CN110531622B - Thrust control method of solid rocket engine based on radial basis function neural network - Google Patents

Thrust control method of solid rocket engine based on radial basis function neural network Download PDF

Info

Publication number
CN110531622B
CN110531622B CN201910836072.2A CN201910836072A CN110531622B CN 110531622 B CN110531622 B CN 110531622B CN 201910836072 A CN201910836072 A CN 201910836072A CN 110531622 B CN110531622 B CN 110531622B
Authority
CN
China
Prior art keywords
neural network
model
solid rocket
rocket engine
output
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910836072.2A
Other languages
Chinese (zh)
Other versions
CN110531622A (en
Inventor
齐义文
陈铖
李献领
刘金福
卢少微
刘远强
喻勇涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenyang Aerospace University
719th Research Institute of CSIC
Original Assignee
Shenyang Aerospace University
719th Research Institute of CSIC
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenyang Aerospace University, 719th Research Institute of CSIC filed Critical Shenyang Aerospace University
Priority to CN201910836072.2A priority Critical patent/CN110531622B/en
Publication of CN110531622A publication Critical patent/CN110531622A/en
Application granted granted Critical
Publication of CN110531622B publication Critical patent/CN110531622B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
    • F02K—JET-PROPULSION PLANTS
    • F02K9/00—Rocket-engine plants, i.e. plants carrying both fuel and oxidant therefor; Control thereof
    • F02K9/80—Rocket-engine plants, i.e. plants carrying both fuel and oxidant therefor; Control thereof characterised by thrust or thrust vector control
    • F02K9/805—Rocket-engine plants, i.e. plants carrying both fuel and oxidant therefor; Control thereof characterised by thrust or thrust vector control servo-mechanisms or control devices therefor
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B11/00—Automatic controllers
    • G05B11/01—Automatic controllers electric
    • G05B11/36—Automatic controllers electric with provision for obtaining particular characteristics, e.g. proportional, integral, differential
    • G05B11/42—Automatic controllers electric with provision for obtaining particular characteristics, e.g. proportional, integral, differential for obtaining a characteristic which is both proportional and time-dependent, e.g. P. I., P. I. D.
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
    • G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming

Landscapes

  • Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Automation & Control Theory (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Physiology (AREA)
  • Combustion & Propulsion (AREA)
  • Mechanical Engineering (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Chemical & Material Sciences (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention provides a thrust control method of a solid rocket engine based on a radial basis function neural network, which is characterized by comprising the following steps: step 1, determining a control object, wherein an actuating mechanism adopts a gas regulating system and a pneumatic servo system, and a solid rocket engine is selected as a power device; step 2, establishing a mathematical model of a controlled object of the solid rocket engine; and 3, establishing a reference model and an actual model according to the expected control performance requirement: step 4, designing an RBF neural network controller for self-adaptive control of thrust of the solid rocket engine, and selecting a learning index of the neural network; step 5, optimizing parameters of the RBF neural network controller by using a genetic algorithm; the method optimizes the neural network parameter setting through the genetic algorithm, so that the RBF neural network-based solid rocket engine control system has good tracking capability on the characteristics of the given reference model and has better anti-interference performance and robustness.

