]

[0113] In the design process, only the state weighting matrix Q and the input weighting matrix R are designed, and Q and R are usually selected as diagonal matrices. The last two diagonal elements of the Q matrix are the weights of the error components in the state variables (n L , n H ), and in order to improve the response speed of the control system and realize high-precision tracking of the engine speed instruction, the weights can be appropriately increased. Finally, Q = diag([1 0.5 1 1 1 2 80 100 150]) and R = ([0.8 0.8 1]) are selected, and by solving the Riccati equation, the following can be obtained,

[0114]

[0115] After the block is obtained,

[0116]

[0117] In the LADRC compensator of the application, the ideal characteristic equation of the closed-loop system is selected as λ(s) = (s + ω c ) 3 , ω0 = 12, ω c = 6, according to the 3ω parameter configuration method, k p = ω c 2 , k d = 2ω c , further calculation can obtain β1 = 36, β2 = 432, β3 = 1728, k p = 36, k d = 12.

[0118] In order to verify the effectiveness of the above controller, the throttle lever step simulation is carried out under the mode transition point flight condition, the thrust instruction is greatly stepped at t = 10s, the simulation of the engine two rotor speed is greatly stepped at t = 30s, and the engine parameter change is as shown in Figure 6 From Figure 6 , it can be seen that when the thrust instruction is stepped or the speed instruction is stepped, the corresponding controlled parameters of the engine can quickly track the instruction signal, and there is no steady-state error. In order to improve the tracking performance of the controller, when a large step is performed, a little overshoot appears in the two speeds, as shown in Figure 6 (a), (b), but the overshoot is controlled within 0.6%, and when the controlled quantity instruction is continuously changed in the actual mode transition process, no overshoot phenomenon appears. Figure 6 (c), the thrust control loop is in the LA-ALQR control mode after the core engine speed is stepped (t = 30s), and the amplitude of the thrust F is reduced from 8.15% to 3.82%. FromFigure 6 (d)It can be seen that the addition of the LADRC anti-interference compensator plays a role in advance compensation for the afterburner fuel flow W fa .

[0119] The above simulation shows that both the LA-ALQR control method and the ALQR control method have good tracking performance, but the LA-ALQR control method performs better in thrust anti-interference.

[0120] Embodiment 2

[0121] Next, the closed-loop control effects of ALQR and LA-ALQR are analyzed through modal transition process performance comparison, and the technical solution of the application is further introduced and described.

[0122] See Figure 7 It can be seen that the addition of the LADRC anti-interference compensator plays a role in advance compensation for the afterburner fuel flow W fa , and it is this advance compensation that suppresses the fluctuation of the thrust. In actual flight, the influence of atmospheric disturbance is inevitable, especially for high Mach number flight states, and it is unrealistic not to consider the influence of external disturbance on the modal transition process. Therefore, the atmospheric turbulence model disturbance output is applied to the TBCC propulsion system modal transition process, and the modal transition control effects under different control modes under atmospheric turbulence disturbance are simulated, as shown in Figure 8 . From Figure 8 (a), it can be seen that under the open-loop control mode, the engine thrust appears sustained large fluctuations, and the fluctuation amplitude is ΔF = 0.74%; compared with the open-loop modal transition control, under the closed-loop control mode, the engine thrust fluctuation amplitude is obviously decreased. Under the ALQR control, except that the thrust fluctuation amplitude is large at the beginning and end of the modal transition process, the maximum thrust fluctuation amplitude is ΔF = 0.36%, which is less than 50% of the open-loop. And under the LA-ALQR control, the maximum thrust fluctuation amplitude is further reduced to ΔF = 0.19%, and even at the beginning and end of the modal transition process, the thrust fluctuation amplitude is not greater than 0.3%. It is proved that the LA-ALQR-based modal transition closed-loop control advantage is more significant. From Figure 6 (b), it can be seen that under the external disturbance condition of atmospheric turbulence, the LADRC interference compensator also plays a role in advance compensation.

