A method for controlling stealth signal characteristics of aircraft engines

Abstract

The present application belongs to the field of aircraft engine design, and is a method for controlling the stealth signal characteristics of an aircraft engine, comprising: constructing a neural network, a PID controller, and an engine stealth model to form an engine stealth characteristic control system based on a neural network; by constructing a neural network based on PID regulation, a closed-loop control is formed between the control amount of the engine and the control object, and three parameters of the PID controller are adjusted online by the neural network. When the stealth signal characteristics of the engine need to be controlled, only the target value to be achieved needs to be input, and the arbitrary nonlinear expression ability of the neural network is utilized. According to the engine operating state and stealth control requirements, the characteristic parameters are adaptively adjusted online to realize automatic learning of the neural network, and the desired target value is achieved by continuously correcting the amount, thereby finally achieving optimized control of stealth performance, which is simple and stable to use.

Description

Technical Field

[0001] The present application belongs to the field of aircraft engine design, and in particular relates to a method for controlling stealth signal characteristics of an aircraft engine. Background Art

[0002] Excellent stealth capability can greatly improve the penetration capability and battlefield survivability of military aircraft, while the stealth performance of aircraft engines is a key factor in whether an aircraft can achieve omnidirectional stealth, and is also one of the important technical indicators of advanced aircraft engines of the fourth generation and above. The aircraft engine is a highly nonlinear, time-varying, and complex aerodynamic and thermal system with a wide operating envelope. Within the operating range, as environmental conditions and operating conditions change, the engine's infrared and radar stealth signal characteristics will change greatly.

[0003] At present, there is no method to control the characteristics of engine stealth signals. Traditional engines all use the engine's altitude and speed characteristics as control targets. The engine's infrared and radar stealth performance is only a characteristic performance under a certain state and cannot be accurately controlled. It cannot fully adapt to the wide envelope of modern engines that is constantly extending, and at the same time achieve a balance between engine aerodynamic performance and stealth performance.

[0004] Therefore, how to achieve dynamic adaptive intelligent control of the engine's stealth characteristics while ensuring stable engine operation is a problem that needs to be solved. Summary of the invention

[0005] The purpose of this application is to provide a method for controlling the stealth signal characteristics of an aircraft engine to solve the problem in the prior art that it is impossible to accurately control the stealth performance of the engine, such as infrared and radar.

[0006] The technical solution of the present application is: a method for controlling the stealth signal characteristics of an aircraft engine, comprising: constructing a neural network, a PID controller, and an engine stealth model, determining the number of nodes in the input layer, the number of nodes in the hidden layer, and the number of nodes in the output layer, determining the control quantity, the control object, the error function, the state function of each layer of the neural network, and the initial weight of each node in the neural network, and forming an engine stealth characteristic control system based on a neural network; sampling and obtaining a group of engine control quantities, inputting them into the engine stealth model, obtaining an actual stealth signal value of a single output, calculating the deviation between the system output quantity and the single output value through the error function, and obtaining the deviation quantity; normalizing the deviation quantity, the actual stealth signal value, the rated stealth signal value, and the threshold, and inputting them into the input layer of the neural network; the input layer sequentially calculates each input quantity through the state function and the initial weight of each node, and reaches the hidden layer, and the hidden layer calculates the input quantity reaching the hidden layer through the state function and the initial weight of each node, and reaches the output layer, and the output layer outputs the proportional coefficient k p , integration coefficient k i, differential coefficient k d To the PID controller; the PID controller calculates the output of the neural network output layer and outputs a new control quantity. The control quantity enters the engine stealth model for calculation again to form a secondary output actual stealth signal value, and again obtains a new deviation through the error function, and repeats this process until the actual stealth signal value and the system output are within the set threshold range.

[0007] Preferably, the specific construction method of the engine stealth feature control system is: using a neural network to calculate the proportional coefficient k p , integration coefficient k i , differential coefficient k d The parameters of the three links are adjusted through online adaptive learning to control the controlled object; the control quantity and control target are selected according to the engine type and scheme; the control object is selected according to the engine stealth control target; the engine parameters constituting the object function S(t) are determined according to the engine control object; the parameters are adjusted through adaptive learning of neural network according to the system output S(t) and the target error error(t), the engine control quantity is dynamically adjusted, and the expected target value is output.