Description

Thrust control method of solid rocket engine based on radial basis function neural network
Technical Field
The invention relates to the field of automatic control, in particular to a thrust control method of a solid rocket engine based on a radial basis function neural network.
Background
The solid rocket engine is also called a solid propellant rocket engine, uses a solid propellant, can meet the requirements of a new generation of tactical missile on a power device thereof to the maximum extent, has the advantages of small volume, light weight, high speed, good maneuverability and the like, is an ideal propulsion system, and is a strategic weapon preferentially developed by all countries.
A large number of theories and experiments are made at home and abroad for the solid rocket engine with controllable thrust. In order to ensure that the solid rocket can fly according to a preset orbit, corresponding guidance and control are required. The rocket attitude control is realized through a thrust vector control system, and the deviation caused by various disturbances in flight and the thrust eccentricity of the carrier rocket is compensated by adopting thrust vector control technologies such as a mechanical method or a fluid jet pneumatic method. In recent years, due to the development of neural networks, a radial basis function neural network is widely researched to be applied to the control of a dynamic object, but from the control method perspective, the radial basis function neural network is still the initiative for the application of the radial basis function neural network to the control of a solid rocket, for the control of a solid rocket engine, parameters of the radial basis function neural network can be changed violently along with the environment, meanwhile, unmodeled dynamics also bring certain difficulty to the control, and the radial basis function neural network has a good self-adaptive function and can well overcome the problem.
Disclosure of Invention
The technical task of the invention is to provide a thrust control method of a solid rocket engine based on a radial basis function neural network aiming at the defects of the prior art, and the method provides a thrust control technology capable of adapting to the working environment change and different working conditions of a rocket and improving the control performance of the engine aiming at the problem that the parameters of the solid rocket engine change along with the external environment.
The technical scheme adopted by the invention for solving the technical problems is as follows: a thrust control method of a solid rocket engine based on a radial basis function neural network comprises the following steps:
step 1, determining a control object, wherein an actuating mechanism adopts a gas regulating system and a pneumatic servo system, and a solid rocket engine is selected as a power device;
step 2, establishing a controlled object mathematical model of the solid rocket engine, and selecting a specific working ignition rocket model parameter and a corresponding value as an engine nominal model;
and 3, establishing a reference model and an actual model according to the expected control performance requirement: designing by adopting a proportional-integral controller, taking a closed-loop system under the action of the controller as a reference model of thrust control of the solid rocket engine based on the RBF neural network, and defining a nominal model perturbed by parameters as an actual engine model;
step 4, designing an RBF neural network controller for self-adaptive control of thrust of the solid rocket engine, and selecting a learning index of the neural network;
and 5, optimizing parameters of the RBF neural network controller by using a genetic algorithm: and adjusting the weight of the controller to approach the performance of the reference model according to the error change condition of the reference model output value and the actual model output value of the control system.
Further, the step 2 specifically includes:
the transfer function form of the solid rocket engine control model with the actuating mechanism is as follows:
Figure BDA0002192201760000021
the model parameters at a certain specific working point are as follows: k1Is 0.56, τ1Is 0.0091, tau2Is 0.0039, τ3Is 0.0014, τ40.0022 and mu 0.0026, and the model under the parameters is defined as a nominal model of the engine, and the input and the output are respectively um(t) and ym(t) wherein ym(t)＝L-1[ΔY(s)]，
Figure BDA0002192201760000022
In which the input and output of a complex function are converted by inverse Laplace transform
Figure BDA0002192201760000023
Δ Y(s), input and output u converted to a function of timem(t)、ym(t)。
Further, the step 3 specifically includes:
designing a nominal model by adopting a proportional integral PI controller:
um(t)＝Kpe(t)+Ki∫e(t)dt
wherein, Kp、KiTaking a closed loop system under the action of the parameters of the PI controller as a reference model based on thrust control of the RBF neural network solid rocket engine, wherein the parameters are PI controller parameters, and e (t) is y (t) -r (t), e (t) is a tracking error, and r (t) is a tracking instruction;
the nominal model perturbed by the parameters is an engine actual model, the range of the parameter perturbation is +/-5%, and the input and the output of the actual engine model are defined as u (t) and y (t) respectively.
Further, the method for designing the RBF neural network controller in step 4 includes:
the dynamic RBF neural network topological structure comprises three layers: the method comprises the following steps that an input layer, a hidden layer and an output layer are determined, wherein the network is in an n-m-1 type connection mode, namely the number of neurons in the input layer is n, the number of neurons in the hidden layer is m, and the number of neurons in the output layer is 1;