[0123] As shown in Figure 9 , under atmospheric disturbance, the position x A of the inlet normal shock wave fluctuates significantly under both open-loop and closed-loop control. This is because when flying at supersonic speed, the downstream engine control is difficult to affect the upstream inlet state. A The fluctuation of x a21 directly affects the total pressure recovery coefficient, and further causes the key parameters (fan / ram duct flow W / WaR1 , combustion chamber outlet temperature (T4, T7), surge margin (S mL , S mH )) fluctuate greatly. Among them, the afterburner outlet total temperature T7 fluctuates particularly severely, and even appears long-time large over-temperature under open-loop control, which is not allowed in flight; under closed-loop control, although the x A dither forces the afterburner fuel to be adjusted greatly to maintain the thrust, thereby causing T7 to change, but the closed-loop control of the present application significantly reduces the fluctuation amplitude of T7, effectively reduces the over-temperature time, and achieves the expected goal.

[0124] Overall, under the closed-loop control based on LA-ALQR, the fluctuation amplitude of the thrust is not greater than 0.3%, and the fluctuation amplitude of the afterburner outlet total temperature is effectively suppressed. Compared with the open-loop control, the overall effect of the closed-loop control proposed by the present application is satisfactory.

[0125] The above-described embodiments only express several embodiments of the present application, which are described in detail and specifically, but should not be understood as a limitation on the scope of the patent. It should be noted that for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, which are all within the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A TBCC engine high-disturbance-capability LA-ALQR mode transition closed-loop control method, characterized in that: The LA-ALQR modal transformation closed-loop control method is designed based on an augmented linear quadratic regulator (ALQR) and a linear active disturbance rejection compensator (LADRC); Wherein LA-ALQR control structure is main combustion chamber fuel quantity W fb , nozzle throat area A8, afterburner fuel quantity W fa Respectively closed loop control compressor rotor speed n H , fan rotor speed n L And engine thrust F;The engine thrust F is obtained by the prediction calculation of the thrust prediction model of the mode conversion process established by the back propagation neural network BP neural network;Mode selection valve MSV and front / rear variable area duct ejector FVABI / RVABI are open loop control.

2. The TBCC engine high-disturbance capability LA-ALQR mode transition closed-loop control method of claim 1, wherein: The thrust prediction model of the modal transformation process is established by using a BP neural network, and the thrust prediction model of the modal transformation process is obtained by using the neural network offline training based on the small batch gradient descent method.

3. The TBCC engine high-disturbance capability LA-ALQR mode transition closed-loop control method according to claim 1 or 2, characterized in that: The prediction model comprises an input layer, a hidden layer and an output layer. The input layer is equivalent to a second-order system of the modal conversion process, and the value parameters of the current time k and the historical time k-1, k-2 include the deviation amount ΔT1, ΔP1, ΔMa of the inlet total temperature, total pressure and Mach number of the inlet of the air inlet from the standard value, the closed-loop control variable fuel flow W of the engine, the throat area A8 of the tail nozzle and the afterburner fuel flow W. fb fa ​​ 4. The TBCC engine high-disturbance capability LA-ALQR mode transition closed-loop control method of claim 3, wherein: The implicit layer is 3 layers, the number of nodes of the implicit layer is set to 25; the output layer is the engine thrust prediction value at the current time k The node functions of the implicit layer and the output layer are set to Sigmoid functions.

5. The TBCC engine high-disturbance capability LA-ALQR mode transition closed-loop control method of claim 3, wherein: The neural network output is a predicted thrust value at the current time The engine actual thrust F, the relative error e of setting the neural network prediction model thrust output F The threshold value of the relative error e F The calculation formula of the relative error e is as follows:

6. The TBCC engine high-disturbance capability LA-ALQR mode transition closed loop control method of claim 1, wherein: The LA-ALQR control structure includes a baseline controller and a disturbance compensator, the baseline controller is a multivariable controller, including the main combustion chamber fuel quantity W fb The closed loop control compressor rotor speed n H That is, W fb →n H Loop, the tail nozzle throat area A8 closed loop control fan rotor speed n L That is, A8→n L Loop, the afterburner fuel quantity W fa The closed loop control engine thrust F, that is, W fa →F loop; the disturbance compensator is a single variable controller, which is designed on the basis of the baseline controller W fa →F control loop, and the disturbance compensator is used to enhance the anti-interference ability of the thrust in the conversion process, the disturbance compensator takes the difference ΔF between the actual engine thrust and the current thrust set value as input, and outputs the afterburner fuel correction amount ΔW fa To realize the control target r(t)=0; the actual afterburner fuel quantity of the LA-ALQR controller is calculated as follows W fa = W fa,ALQR + ΔW fa (2) where W fa,ALQR is the afterburner fuel quantity calculated by the ALQR controller, ΔW fa is the afterburner fuel compensation quantity calculated by the compensator.

7. The TBCC engine high-disturbance capability LA-ALQR mode transition closed-loop control method of claim 1 or 6, wherein: The baseline controller is a control structure based on the ALQR control theory, and the structure specifically includes setting the engine as a controlled object state variable model during operation: In the formula, u is a system input variable, x is a state variable, y is a system output variable, A, B, C, and D are state variable model coefficients; the control objective is to make the state variable model output variable y track the given signal r(t)=a*I(t), wherein a is a constant, I(t) is a unit step function, the control error e=r-y, and e is an error signal; The differential of the state variable model is, The augmented state vector is The design object can be represented as In the formulae, For the LQR state regulator represented by formula (6), let That is, the state variables are all zero; while eliminating the steady-state error e = 0, each parameter variable is unchanged, then it meets the control requirements, set the performance functional J of formula (4) as where Q = Q T ≥ 0, R = R T > 0, Q ∈ R 9×9 , R ∈ R 3×3 are appropriate dimensional matrices, respectively, then the control law where and P satisfies the following Riccati equation, will be described below. According to the present application, the following embodiments are provided. and e are expressed as block matrices Then, The Laplace transform of formula (9) is performed, and the system control law is, In the formula, s is a complex frequency.

8. The TBCC engine high-disturbance capability LA-ALQR mode transition closed-loop control method of claim 1 or 6, wherein: The anti-disturbance compensator is designed based on the LADRC control theory to complete the anti-interference, and specifically includes that the combined circulation engine as a controlled object can be regarded as a second-order controlled object, as shown in the following formula, In the formula, x, x', and x" are system state variables and their first and second derivatives, respectively, f(x, x', ω) is a generalized disturbance of the system, x1=y, x2=y', and f'=h are assumed, and the extended state variable x3=f, then the extended state equation of the controlled object is, The linear extended state observer (LESO) is established as, In the formula, β1, β2, and β3 are adjustable parameters in the LESO, and z1, z2, and z3 are estimated values of the state variables x1, x2, and x3, respectively; From the above formula, the characteristic equation of the LESO is, λ(s) = s 3 + β1s 2 + β2s + β3 (14) According to the 3ω parameter configuration method, an ideal characteristic equation λ(s) = (s + ω0) is selected 3 , and β1 = 3ω0, β2 = 3ω0, and β3 = ω0 are obtained 2 , where ω0 is the bandwidth of the observer 3 ​ The control law part of the LADRC adopts proportional-derivative (PD) control, and the PD control law is designed as, where k p , k d and b0 are control parameters of the PD element, which correct the system after disturbance compensation to meet the system requirements; if the disturbance of the extended observer is basically accurate, i.e. z3≈f, then And the closed-loop dynamic equation and the transfer function of the entire LADRC compensator are Similarly, according to the 3ω parameter configuration method, the ideal characteristic equation of the closed loop system λ(s) = (s + ω c ) 3 , k p = ω c 2 , k d = 2ω c , ω c is the controller bandwidth, and by selecting appropriate ω0, ω c and b0 parameters, a satisfactory control effect is obtained.

Images