[0008] Preferably, the method for determining the neural network structure is: determining the input layer of the neural network according to the engine rated stealth signal value RCS or IR, the engine actual real-time stealth signal value RCS or IR, the deviation error(t) and a stable constant term; determining the hidden layer of the neural network according to the neuron error training function, and the neuron error training function is:

[0009]

[0010] Among them, p is the number of samples, m is the number of neurons in the output layer, and q jp Represents the model output, y jp Represents the sample output; observe the changes in the network's training error and prediction error until the training error reaches the minimum and stabilizes, and the prediction error reaches the minimum.

[0011] Preferably, the method for constructing the neuron state function of the neural network is: determining the hidden layer state function as: Determine the output layer state function as:

[0012] Preferably, the system output and system control quantity are determined according to the real-time deviation obtained by the neural network performance index function; an incremental control algorithm is used to make the output node of the neural network correspond to the three parameters O1=k of the PID controller. p 、O2=k i 、O3=k d; By constructing a learning performance indicator function and assigning initial weight values ​​to each node of the neural network, the neural network can be self-learned.

[0013] Preferably, the real-time deviation is:

[0014] error(t)=Sr-S(t)

[0015] ; The system output is:

[0016] S(t)=S(t-1)+k p (error(t)-error(t-1))+k i error(t)+k d (error(t)-2error(t)+error(t-2))

[0017] ; The system control quantity is:

[0018] u(K)=u(K-1)+k p [e(K)-e(K-1)]+k i e(K)+k d [e(K)-2e(K)+e(k-2)].

[0019] Preferably, the calculation method of the three output parameters of the PID controller is:

[0020]

[0021]

[0022]

[0023] Preferably, the specific method of the neural network self-learning is: establish a learning performance index function E(t)=lr(Sr-S(t)), where lr∈(0-1); modify the weight coefficient of the network according to the gradient descent method, and add an inertia term that enables the search to converge quickly to the global minimum; the weight coefficient is:

[0024]

[0025] Where η is the learning efficiency, α is the inertia coefficient, α∈(0, 0.1); the weight k-time weight determination method is adopted, k is a natural number not less than 2; the initial value matrix of weight w1 is selected as a 5×4 random number matrix on the interval [a, b], a, b∈(-0.5, 0.5); the initial value matrix of weight w2 is selected as a 3×5 random number matrix on the interval [c, d], c, d∈(-0.5, 0.5), and the initial value of network weight is determined; neural network learning is carried out and the weight coefficient Δw2 is adjusted kj (t) and Δw2ji (t), realize adaptive learning and adjustment of control parameters.

[0026] Preferably, the method for constructing the engine stealth model of the stealth characteristic signal S(t) is as follows: the relationship between the change of the radiation energy of the engine when it is transmitted in the medium and the absorption, emission and scattering of the medium is given to determine the infrared radiation intensity; the relationship between the effective radiation energy of the engine gray body wall and the gray body's own radiation and reflected radiation is given; the spatial distribution characteristics of the radiation energy emitted by the radiation source are given to determine the infrared radiation characteristic signal; an expression for the engine radar radiation area is given; an expression for the surface scattering field of the engine scattering source is given; an engine state characteristic model is established to determine the radar scattering characteristic signal.

[0027] Preferably, the relationship between the change of the radiation energy of the engine when it is transmitted in the medium and the absorption, emission and scattering of the medium is:

[0028]

[0029] Where s represents the microelement position of the engine radiation source, L λ and L bλ denote the spectral radiance and blackbody spectral radiance respectively, κ λ is the medium spectral absorption coefficient, γ λ is the medium spectral scattering coefficient, ω is the solid angle;

[0030] The relationship between the effective radiation energy of the engine gray body wall and the gray body's own radiation and reflected radiation is:

[0031] H λ =∑ Ω L λ (s,ω)cosβ ω dω

[0032] ; The spatial distribution characteristics of the radiation energy emitted by the radiation source are

[0033]

[0034] Among them, I λ Indicates the target's own radiation intensity, The illumination from the radiation source at the detection point, R is the distance between the target and the detector, τ λ,atm is the atmospheric transmittance between the target and the detector;

[0035] The engine radar scattering area is:

[0036]

[0037] Where r is the distance from the target to the receiver, S sis the power density of the target scattered field, S i is the power density of the incident wave, E s is the electric field strength of the scattered field, E i is the electric field strength of the incident field;

[0038] The surface scattering field of the engine scattering source:

[0039]

[0040] Where s represents the microelement position of the engine radiation source, L λ and L bλ denote the spectral radiance and blackbody spectral radiance respectively, κ λ is the medium spectral absorption coefficient, γ λ is the medium spectral scattering coefficient, ω is the solid angle;

[0041] The engine state characteristic model is expressed as:

[0042]

[0043]

[0044] where π T is the turbine pressure drop ratio, T6 is the total temperature of the turbine outlet, W is the fuel supply, A8 is the nozzle throat area, H and Ma are the height and Mach number respectively, ε, ε′ and μ′ are the emissivity, dielectric constant and magnetic permeability of the engine target material respectively.

