Reverse overrun protection control method and device for recession-oriented variable cycle engine model

Abstract

The invention discloses a recession-oriented variable cycle engine model inverse overrun protection control method. A main control loop and at least one self-adaptive limit protection control loop based on over-limit prediction parallel to the main control loop are used for carrying out the following over-limit protection control on the variable-cycle engine: the main control loop and the at least one self-adaptive limit protection control loop based on over-limit prediction are used; the adaptive limit protection control loop uses a neural network state variable model to predict whether a safety limit parameter which is easy to exceed the safety limit after performance degradation exceeds the limit in the future, and if the prediction exceeds the limit, a model inverse limit protection controller is constructed based on the neural network state variable model. Calculating a value of a control quantity enabling the safety limit parameter not to exceed the limit and transmitting the value to the engine; otherwise, the value of the control quantity generated by the main control loop is transmitted to the engine. The invention further discloses a recession-oriented variable cycle engine model inverse overrun protection control device. According to the invention, the thrust of the engine can be ensured not to be reduced to the greatest extent after safety limit protection is carried out.

Description

Technical Field

[0001] The present invention relates to a control method for over-limit protection of an aero-engine, and in particular to a control method and device for over-limit protection of a variable cycle engine, belonging to the technical field of aero-engine control. Background Art

[0002] The main function of an aero-engine is to provide thrust corresponding to the throttle angle for an aircraft within the safety limits. Over-limit protection control is an important aspect of engine control, and its goal is to prevent the engine from entering working states such as overspeed, over-temperature, and over-pressure, ensure that the engine operates within the safe range, and avoid the aggravation of degradation caused by over-limits, and even threaten the operation safety.

[0003] The variable cycle engine combines the economy of the turbofan engine and the high thrust of the turbojet engine, and has a wider flight envelope and more working states. For future advanced variable cycle engines, advanced multi-operating point matching design technology is used in the design process. Therefore, over-limit conditions usually do not occur in the design state. However, with the performance degradation of the engine, changes in parameters such as the flow capacity and adiabatic efficiency of the engine are brought about. The control plan designed according to the rated working state may limit the operation of the engine, and usually the engine needs to operate in a reduced state. The variable cycle engine has multiple adjustable parameters. By adopting multi-variable participation in limit protection control, the performance potential of the degraded engine can be exploited, and the reserved safety margin can be reduced. The traditional over-limit protection adopts the high-low selection control method. After exceeding the limit, the fuel flow rate of the main control loop is adjusted by the high-low selection (Min-Max) of multiple limit loops to make the variable exceeding the limit return to the limit line. However, at the same time, the thrust of the system will also be reduced, and the maneuverability will be affected, which is not conducive to giving full play to the advantages of multi-variable control. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the deficiencies of the existing over-limit protection control technology for aero-engines, and provide an inverse over-limit protection control method for a variable cycle engine model facing degradation. By online real-time predicting the change trend of variables that are likely to exceed the safety limit after the engine degrades, the switching between the main loop and the limit protection loop is controlled, and it can be realized that the thrust of the engine is not reduced to the greatest extent after entering the safety limit protection.

[0005] The present invention specifically adopts the following technical solutions to solve the above technical problems:

[0006] An inverse over-limit protection control method for a variable cycle engine model facing degradation uses a main control loop and at least one adaptive limit protection control loop based on over-limit prediction parallel to the main control loop to perform the following over-limit protection control on the variable cycle engine:

[0007] In the adaptive limit protection control loop based on overrun prediction, there is a neural network state variable model constructed by using an extreme learning machine for online learning, which takes a safety limit parameter that is prone to exceed the safety limit after performance degradation as the output, and takes the state variables and control variables participating in the limit protection of the safety limit parameter as the inputs. The adaptive limit protection control loop based on overrun prediction uses the neural network state variable model to predict whether the safety limit parameter will exceed the limit at a certain future moment. If it is predicted that the limit will be exceeded, a corresponding model inverse limit protection controller is constructed based on the neural network state variable model. The model inverse limit protection controller takes the limit value of the safety limit parameter, the modeling error of the neural network state variable model, the state variables participating in the limit protection of the safety limit parameter, and the output matrix and direct transmission matrix in the neural network state variable model as input quantities, calculates the value of the control variable that makes the safety limit parameter not exceed the limit, and transmits it to the engine; otherwise, the value of the control variable generated by the main control loop is transmitted to the engine.