randomly assigning a weight value of the network during initialization, wherein the input vector of the network is as follows: xi＝[x1 x2…xn]T(ii) a Wherein i represents the number of iterations of the network;
radial basis function selection hj＝exp(-‖xi-cj‖2/2b2) (ii) a Wherein, cj>0 represents the height of the Gaussian base function of the hidden layer of the jth neuron, | x-cjII denotes x to cj(>0) Radial distance of (b)>0 represents the Gaussian basis function width of the hidden layer neurons:
the hidden layer output is defined as H ═ H1 h2…hj…hm]T(ii) a Wherein h isjThe output of the jth neuron of the hidden layer;
RBF network weight is defined as wi＝[w1 w2…wm]T；
The error in defining the outputs of the reference model and the actual model is ec(t)＝ym(t) -y (t); wherein, ym(t), y (t) representing the reference model output and the actual model output, respectively;
the network learning error index is
Figure BDA0002192201760000031
The input/output mapping relation of the network is u (t) ═ w1h1+w2h2+…+wmhm。
Further, the neural network learning algorithm in the step 4 adopts a gradient descent method, namely
Figure BDA0002192201760000032
wi(t)＝wi(t-1)+Δwi(t)+αΔwi(t); wherein eta is>0 is the learning rate, α>0 is the momentum factor.
Further, the specific optimization method of step 5 is as follows:
1. performing population initialization, and performing entity coding on parameters of a neural network controller, wherein each individual is a string of real numbers which are respectively a hidden layer weight of the RBF neural network, a Gaussian basis function width b, a learning rate eta and a momentum factor alpha;
2. the tracking error is chosen as a function of the degree of adaptation, i.e.
Figure BDA0002192201760000041
Figure BDA0002192201760000042
Wherein, ym(t) is the reference model output, and y (t) is the actual model output; k is a constant and is taken as 1;
3. the roulette method is adopted as a selection strategy, and each individual is selected with the probability of
Figure BDA0002192201760000043
Figure BDA0002192201760000044
Wherein, FiIs a value contained in F (t), fiIs the reciprocal of the individual fitness, piA probability of being selected for an individual;
4. real number crossing is used as a crossing strategy, and the probability is selected to be 20%;
5. selecting the ith and jth genes for mutation, wherein the function expression is as follows:
Figure BDA0002192201760000045
wherein, aijIs a gene, amaxAnd aminAs the upper and lower bounds of the gene; r is [ 0-1]A random number in between; f (g) r2(1-g/Gmax)，r2Is a random number, G is the current iteration number, GmaxThe maximum number of evolutions;
6. calculating the fitness F (t), judging whether the minimum error requirement or the maximum iteration times is met, and stopping searching if the minimum error requirement or the maximum iteration times is met; if not, the updating is continued.
The invention has the beneficial effects that: compared with the traditional method for controlling the thrust of the solid rocket engine, the method for controlling the thrust of the solid rocket engine based on the radial basis function neural network optimized by the genetic algorithm utilizes the online learning capacity of the RBF neural network to realize the function of regulating the gas flow of the rocket engine by approaching a reference model; meanwhile, the approximation effect of the reference model is greatly influenced by network parameter selection, inappropriate parameters easily cause network divergence, the overall stability cannot be guaranteed, and the performance of the control system is influenced.
Drawings
FIG. 1 is a diagram of a pneumatic servo system of a solid rocket engine;
FIG. 2 is a schematic diagram of the controller of the present invention;
FIG. 3 is a reference closed loop model structure according to the present invention;
FIG. 4 is a block diagram of the topology of an RBF neural network;
FIG. 5 is a flow chart of a genetic algorithm;
FIG. 6 shows the result of the optimization of the genetic algorithm according to the embodiment of the present invention;
fig. 7 is a diagram of simulation results according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The invention provides a thrust control technology of a radial basis function neural network solid rocket engine based on genetic algorithm optimization, which comprises the following steps:
step 1, determining a control object, wherein a gas regulating system and a pneumatic servo system are adopted as an actuating mechanism, a solid rocket engine is selected as a power device, the pneumatic servo system comprises a high-pressure gas cylinder 1, a switch valve 2, a pressure reducing valve 3, a first control valve 4, a needle valve head 8 and a needle valve head chamber 9, and the formed pneumatic servo system can provide variable combustion chamber pressure so as to push the valve head to move. When the demand for gas flow into afterburner chamber 12 increases, the control valve is adjusted and gas flow from the inlet to the valve head chamber through control valve nozzle 11 causes the gas pressure in the needle head chamber to rise while progressing along valve stem 10. The pressure in the gas generator 5 in the throat region 7 then decreases and the operating pressure and gas generation rate in the gas generator increases, so that the flow through the control valve two 6 increases.
And 2, the transfer function form of the solid rocket engine control model with the actuating mechanism is as follows:
Figure BDA0002192201760000061