[0045] The present invention discloses a method for controlling the stealth signal characteristics of an aero-engine, comprising: constructing a neural network, a PID controller, and an engine stealth model, determining the number of input layer nodes, the number of hidden layer nodes, and the number of output layer nodes, determining the control quantity, the control object, the error function, the state functions of each layer of the neural network, and the initial weight of each node of the neural network, so as to form an engine stealth characteristic control system based on a neural network; by constructing a neural network based on PID regulation, the control quantity of the engine and the control object are closed-loop controlled, and three parameters of the PID controller are adjusted online by the neural network. When the stealth signal characteristics of the engine need to be controlled, only the target value to be achieved needs to be input, and the arbitrary nonlinear expression ability of the neural network is utilized. According to the engine operating state and the stealth control requirements, the characteristic parameters are adjusted online and adaptively, so as to realize automatic learning of the neural network, and the desired target value is achieved by continuously correcting the quantity, so as to finally achieve stealth performance optimization control, which is simple and stable to use. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solution provided by the present application, the following is a brief introduction to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present application.

[0047] Figure 1 This is a schematic diagram of the overall process of this application;

[0048] Figure 2 This is a schematic diagram of the overall structure of the engine stealth feature control system of this application;

[0049] Figure 3 Schematic diagram of the neural network structure of this application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the implementation of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below in conjunction with the drawings in the embodiments of this application.

[0051] A method for the characteristic of aircraft engine stealth signal is proposed. The engine stealth performance control is taken as the goal, the classic PID control method is improved, and a multi-layer artificial neural network model is constructed. Through the self-learning and error feedback of the multi-layer BP artificial neural network, the three control functions of proportion, integration and differentiation are adjusted to form a relationship of mutual coordination and mutual restriction in the control quantity, and the best combination can be found from the changing nonlinear combination. The specific scheme is as follows.

[0052] like Figure 1 As shown, the following steps are included:

[0053] like Figure 2 , Figure 3 As shown, step S100, construct a neural network, a PID controller, and an engine stealth model, wherein the neural network includes an input layer, a hidden layer, and an output layer; by determining the number of nodes in the input layer, the number of nodes in the hidden layer, and the number of nodes in the output layer, the control quantity, the control object, the system output quantity, the error function, the state function of each layer of the neural network, and the initial weight of each node in the neural network, a neural network-based engine stealth feature control system is formed;

[0054] The specific construction method of the engine stealth feature control system is:

[0055] The neural network is a BP neural network, and the BP neural network has a proportional coefficient k p , integration coefficient k i , differential coefficient k d The parameters of the three links are adjusted through online adaptive learning to control the controlled object S(t).

[0056] The control quantity is selected according to the engine type and scheme. Preferably, for an aviation turbofan engine, the parameters of engine turbine post-temperature T6(t), nozzle throat area A8(t), and fuel supply amount W(t) are selected as the control quantities.

[0057] The control object is selected through the engine stealth control target. Preferably, the engine infrared radiation intensity IR and radar scattering cross section RCS are used as the target value of the stealth feature control object S(t), thereby forming the object function S(t).

[0058] According to the engine control object, the engine stealth model constituting the object function S(t) is determined. Preferably, the flight altitude H of the engine, the flight Mach number Ma, the surface emissivity ε of the key signal source components, the material dielectric constant ε', the magnetic permeability μ, etc. are used as input parameters of the engine stealth model.

[0059] The control system scheme uses a neural network to adaptively learn and adjust parameters according to the system output S(t) and the target error error(t), dynamically adjusts the engine control quantity, and finally outputs the expected target value to achieve the expected stealth performance optimization control of the engine.

[0060] The neural network uses a 4-5-3 quantitative structure, that is, the number of output layer nodes is 4, the number of hidden layer nodes is 5, and the number of output layer nodes is 3.

[0061] Preferably, the method for constructing the neural network structure is:

[0062] The input layer of the neural network is determined according to the engine rated stealth signal value RCS or IR, the engine actual real-time stealth signal value RCS or IR, the deviation error(t) and the threshold. The threshold is a stable constant term, preferably 1 or 2; the hidden layer also contains a threshold node, preferably 1 or 2.