[0008] Preferably, the extreme learning machine for online learning uses the state variable with the same response characteristic as the safety limit parameter as the scheduling parameter, and adopts a single hidden layer network structure with a multiplication layer; the hidden layer nodes are divided into two groups, and the outputs of one group are multiplied by the state variables participating in the limit protection of the safety limit parameter in the multiplication layer, and the outputs of the other group are multiplied by the control variables participating in the limit protection of the safety limit parameter in the multiplication layer.

[0009] Preferably, the control variable participating in the limit protection of the safety limit parameter is the control variable with the highest sensitivity to the safety limit parameter, which is determined in advance through sensitivity analysis.

[0010] Preferably, the safety limit parameter that is prone to exceed the safety limit after performance degradation is obtained through the following method in advance: simulating the phenomenon of engine component performance degradation based on the Monte Carlo method, inputting the degradation amount into the engine component-level closed-loop model, and statistically analyzing the variation law of each safety limit parameter of the engine caused by the degradation, so as to determine the safety limit parameter that is prone to exceed the safety limit after engine performance degradation.

[0011] Based on the same inventive concept, the following technical solutions can also be obtained:

[0012] Variable cycle engine model inverse overlimit protection control device for facing decline, including a main control loop and at least one adaptive limit protection control loop based on overlimit prediction in parallel with the main control loop; in the adaptive limit protection control loop based on overlimit prediction, there is a neural network state variable model constructed by an extreme learning machine with online learning, which takes a safety limit parameter that is likely to exceed the safety limit after performance decline as the output, and the state variables and control variables participating in the limit protection of this safety limit parameter as the inputs; the adaptive limit protection control loop based on overlimit prediction uses the neural network state variable model to predict whether the safety limit parameter will exceed the limit at a certain future moment. If it is predicted to exceed the limit, a corresponding model inverse limit protection controller is constructed based on the neural network state variable model. The model inverse limit protection controller takes the limit value of the safety limit parameter, the modeling error of the neural network state variable model, the state variables participating in the limit protection of this safety limit parameter, and the output matrix and direct transfer matrix in the neural network state variable model as input quantities, calculates the value of the control variable that makes the safety limit parameter not exceed the limit, and transmits it to the engine; otherwise, the value of the control variable generated by the main control loop is transmitted to the engine.

[0013] Preferably, the extreme learning machine with online learning uses the state variable with the same response characteristic as the safety limit parameter as the scheduling parameter, and adopts a single hidden layer network structure with a multiplication layer; the hidden layer nodes are divided into two groups. One group of outputs is multiplied by the state variables participating in the limit protection of this safety limit parameter in the multiplication layer, and the other group of outputs is multiplied by the control variables participating in the limit protection of this safety limit parameter in the multiplication layer.

[0014] Preferably, the control variable participating in the limit protection of this safety limit parameter is the control variable with the highest sensitivity to this safety limit parameter, which is determined in advance through sensitivity analysis.

[0015] Preferably, the safety limit parameter that is likely to exceed the safety limit after performance decline is obtained through the following method in advance: based on the Monte Carlo method, simulate the phenomenon of engine component performance decline, input the degradation amount into the engine component-level closed-loop model, and statistically analyze the change rules of each safety limit parameter of the engine caused by degradation, so as to determine the safety limit parameter that is likely to exceed the safety limit after engine performance decline.