the model parameters at a certain specific working point are as follows: k1Is 0.56, τ1Is 0.0091, tau2Is 0.0039, τ3Is 0.0014, τ40.0022 and mu 0.0026, and the model under the parameters is defined as a nominal model of the engine, and the input and the output are respectively um(t) and ym(t) wherein ym(t)＝L-1[ΔY(s)]，
Figure BDA0002192201760000062
In which the input and output of a complex function are converted by inverse Laplace transform
Figure BDA0002192201760000063
Δ Y(s), input and output u converted to a function of timem(t)、ym(t)；
Step 3, according to the expected control performance requirement, a reference model is established, the structure of the reference closed-loop model is shown in fig. 3, and a Proportional Integral (PI) controller is adopted to design a nominal model:
um(t)＝Kpe(t)+Ki∫e(t)dt
wherein, Kp、KiTaking a closed loop system under the action of the parameters of the PI controller as a reference model based on thrust control of the RBF neural network solid rocket engine, wherein the parameters are PI controller parameters, and e (t) is y (t) -r (t), e (t) is a tracking error, and r (t) is a tracking instruction;
defining a nominal model perturbed by parameters as an engine actual model, wherein the parameter perturbation range is +/-5%, and the input and output of the actual engine model are respectively u (t) and y (t);
step 4, designing an RBF neural network controller for self-adaptive control of thrust of the solid rocket engine, wherein FIG. 2 is a controller schematic diagram; determining that the network is in an n-m-1 type connection mode, and fig. 4 is a topological structure diagram of the RBF neural network, namely the number of neurons in an input layer is n, the number of neurons in a hidden layer is m, and the number of neurons in an output layer is 1;
randomly assigning a weight value of the network during initialization, wherein the input vector of the network is as follows: xi＝[x1 x2…xn]T(ii) a Wherein i represents the number of iterations of the network;
radial basis function selection hj＝exp(-‖xi-cj‖2/2b2) (ii) a Wherein, cj>0 represents the height of the Gaussian base function of the hidden layer of the jth neuron, | x-cjII denotes x to cj(>0) Radial distance of (b)>0 represents the Gaussian basis function width of the hidden layer neurons:
the hidden layer output is defined as H ═ H1 h2…hj…hm]T(ii) a Wherein h isjThe output of the jth neuron of the hidden layer;
RBF network weight is defined as wi＝[w1 w2…wm]T；
The output tracking error defining the model reference is ec(t)＝ym(t) -y (t); wherein, ym(t), y (t) representing the reference model output and the actual model output, respectively;
the network learning error index is
Figure BDA0002192201760000071
The input/output mapping relation of the network is u (t) ═ w1h1+w2h2+…+wmhm；
NetworkLearning by gradient descent, i.e.
Figure BDA0002192201760000072
wi(t)＝wi(t-1)+Δwi(t)+αΔwi(t); wherein eta is>0 is the learning rate, α>0 is a momentum factor;
step 5, optimizing the controller by utilizing a genetic algorithm to overcome the limitation in practical engineering, wherein a flow chart of the algorithm is shown in FIG. 5, the RBF neural network controller is simple in structural design, but the convergence effect of the RBF neural network controller depends on the selection of initial parameters to a great extent and is easy to fall into local stability;
firstly, carrying out population initialization, designing the population scale to be 20, carrying out iteration times to be 200, carrying out entity coding on parameters of a neural network controller, wherein each individual is a string of real number strings which are respectively a hidden layer weight of the RBF neural network, a Gaussian basis function width b, a learning rate eta and a momentum factor alpha;
secondly, the tracking error is chosen as a function of the degree of adaptation, i.e.
Figure BDA0002192201760000073
Figure BDA0002192201760000074
Wherein, ym(t) is the reference model output, and y (t) is the actual model output; k is a constant and is taken as 1;
the selection strategy adopts roulette method, and the probability of each individual being selected is
Figure BDA0002192201760000075
Wherein, FiIs a value contained in F (t), fiIs the reciprocal of the individual fitness, piA probability of being selected for an individual;
real number crossing is used as a crossing strategy, and the probability is selected to be 0.2;
selecting the ith and jth genes for mutation, wherein the function expression is as follows:
Figure BDA0002192201760000076
wherein, aijIs a gene, amaxAnd aminAs the upper and lower bounds of the gene; r is [ 0-1]A random number in between; f (g) r2(1-g/Gmax)，r2Is a random number, G is the current iteration number, GmaxThe maximum number of evolutions; the mutation probability is 0.15;
finally, calculating the fitness, judging whether the minimum error requirement or the maximum iteration number is met, and stopping searching if the minimum error requirement or the maximum iteration number is met; if not, the updating is continued, and FIG. 6 shows the optimization result of the genetic algorithm;
step 6, simulation verification: the method utilizes online data to train the RBF neural network, designs the model reference adaptive controller, and takes the sampling period ts as 0.001s and the reference model ym(t) is the output of the solid rocket actuator in a better state, r (t) is a step given signal, and the simulation result is shown in fig. 7.
The technical idea of the present invention is described in the above technical solutions, and the protection scope of the present invention is not limited thereto, and any changes and modifications made to the above technical solutions according to the technical essence of the present invention belong to the protection scope of the technical solutions of the present invention.