[0063] The hidden layer of the neural network is determined according to the neuron error training function, wherein the neuron error training function is:

[0064]

[0065] Among them, p is the number of samples, m is the number of neurons in the output layer, and q jp Represents the model output, y jp Represents sample output;

[0066] Observe the changes in the network's training error and prediction error until the training error reaches the minimum and stabilizes, and the prediction error reaches the minimum.

[0067] Preferably, the method for constructing the neuron state function of the neural network is:

[0068] Determine the hidden layer state function as:

[0069] Determine the output layer state function as:

[0070] That is, the parameters reaching the input layer enter the hidden layer after being calculated by the hidden layer state function, and the parameters of the hidden layer enter the output layer after being calculated by the output layer state function, forming the proportional coefficient k p , integration coefficient k i , differential coefficient k d Three link parameters.

[0071] Preferably, the neural network algorithm is constructed as follows:

[0072] Obtain the real-time deviation of the neural network performance index function to determine the system output and system control quantity; the real-time deviation is:

[0073] error(t)=Sr-S(t)

[0074] ; The system output is:

[0075] S(t)=S(t-1)+k p (error(t)-error(t-1))+k i error(t)+k d (error(t)-2error(t)+error(t-2))

[0076] ; The system control quantity is:

[0077] u(K)=u(K-1)+k p [e(K)-e(K-1)]+k i e(K)+k d [e(K)-2e(K)+e(k-2)].

[0078] Using the incremental control algorithm, the output nodes of the neural network correspond to the three parameters of the PID controller O1 = k p 、O2=k i 、O3=k d ; Among them, the calculation methods of the proportional, integral and differential coefficients of the PID controller are:

[0079]

[0080]

[0081]

[0082] By constructing a learning performance index function and assigning initial weight values ​​to each node of the neural network, the neural network can be self-learned. The specific method of neural network self-learning is:

[0083] Establish a learning performance indicator function E(t) = lr(Sr-S(t)), where lr∈(0-1);

[0084] The weight coefficients of the network are modified according to the gradient descent method, and an inertia term is added to make the search converge quickly to the global minimum; the weight coefficients are:

[0085]

[0086] Among them, η is the learning efficiency, α is the inertia coefficient, α∈(0, 0.1);

[0087] The weight k-time weight determination method is adopted, where k is a natural number not less than 2; the initial value matrix of weight w1 is selected as a 5×4 random number matrix on the interval [a, b], a, b∈(-0.5, 0.5); the initial value matrix of weight w2 is selected as a 3×5 random number matrix on the interval [c, d], c, d∈(-0.5, 0.5), and the initial value of network weight is determined;

[0088] Perform neural network learning and adjust the weighting coefficient Δw2 kj (t) and Δw2 ji (t), realize the adaptive learning adjustment of control parameters. K, j, i represent different positions of neural network nodes, such as j = 3, i = 2, representing the second node from top to bottom of the output layer.

[0089] Preferably, the method for constructing the engine stealth model of the stealth characteristic signal S(t) is:

[0090] 1) Establish infrared radiation characteristic signal model

[0091] The relationship between the change of radiation energy when the engine radiation energy is transmitted in the medium and the absorption, emission and scattering of the medium is given to determine the infrared radiation intensity; the relationship is:

[0092]

[0093] Where s represents the microelement position of the engine radiation source, L λ and L bλ denote the spectral radiance and blackbody spectral radiance respectively, κ λ is the medium spectral absorption coefficient, γ λ is the medium spectral scattering coefficient, ω is the solid angle;

[0094] The relationship between the effective radiation energy of the engine gray body wall and the gray body's own radiation and reflected radiation is given as H λ =∑ Ω L λ (s,ω)cosβ ω dω;

[0095] Given the spatial distribution characteristics of the radiation energy emitted by the radiation source, the infrared radiation characteristic signal is determined as

[0096] Among them, I λ Indicates the target's own radiation intensity, The illumination from the radiation source at the detection point, R is the distance between the target and the detector, τ λ,atm is the atmospheric transmittance between the target and the detector.