[0016] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0017] It can better exert the advantages of the multivariable control of aeroengines in the aspect of limit protection control, fully consider the influence of degradation on the engine safety limit parameters, pre-select the safety limit parameters that are likely to exceed the safety limits after performance degradation to construct the corresponding adaptive limit protection control loop, and actively switch between the main loop control output and the limit control loop output based on over-limit prediction, which can ensure that the thrust of the engine will not be reduced after entering the safety limit.

[0018] The present invention adopts a limit protection control method based on model inversion, constructs a model inversion safety limit controller by online training a neural network state variable model, and effectively improves the control accuracy and adaptive ability of the limit protection control. Brief Description of the Drawings

[0019] Figure 1 is a schematic structural diagram of a high-flow dual-variable cycle engine;

[0020] Figure 2 is a box plot of the influence of degradation on the engine input and output;

[0021] Figure 3 is a schematic structural principle diagram of a high-flow dual-variable cycle engine model inversion over-limit protection control device constructed in a specific embodiment;

[0022] Figure 4 is a structural diagram of a neural network state variable model;

[0023] Figure 5 is a flowchart of the model inversion over-limit protection control method;

[0024] Figure 6 is the simulation result of the acceleration and deceleration process of the engine at the altitude of the economic supercruise operating point in the turbofan mode; where (a) is the change curve of the thrust F, (b) is the change curve of the low-pressure turbine expansion ratio π LT , (c) is the change curve of the temperature T at the outlet of the inter-stage combustion chamber 45 , (d) is the response curve of the high-pressure speed N H , (e) is the response curve of the medium-pressure speed N M , (f) is the response curve of the low-pressure speed N L , (g) is the response curve of the compressor outlet pressure P3, (h) is the response curve of the main fuel flow rate W fz , (i) is the response curve of the fuel flow rate W at the inter-stage combustion chamber fs , (j) is the response curve of the nozzle throat area A8, (k) is the change curve of the control loop signal. Detailed Embodiments

[0025] Aiming at the deficiencies of the prior art, the solution idea of the present invention is to construct a neural network state variable model to online and real-time predict the change trend of the safety limit parameters that are prone to exceed the safety limits after the performance degradation of the variable cycle engine, and generate corresponding control quantities through a limit protection controller based on the inverse of the state variable model to replace the corresponding control quantities generated by the main control loop when the prediction exceeds the limit.

[0026] The method for inverse over-limit protection control of a variable cycle engine facing degradation proposed by the present invention uses a main control loop and at least one adaptive limit protection control loop based on over-limit prediction in parallel with the main control loop to perform the following over-limit protection control on the variable cycle engine:

[0027] In the adaptive limit protection control loop based on over-limit prediction, there is a neural network state variable model constructed by using an extreme learning machine for online learning, with a safety limit parameter that is prone to exceed the safety limit after performance degradation as the output, and the state variables and control quantities participating in the limit protection of this safety limit parameter as the inputs; the adaptive limit protection control loop based on over-limit prediction uses the neural network state variable model to predict whether the safety limit parameter will exceed the limit at a certain future moment. If the prediction exceeds the limit, a corresponding model inverse limit protection controller is constructed based on the neural network state variable model. The model inverse limit protection controller takes the limit value of this safety limit parameter, the modeling error of the neural network state variable model, the state variables participating in the limit protection of this safety limit parameter, and the output matrix and direct transfer matrix in the neural network state variable model as input quantities, calculates the value of the control quantity that makes this safety limit parameter not exceed the limit and transmits it to the engine; otherwise, transmits the value of the control quantity generated by the main control loop to the engine.