Claims (4)

1. A thrust control method of a solid rocket engine based on a radial basis function neural network is characterized by comprising the following steps:
step 1, determining a control object, wherein an actuating mechanism adopts a gas regulating system and a pneumatic servo system, and a solid rocket engine is selected as a power device;
step 2, establishing a controlled object mathematical model of the solid rocket engine, and selecting a specific working ignition rocket model parameter and a corresponding value as an engine nominal model;
and 3, establishing a reference model and an actual model according to the expected control performance requirement: designing by adopting a proportional-integral controller, taking a closed-loop system under the action of the controller as a reference model of thrust control of the solid rocket engine based on the RBF neural network, and defining a nominal model perturbed by parameters as an actual engine model;
step 4, designing an RBF neural network controller for self-adaptive control of thrust of the solid rocket engine, and selecting a learning index of the neural network;
and 5, optimizing parameters of the RBF neural network controller by using a genetic algorithm: adjusting the weight of the controller to approach the performance of the reference model according to the error change condition of the output value of the reference model and the output value of the actual model;
the method for designing the RBF neural network controller in the step 4 comprises the following steps:
the dynamic RBF neural network topological structure comprises three layers: the method comprises the following steps that an input layer, a hidden layer and an output layer are determined, wherein the network is in an n-m-1 type connection mode, namely the number of neurons in the input layer is n, the number of neurons in the hidden layer is m, and the number of neurons in the output layer is 1;
randomly assigning a weight value of the network during initialization, wherein the input vector of the network is as follows: xi＝[x1 x2…xn]T(ii) a Wherein i represents the number of iterations of the network;
radial basis function selection hj＝exp(-||xi-cj||2/2b2) (ii) a Wherein, cj>0 represents the height of the Gaussian base function of the hidden layer of the jth neuron, | | x-cjI denotes x to cj，cj>Radial distance of 0, b>0 represents the Gaussian basis function width of the hidden layer neurons;
the hidden layer output is defined as H ═ H1 h2…hj…hm]T(ii) a Wherein h isjThe output of the jth neuron of the hidden layer;
RBF network weight is defined as wi＝[w1 w2…wm]T；
Defining the error of the output of the reference model and the actual model as ec(t)＝ym(t) -y (t); wherein, ym(t), y (t) representing the reference model output and the actual model output, respectively;
the network learning error index is
Figure FDA0003502921820000021
The input/output mapping relation of the network is u (t) ═ w1h1+w2h2+…+wmhm；
The specific optimization method of the step 5 comprises the following steps:
s01, carrying out population initialization, carrying out entity coding on parameters of a neural network controller, wherein each individual is a string of real numbers which are respectively a hidden layer weight of the RBF neural network, a Gaussian function width b, a learning rate eta and a momentum factor alpha;
s02. selecting the tracking error as the adaptive function, i.e.
Figure FDA0003502921820000022
Figure FDA0003502921820000023
S03, adopting a roulette method as a selection strategy, wherein the probability of each individual being selected is
Figure FDA0003502921820000024