[0097] 2) Establish radar scattering characteristic signal model

[0098] The expression of engine radar radiation area is given as:

[0099]

[0100] Where r is the distance from the target to the receiver, S s is the power density of the target scattered field, S i is the power density of the incident wave, E s is the electric field strength of the scattered field, E i is the electric field strength of the incident field;

[0101] The expression of the surface scattering field of the engine scattering source is given as:

[0102]

[0103] Where s represents the microelement position of the engine radiation source, L λ and L bλ denote the spectral radiance and blackbody spectral radiance respectively, κ λ is the medium spectral absorption coefficient, γ λ is the medium spectral scattering coefficient, ω is the solid angle;

[0104] The engine state characteristic model is established as follows:

[0105]

[0106]

[0107] where π T is the turbine pressure drop ratio, T6 is the total temperature of the turbine outlet, W is the fuel supply, A8 is the nozzle throat area, H and Ma are the height and Mach number respectively, ε, ε′ and μ′ are the emissivity, dielectric constant and magnetic permeability of the engine target material respectively.

[0108] Thereby determining the radar scattering characteristic signal.

[0109] Step S200, sampling and acquiring a group of engine control quantities, inputting them into the engine stealth model, obtaining an actual stealth signal value of a primary output, and calculating the deviation between the system output quantity and the primary output value through an error function to obtain a deviation quantity;

[0110] Step S300, normalizing the deviation, the actual stealth signal value, the rated stealth signal value and the threshold, and inputting them into the input layer of the neural network;

[0111] Step S400: The input layer calculates each input quantity in turn through the state function and the initial weight of each node, and reaches the hidden layer. The hidden layer calculates the input quantity reaching the hidden layer through the state function and the initial weight of each node, and reaches the output layer. The output layer outputs the proportional coefficient k p , integration coefficient k i , differential coefficient k d To PID controller;

[0112] In step S500, the PID controller calculates the output of the neural network output layer and outputs a new control quantity. The control quantity enters the engine stealth model for calculation again to form a secondary output actual stealth signal value, and obtains a new deviation through the error function again. This process is repeated until the actual stealth signal value and the system output are within the set threshold range.

[0113] By constructing a neural network based on PID regulation, the control quantity of the engine and the control object form a closed-loop control. The three parameters of the PID controller are adjusted online by the neural network. When the stealth signal characteristics of the engine need to be controlled, only the target value to be achieved needs to be input. By utilizing the arbitrary nonlinear expression ability of the neural network, the characteristic parameters are adaptively adjusted online according to the engine operating status and stealth control requirements, and the neural network is automatically learned. The desired target value is achieved by continuously correcting the amount, and finally the stealth performance optimization control is achieved. It is simple and stable to use.

[0114] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for controlling stealth signal characteristics of an aircraft engine, characterized in that: include: Construct a neural network, a PID controller, and an engine stealth model, determine the number of input layer nodes, the number of hidden layer nodes, and the number of output layer nodes, determine the control quantity, the control object, the error function, the state function of each layer of the neural network, and the initial weight of each node of the neural network, and form an engine stealth feature control system based on a neural network; A set of engine control quantities are sampled and input into the engine stealth model to obtain the actual stealth signal value of the primary output. The deviation between the system output and the primary output value is calculated through the error function to obtain the deviation; The deviation, the actual stealth signal value, the rated stealth signal value and the threshold are normalized and input into the input layer of the neural network; The input layer calculates each input quantity in turn through the state function and the initial weight of each node, and reaches the hidden layer. The hidden layer calculates the input quantity reaching the hidden layer through the state function and the initial weight of each node, and reaches the output layer. The output layer outputs the proportional coefficient k p , integration coefficient k i , differential coefficient k d To PID controller; The PID controller calculates the output of the neural network output layer and outputs a new control quantity. The control quantity enters the engine stealth model for calculation again to form the actual stealth signal value of the secondary output, and obtains a new deviation through the error function again. This process is repeated until the actual stealth signal value and the system output are within the set threshold range. The method for constructing the engine stealth model of the stealth characteristic signal S(t) is: Give the relationship between the change of radiation energy when the engine radiation energy is transmitted in the medium and the absorption, emission and scattering of the medium, and determine the infrared radiation intensity; The relationship between the effective radiation energy of the engine gray body wall and the gray body's own radiation and reflected radiation is given; Give the spatial distribution characteristics of the radiation energy emitted by the radiation source and determine the infrared radiation characteristic signal; Give the expression of engine radar radiation area; The expression of the surface scattering field of the engine scattering source is given; Establish engine status characteristic model and determine radar scattering characteristic signal; The relationship between the change of the radiation energy of the engine when it is transmitted in the medium and the absorption, emission and scattering of the medium is: Where s represents the microelement position of the engine radiation source, L λ and L bλ denote the spectral radiance and blackbody spectral radiance respectively, κ λ is the medium spectral absorption coefficient, γ λ is the medium spectral scattering coefficient, ω is the solid angle; The relationship between the effective radiation energy of the engine gray body wall and the gray body's own radiation and reflected radiation is: H λ =∑ Ω L λ (s,ω)cosβ ω dω; The spatial distribution characteristics of the radiation energy emitted by the radiation source are: Among them, I λ Indicates the target's own radiation intensity, The illumination from the radiation source at the detection point, R is the distance between the target and the detector, τ λ,atm is the atmospheric transmittance between the target and the detector; The engine state characteristic model is expressed as: where π T is the turbine pressure drop ratio, T6 is the total temperature of the turbine outlet, W is the fuel supply, A8 is the nozzle throat area, H and Ma are the height and Mach number respectively, ε, ε′ and μ′ are the emissivity, dielectric constant and magnetic permeability of the engine target material respectively.