[0028] The anti-overlimit protection control device for a variable cycle engine model facing decline proposed by the present invention includes a main control loop and at least one adaptive limit protection control loop parallel to the main control loop based on overlimit prediction; in the adaptive limit protection control loop based on overlimit prediction, there is a neural network state variable model constructed by an extreme learning machine with online learning, which takes a safety limit parameter that is likely to exceed the safety limit after performance decline as the output, and the state variables and control variables participating in the limit protection of this safety limit parameter as the inputs; the adaptive limit protection control loop based on overlimit prediction uses the neural network state variable model to predict whether the safety limit parameter will exceed the limit at a certain future moment. If it is predicted to exceed the limit, a corresponding model inverse limit protection controller is constructed based on the neural network state variable model. The model inverse limit protection controller takes the limit value of the safety limit parameter, the modeling error of the neural network state variable model, the state variables participating in the limit protection of this safety limit parameter, and the output matrix and direct transmission matrix in the neural network state variable model as input quantities, calculates the value of the control variable that makes the safety limit parameter not exceed the limit, and transmits it to the engine; otherwise, the value of the control variable generated by the main control loop is transmitted to the engine.

[0029] Preferably, in order to accurately represent the engine state while minimizing the complexity of the neural network as much as possible, the extreme learning machine with online learning uses the state variable with the same response characteristic as the safety limit parameter as the scheduling parameter and adopts a single hidden layer network structure with a multiplication layer; the hidden layer nodes are divided into two groups, one group of outputs is multiplied by the state variables participating in the limit protection of this safety limit parameter in the multiplication layer, and the other group of outputs is multiplied by the control variables participating in the limit protection of this safety limit parameter in the multiplication layer.

[0030] Preferably, in order to reduce the control complexity, the control variable participating in the limit protection of this safety limit parameter is the control variable with the highest sensitivity to this safety limit parameter, which is determined in advance through sensitivity analysis.

[0031] Preferably, the safety limit parameter that is likely to exceed the safety limit after performance decline is obtained through the following method: based on the Monte Carlo method, the phenomenon of engine component performance decline is simulated, and the degradation amount is input into the engine component-level closed-loop model, and the change rules of each safety limit parameter of the engine caused by degradation are statistically analyzed, so as to determine the safety limit parameter that is likely to exceed the safety limit after engine performance decline.

[0032] For the convenience of public understanding, the technical solution of the present invention will be described in detail below through a specific embodiment in combination with the drawings:

[0033] In this embodiment, the variable cycle engine under study is a high bypass ratio dual variable cycle engine, which adopts a three-rotor and three-combustor structure and a three-duct variable cycle aerodynamic and thermal layout. It has seven rotating components, namely the front fan, rear fan, core fan, compressor, high-pressure turbine, medium-pressure turbine, and low-pressure turbine. As Figure 1 shown, it can achieve a bypass ratio adjustment range far exceeding that of traditional variable cycle engines.

[0034] The adjustable geometric mechanisms of the dual variable cycle engine increase, including the main fuel flow rate W fz , the fuel flow rate W fs of the inter-stage combustor, the throat area A8 of the tail nozzle, the front fan guide vane V Gff , the rear fan guide vane V Grf , the core fan guide vane V Gcf , the compressor guide vane V Gc , etc. The safety limit parameters include: the total pressure P3 at the compressor outlet, the temperature T 45 at the outlet of the inter-stage combustor, and the rotational speeds of the three rotors.

[0035] The increase in adjustable variables makes the engine control system more complex. In order to give full play to the role of adjustable control variables in the engine limit protection control, it is first necessary to carry out a sensitivity analysis of adjustable control variables to safety limit parameters to evaluate the influence of different inputs on safety limit parameters. Taking the supersonic cruise operating point (H = 12 km, Ma = 1.8, both the main combustor and the inter-stage combustor are operating) in the turbofan mode of the high bypass ratio dual variable cycle engine as an example, based on the component-level open-loop model of the engine, small perturbations are made to each control variable, and the changes in the limit protection control parameters caused by the changes in each control variable are shown in Table 1.