Figure FDA0003502921820000025
Wherein, FiIs a value contained in F (t), fiIs the reciprocal of the individual fitness, piSelecting probability for an individual, taking k as a constant and taking 1;
s04, using real number intersection as an intersection strategy, and selecting the probability as 20%;
s05, selecting the ith and jth genes for mutation, wherein the function expression is as follows:
Figure FDA0003502921820000026
wherein, aijIs a gene, amaxAnd aminAs the upper and lower bounds of the gene(ii) a r is a random number between 0 and 1; f (g) r2(1-g/Gmax)，r2Is a random number, G is the current iteration number, GmaxThe maximum number of evolutions;
s06, calculating fitness F (t), judging whether the requirement of minimum error or the maximum iteration number is met, and stopping searching if the requirement of minimum error or the maximum iteration number is met; if not, the updating is continued.
2. The thrust control method of the solid rocket engine based on the radial basis function neural network as claimed in claim 1, wherein said step 2 specifically comprises:
the transfer function form of the solid rocket engine control model with the actuating mechanism is as follows:
Figure FDA0003502921820000031
the model parameters at a certain specific working point are as follows: k1Is 0.56, τ1Is 0.0091, tau2Is 0.0039, τ3Is 0.0014, τ40.0022 and mu is 0.0026, the model is a nominal model of the engine under the definition of the parameters,
Figure FDA0003502921820000032
representing the gas flow, the input and output are respectively um(t) and ym(t) wherein, for
Figure FDA0003502921820000033
And Δ Y(s) by inverse Ralstonian transform to obtainm(t)＝L-1[ΔY(s)]，
Figure FDA0003502921820000034
Figure FDA0003502921820000035
L-1Is an inverse Laplace transform operator.
3. The thrust control method of the solid rocket engine based on the radial basis function neural network as claimed in claim 2, wherein said step 3 is specifically:
designing a nominal model by adopting a proportional integral PI controller:
um(t)＝Kpe(t)+Ki∫e(t)dt
wherein, Kp、KiTaking a closed loop system under the action of the parameters of the PI controller as a reference model based on thrust control of the RBF neural network solid rocket engine, wherein the parameters are PI controller parameters, and e (t) is y (t) -r (t), e (t) is a tracking error, and r (t) is a tracking instruction;
the nominal model perturbed by the parameters is an engine actual model, the range of the parameter perturbation is +/-5%, and the input and the output of the actual engine model are defined as u (t) and y (t) respectively.
4. The thrust control method of the solid rocket engine based on the radial basis function neural network as claimed in claim 1, wherein the neural network learning algorithm in step 4 adopts a gradient descent method, that is, the method is characterized in that
Figure FDA0003502921820000036
wi(t)＝wi(t-1)+Δwi(t)+αΔwi(t); wherein eta is>0 is the learning rate, α>0 is a momentum factor, wi(t) is the updated network weight, wiAnd (t-1) is the network weight at the last moment.
CN201910836072.2A 2019-09-05 2019-09-05 Thrust control method of solid rocket engine based on radial basis function neural network Active CN110531622B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910836072.2A CN110531622B (en) 2019-09-05 2019-09-05 Thrust control method of solid rocket engine based on radial basis function neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910836072.2A CN110531622B (en) 2019-09-05 2019-09-05 Thrust control method of solid rocket engine based on radial basis function neural network