2. The method for controlling the stealth signal characteristics of an aircraft engine according to claim 1, characterized in that: The specific construction method of the engine stealth feature control system is: The proportionality factor k is calculated by neural network p , integration coefficient k i , differential coefficient k d The parameters of the three links are adjusted through online adaptive learning to control the controlled object; Select control quantity and control target according to engine type and scheme; Select the control object according to the engine stealth control target; According to the engine control object, determine the engine parameters constituting the object function S(t); According to the system output S(t) and the target error error(t), the neural network is used to adaptively learn and adjust the parameters, dynamically adjust the engine control quantity, and output the expected target value.

3. The method for controlling the stealth signal characteristics of an aircraft engine according to claim 2, characterized in that: The method for determining the neural network structure is: Determine the input layer of the neural network according to the engine rated stealth signal value RCS or IR, the engine actual real-time stealth signal value RCS or IR, the deviation error(t) and a stable constant term; The hidden layer of the neural network is determined according to the neuron error training function, wherein the neuron error training function is: Among them, p is the number of samples, m is the number of neurons in the output layer, and q jp Represents the model output, y jp Represents sample output; Observe the changes in the network's training error and prediction error until the training error reaches the minimum and stabilizes, and the prediction error reaches the minimum.

4. The method for controlling the stealth signal characteristics of an aircraft engine according to claim 3, characterized in that: The method for constructing the neural network neuron state function is: Determine the hidden layer state function as: Determine the output layer state function as:

5. The method for controlling the stealth signal characteristics of an aircraft engine according to claim 4, characterized in that: The construction method of the neural network algorithm is: Determine the system output and system control quantity according to the real-time deviation obtained by the neural network performance index function; Using the incremental control algorithm, the output nodes of the neural network correspond to the three parameters of the PID controller O1 = k p 、O2=k i 、O3=k d ; By constructing a learning performance indicator function and assigning initial weight values ​​to each node of the neural network, the neural network can be self-learned.

6. The method for controlling the stealth signal characteristics of an aircraft engine according to claim 5, characterized in that: The real-time deviation is: error(t)=Sr-S(t); The system output is: S(t)=S(t-1)+k p (error(t)-error(t-1))+k i error(t)+k d (error(t)-2error(t)+error(t-2)); the system control quantity is: u(K)=u(K-1)+k p [e(K)-e(K-1)]+k i e(K)+k d [e(K)-2e(K)+e(k-2)]。 7. The method for controlling the stealth signal characteristics of an aircraft engine according to claim 6, characterized in that: The calculation method of the three output parameters of the PID controller is:

8. The method for controlling the stealth signal characteristics of an aircraft engine according to claim 7, characterized in that: The specific method of the neural network self-learning is: Establish a learning performance indicator function E(t) = lr(Sr-S(t)), where lr∈(0-1); The weight coefficients of the network are modified according to the gradient descent method, and an inertia term is added to make the search converge quickly to the global minimum; the weight coefficients are: Among them, η is the learning efficiency, α is the inertia coefficient, α∈(0, 0.1); The weight k-time weight determination method is adopted, where k is a natural number not less than 2; the initial value matrix of weight w1 is selected as a 5×4 random number matrix on the interval [a, b], a, b∈(-0.5, 0.5); the initial value matrix of weight w2 is selected as a 3×5 random number matrix on the interval [c, d], c, d∈(-0.5, 0.5), and the initial value of network weight is determined; Perform neural network learning and adjust the weighting coefficient Δw2 kj (t) and Δw2 ji (t), realize adaptive learning and adjustment of control parameters.

Images