[0036] Table 1 Sensitivity analysis of limit protection control

[0037]

[0038]

[0039] It can be found from Table 1 that the sensitivity of the pressure P3 at the outlet of the high-pressure compressor to the main fuel flow rate W fz is the highest, and the sensitivity of the temperature T 45 at the outlet of the inter-stage combustor to the fuel flow rate W fs of the inter-stage combustor is the highest; the sensitivity of the low-pressure rotational speed N L to the V Gff of the front fan guide vane angle is the highest. The sensitivity of the medium-pressure rotational speed N M to the V Gcf of the core fan guide vane angle is the highest, and the sensitivity of the high-pressure rotational speed N H to the V Gc of the high-pressure compressor guide vane angle is the highest. And the influence of the above control variable changes on the limit parameters is monotonic, except for VGc Suppress N H When it exceeds the limit, it will cause T 45 In addition to the increase of T, other changes in control variables can synchronously suppress other output overlimits or cause the change amplitude of other outputs to be more than one order of magnitude smaller. Therefore, the design of one-to-one single-loop overlimit protection control will not bring strong coupling effects.

[0040] At present, the overall performance design of aero-engines can provide high-reliability design solutions under multiple operating conditions. Therefore, for advanced dual variable cycle engines, it can be considered that under the rated state, the overall control law will not cause the engine to enter a restricted state. Therefore, this invention conducts research on the impact of performance degradation on safety limit parameters and conducts Monte Carlo simulations of the degradation of high-flow dual variable cycle engines based on the Ziggurat algorithm.

[0041] Referring to the degradation of each component after the engine has operated for 3000 hours in the NASA report, the flow rate and efficiency degradation are respectively set for seven rotating components. Considering the uncertainty of performance degradation, the deviation of the degradation amount is sampled according to the normal distribution with the deviation μ = 0 and 3σ = 0.15%. The N sets of degradation amount samples are sequentially input into the component-level model of the high-flow dual variable cycle engine. Under the state of closed-loop control of the high-pressure speed and thrust by the main fuel and the area of the tail nozzle and other control variables remaining unchanged, the impact of degradation on the safety limit parameter X is investigated i :

[0042]

[0043] In the formula, x norm is the rated engine output, and x i , i = 1, 2,..., N is the output of the engine under the i-th group of degradation.

[0044] Adopt the main fuel flow rate W fz and the throat area A8 of the tail nozzle to control the thrust F and the low-pressure turbine expansion ratio π LT unchanged, and only investigate the impact of degradation on N L , N M , N H , P3, T 45 , and at the same time give the relative changes of the control variables W fz , W fs and A8, and obtain the data distribution law as Figure 2 shown. From Figure 2 it can be seen that under the condition of closed-loop control of the thrust and the low-pressure turbine expansion ratio, the performance degradation of the engine causes N L , P3, N M to decrease, N H remains almost unchanged, W fz increases by about 1%, and W fsIt increases by about 3.5% to maintain the engine thrust unchanged after recession. The increase in fuel makes T 45 increase by more than 5.5% and A8 decreases. It can be seen that due to the degradation of the compressor performance, the compression ability of air weakens, the power decreases. According to the common working conditions, the turbine work decreases, the turbine outlet temperature increases, and N L 、N M decreases. It can be known from this that when the engine performance degrades, the parameter that is likely to exceed the safety limit is the temperature after the turbine. Therefore, in the working state of the inter-stage combustion chamber, the control quantity W fs is used to perform over-limit protection on the safety limit parameter T 45 .

[0045] For a dual variable cycle engine adopting a direct thrust control structure, this embodiment designs a variable cycle engine model inverse over-limit protection control device facing recession as shown in Figure 3 . It includes a main control loop and an adaptive limit protection control loop based on over-limit prediction. Among them, the main controller is to control the thrust to be constant, adopting a three-input two-output control method. The state quantity x = [N L , N M , N H T , the input quantity u = [W fz , W fs , A8] T , the controlled quantities are the thrust F and the low-pressure turbine expansion ratio π LT , that is, y z = [F, π LT T . In the limit protection control loop, a prediction model is constructed by using a neural network state variable model (NSSE). The input quantities of the NSSE model are the state quantity N H participating in the limit protection and the control quantity W fs . It predicts the output of T 45 at the k + 5 moment and compares it with the corrected limit value (T 45 c). If it exceeds the limit value , then the inter-stage fuel flow command W fs,L of the limit loop is input to the engine, and the limit protection control loop is in the active state; otherwise, the inter-stage fuel flow command W fs,M of the main loop is input to the engine, and the limit loop is in the backup state. The controller of the limit protection control loop is constructed through the NSSE inverse model. The NSSE model is constructed through a new configuration online learning-extreme learning machine (OS-ELM) neural network. e NN4,k is the modeling error of T 45 , which is used for the feedback correction of the limit value.

[0046] ​​In this embodiment, the NSSE model can directly extract the state variable model of a nonlinear system with the help of a new configuration neural network, and then predict the output. At the same time, a model inverse controller can be constructed. To reduce the scale of the network, this paper selects N 45 with the same response characteristics as T H as the scheduling parameter of the network. The network structure is as shown in Figure 4 , where α = N H . As shown in Figure 4 , a single hidden layer network structure with a multiplication layer is adopted (that is, a multiplication layer is added between the hidden layer and the output layer of the neural network). The hidden layer nodes are divided into two groups. One group of outputs is multiplied by the state variable N H in the multiplication layer, and the other group of outputs is multiplied by the control variable W fs in the multiplication layer. Its mathematical expression is as follows:

[0047]

[0048] where W is the connection weight of the input layer, b is the bias of the hidden layer, and β k represents the connection weight of the output layer. σ() is the activation function of the hidden layer, H k = [H 1k , H 2k , M k , are the outputs of the hidden layer, multiplication layer, and output layer respectively. The subscripts k and k + 1 represent the sampling times.

[0049] Then the neural network state variable model constructed in this embodiment is specifically as follows:

[0050]

[0051] where

[0052]

[0053] The subscript j represents the jth node of the hidden layer.

[0054] According to the solution method of the discrete system state variable equation, the predicted value of the safety limit parameter at time k + i can be obtained

[0055]

[0056] At the same time, a feedback correction is introduced to the predicted value. The prediction error of the predicted value at time k is:

[0057]

[0058] where represents the predicted value at time k obtained by using the neural network state variable model at time k - 1.

[0059] The predicted value after feedback correction is:

[0060]

[0061] Based on the predicted value of the safety limit parameter, select the safety limit control quantity transmitted to the engine:

[0062]

[0063] In the formula, T 45,lim is the input of the limit value of the safety limit parameter, u L,k is the control quantity obtained by the limit protection controller, u M,k is the control quantity obtained by the main controller.

[0064] The specific form of the model inverse limit protection controller constructed based on the above neural network state variable model is as follows:

[0065]

[0066] In the formula, is the modeling error of the neural network at time k, T 45,k is the output of the engine at time k, T 45 is the actual value.

[0067] The OS-ELM method with a forgetting factor is used to train the network online, so the established NSSE can accurately represent the engine dynamics at time k, and then the over-limit prediction model and the model inverse controller constructed both have high accuracy.

[0068] In this embodiment, the main control loop adopts a three-input and two-output control method, and the state variables x = [N L , N M , N H T are selected, the input variables u = [W fz , W fs , A8] T are selected, and the controlled variables y z = [F, π LT T . The discrete multivariable robust ALQR control method is adopted for design based on the state variable model established offline.

[0069] For the system

[0070]

[0071] The tracking error e z,k = r - y z,k , and the augmented state vector is taken as Input variable​​ The augmented state equation can be obtained as:

[0072]

[0073] in,

[0074] The performance indicators of ALQR control are defined as:

[0075]

[0076] Among them, Q≥0, R>0 is the weighting matrix, T s is the sampling step size.

[0077] By solving the Riccati equation, the main loop control output can be obtained

[0078]

[0079] The second element u of the main loop control output k (2) Participation restriction protection control, that is, u M,k =u k (2).

[0080] In order to verify the effectiveness of the model inverse limit protection control method proposed in this invention, Figure 5 The process shown in the figure is used to carry out simulation verification of the limited protection control. The limited protection control performance simulation verification of the acceleration and deceleration process is carried out at the supersonic cruise operating point of the high-throughput dual-variable cycle engine turbofan mode. During the period of 5 to 8 seconds, the thrust command increases from 0.9 at the design point to 1.46, and remains unchanged after 8 seconds. During the period of 27 to 30 seconds, the thrust command is reduced from 1.46 to 0.9, as shown in the figure. Figure 6 (a). To ensure that the over-limit state can be entered, the limit value is set to T 45,lim =2300K, the simulation results are as follows Figure 6 The three rotor speeds in the figure are expressed as percentage speeds, and the turbine after temperature T 45 All other physical quantities except for are normalized relative to the design point. For the sake of comparison, the "Main control" in the figure represents the simulation results using only the main loop controller (unlimited protection control), and the "Min-max" represents the simulation results of the limited protection control based on the high-low selection (Min-Max) logic, where the limited protection loop uses W fs,L Control T 45 The single-loop PI controller “NSSE” is the simulation result of the proposed model inverse limiting protection control. Figure 6(k) When the over-limit protection flag signal flag = 1, it indicates that the over-limit state has not been entered and the over-limit protection control loop has not been activated; when flag = 2, the over-limit protection control state is entered and the limit loop is activated. After the limit protection controller takes effect, the controller regulates the fuel flow rate W of the inter-stage combustion chamber fs to decrease, so that T 45 can be limited to 2300K. To maintain the engine thrust F from decreasing, the main loop controller regulates the main fuel flow rate W fz to increase. It can be seen that if the traditional Min-Max selection over-limit protection control method is adopted, T 45 will be quickly limited to the maximum allowable value. Although there are slight fluctuations, the amplitude of the fluctuations is not large. However, the multi-variable control outputs (W fz , A8) after low selection and W fs after low selection are significantly mismatched, causing the responses of other variables, except for the unchanged T 45 , to become slower. During deceleration, since the W fs obtained by the main control loop in the initial stage of deceleration is still higher than the limit loop, it remains low-selected in the PI limit protection control loop for a period of time, with a slow exit time and a secondary switching phenomenon, which is not conducive to stability. And the method proposed in this paper is that during the acceleration stage, when the NSSE controller predicts that the T 45 at k + 5 steps will exceed the limit value, it will actively switch to the limit protection loop and can quickly stabilize at the limit value of T 45 , with a good response. During the deceleration stage, when it predicts that the T 45 at k + 5 steps is less than the set limit value, it will actively switch back to the main loop controller, verifying the effectiveness of the inverse over-limit protection control method of the model of the present invention.

Claims

1. A reverse over-limit protection control method for a variable cycle engine model facing decline, characterized in that, The following over-limit protection control is performed on a variable cycle engine using a main control loop and at least one over-limit prediction-based adaptive limit protection control loop parallel to the main control loop: In the over-limit prediction-based adaptive limit protection control loop, a neural network state variable model is established using an extreme learning machine for online learning, with a safety limit parameter that is prone to exceeding the safety limit after performance degradation as the output, and the state variables and control variables participating in the limit protection of the safety limit parameter as the inputs. The over-limit prediction-based adaptive limit protection control loop uses the neural network state variable model to predict whether the safety limit parameter will exceed the limit at a future moment. If it is predicted to exceed the limit, a corresponding model inverse limit protection controller is constructed based on the neural network state variable model. The model inverse limit protection controller takes the limit value of the safety limit parameter, the modeling error of the neural network state variable model, the state variables participating in the limit protection of the safety limit parameter, and the output matrix and direct transfer matrix in the neural network state variable model as input quantities, calculates the value of the control variable that makes the safety limit parameter not exceed the limit, and transmits it to the engine; Otherwise, the value of the control variable generated by the main control loop is transmitted to the engine.

2. The recession-oriented variable cycle engine model inverse overlimit protection control method according to claim 1, wherein, The extreme learning machine for online learning uses the state variable with the same response characteristic as the safety limit parameter as the scheduling parameter and adopts a single hidden layer network structure with a multiplication layer; the hidden layer nodes are divided into two groups, one group of outputs is multiplied by the state variables participating in the limit protection of the safety limit parameter in the multiplication layer, and the other group of outputs is multiplied by the control variables participating in the limit protection of the safety limit parameter in the multiplication layer.

3. The recession-oriented inverse over-limit protection control method for a variable cycle engine model according to claim 1, characterized in that, The control variable participating in the limit protection of the safety limit parameter is the control variable with the highest sensitivity to the safety limit parameter, which is determined in advance through sensitivity analysis.

4. The reverse over-limit protection control method for the recession-oriented variable cycle engine model according to claim 1, wherein The safety limit parameter that is prone to exceeding the safety limit after performance degradation is obtained through the following method in advance: simulating the phenomenon of engine component performance degradation based on the Monte Carlo method, inputting the degradation amount into the engine component-level closed-loop model, and statistically analyzing the variation law of each safety limit parameter of the engine caused by the degradation, so as to determine the safety limit parameter that is prone to exceeding the safety limit after engine performance degradation.

5. Variable cycle engine model reverse over-limit protection control device for facing decline, characterized in that It includes a main control loop and at least one adaptive limit protection control loop based on overrun prediction that is parallel to the main control loop; in the adaptive limit protection control loop based on overrun prediction, there is a neural network state variable model constructed by an extreme learning machine for online learning, with a safety limit parameter that is prone to exceed the safety limit after performance degradation as the output, and the state variables and control variables that participate in the limit protection of this safety limit parameter as the inputs; the adaptive limit protection control loop based on overrun prediction uses the neural network state variable model to predict whether the safety limit parameter will exceed the limit at a certain future moment. If it is predicted to exceed the limit, a corresponding model inverse limit protection controller is constructed based on the neural network state variable model. The model inverse limit protection controller takes the limit value of this safety limit parameter, the modeling error of the neural network state variable model, the state variables that participate in the limit protection of this safety limit parameter, and the output matrix and direct transfer matrix in the neural network state variable model as input quantities, calculates the value of the control variable that makes this safety limit parameter not exceed the limit, and transmits it to the engine; Otherwise, the value of the control variable generated by the main control loop is transmitted to the engine.

6. The reverse over-limit protection control device for a recession-oriented variable cycle engine model according to claim 5, characterized in that, The extreme learning machine for online learning uses the state variable with the same response characteristic as the safety limit parameter as the scheduling parameter and adopts a single hidden layer network structure with a multiplication layer; the hidden layer nodes are divided into two groups. The output of one group is multiplied by the state variables that participate in the limit protection of this safety limit parameter in the multiplication layer, and the output of the other group is multiplied by the control variables that participate in the limit protection of this safety limit parameter in the multiplication layer.

7. The recession-oriented variable cycle engine model inverse overlimit protection control device according to claim 5, characterized in that, The control variable that participates in the limit protection of this safety limit parameter is the control variable with the highest sensitivity to this safety limit parameter, which is determined in advance through sensitivity analysis.

8. The reverse over-limit protection control device for a recession-oriented variable cycle engine model according to claim 5, wherein, The safety limit parameter that is prone to exceed the safety limit after performance degradation is obtained through the following method in advance: Simulate the phenomenon of engine component performance degradation based on the Monte Carlo method, input the degradation amount into the engine component-level closed-loop model, and statistically analyze the variation law of each safety limit parameter of the engine caused by the degradation, so as to determine the safety limit parameter that is prone to exceed the safety limit after engine performance degradation.

Images