Publications (2)

Publication Number Publication Date
CN110531622A CN110531622A (en) 2019-12-03
CN110531622B true CN110531622B (en) 2022-04-05

Family

ID=68666888

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910836072.2A Active CN110531622B (en) 2019-09-05 2019-09-05 Thrust control method of solid rocket engine based on radial basis function neural network

Country Status (1)

Country Link
CN (1) CN110531622B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111042928B (en) * 2019-12-30 2020-12-11 大连理工大学 Variable cycle engine intelligent control method based on dynamic neural network
CN111237090B (en) * 2020-01-20 2021-04-16 中国航发沈阳发动机研究所 Vector nozzle deflection control method and system
CN111624886B (en) * 2020-06-05 2022-04-08 沈阳航空航天大学 Variable-cycle aero-engine thrust control method based on SARSA
CN112965387B (en) * 2021-03-31 2022-09-23 西安理工大学 Pneumatic servo system adaptive neural network control method considering state limitation
CN113107711B (en) * 2021-04-20 2022-03-04 北京控制工程研究所 Initial parameter design method of two-component centrifugal engine

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105975651A (en) * 2016-04-12 2016-09-28 哈尔滨工业大学 Genetic particle swarm multi-disciplinary design optimization algorithm-based guided missile parameter design method
CN107423556A (en) * 2017-06-09 2017-12-01 南京理工大学 A kind of computational methods of the Long Range Rocket Gun launch data based on RBF agent model
CN107589665A (en) * 2017-08-25 2018-01-16 中国人民解放军火箭军装备研究院 A kind of trajectory fast Optimization and device based on gradient particle group's algorithm
CN108386276A (en) * 2018-02-27 2018-08-10 南京航空航天大学 Aero-engine based on control of lengthening the life accelerates control real-time optimization method, apparatus
CN108931987A (en) * 2018-07-09 2018-12-04 北京航空航天大学 A kind of Attitude Control System Design method
CN109101717A (en) * 2018-08-07 2018-12-28 重庆大学 Solid propellant rocket Reliability Prediction Method based on reality with the study of fuzzy data depth integration
CN109343341A (en) * 2018-11-21 2019-02-15 北京航天自动控制研究所 It is a kind of based on deeply study carrier rocket vertically recycle intelligent control method
CN109541941A (en) * 2018-11-19 2019-03-29 哈尔滨工业大学 A kind of adaptive augmentation anti-interference fault-tolerance approach for VTOL vehicle boosting flight
CN109779791A (en) * 2019-03-24 2019-05-21 西安航天动力测控技术研究所 Abnormal data intelligent diagnosing method in a kind of solid propellant rocket
CN109884887A (en) * 2019-04-03 2019-06-14 哈尔滨理工大学 A kind of Continuous rotary electrohydraulic servomotor control method based on self-correcting wavelet neural network

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105975651A (en) * 2016-04-12 2016-09-28 哈尔滨工业大学 Genetic particle swarm multi-disciplinary design optimization algorithm-based guided missile parameter design method
CN107423556A (en) * 2017-06-09 2017-12-01 南京理工大学 A kind of computational methods of the Long Range Rocket Gun launch data based on RBF agent model
CN107589665A (en) * 2017-08-25 2018-01-16 中国人民解放军火箭军装备研究院 A kind of trajectory fast Optimization and device based on gradient particle group's algorithm
CN108386276A (en) * 2018-02-27 2018-08-10 南京航空航天大学 Aero-engine based on control of lengthening the life accelerates control real-time optimization method, apparatus
CN108931987A (en) * 2018-07-09 2018-12-04 北京航空航天大学 A kind of Attitude Control System Design method
CN109101717A (en) * 2018-08-07 2018-12-28 重庆大学 Solid propellant rocket Reliability Prediction Method based on reality with the study of fuzzy data depth integration
CN109541941A (en) * 2018-11-19 2019-03-29 哈尔滨工业大学 A kind of adaptive augmentation anti-interference fault-tolerance approach for VTOL vehicle boosting flight
CN109343341A (en) * 2018-11-21 2019-02-15 北京航天自动控制研究所 It is a kind of based on deeply study carrier rocket vertically recycle intelligent control method
CN109779791A (en) * 2019-03-24 2019-05-21 西安航天动力测控技术研究所 Abnormal data intelligent diagnosing method in a kind of solid propellant rocket
CN109884887A (en) * 2019-04-03 2019-06-14 哈尔滨理工大学 A kind of Continuous rotary electrohydraulic servomotor control method based on self-correcting wavelet neural network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Neural network based sensor validation for reusable rocket engines;T.H.Guo等;《Proceedings of 1995 American Control Conference - ACC"95》;20020806;第1367-1372页 *
基于遗传算法的固体火箭发动机参数优化设计;何景轩等;《固体火箭技术》;20041231;第27卷(第4期);第250-254页 *
遗传算法在某固体火箭发动机中的应用;张阳新等;《弹箭与制导学报》;20130228;第33卷(第1期);第129-131页 *

Also Published As

Publication number Publication date
CN110531622A (en) 2019-12-03

Similar Documents

Publication Publication Date Title
CN110531622B (en) Thrust control method of solid rocket engine based on radial basis function neural network
CN107193212B (en) Aero-engine nonlinear predictive control method based on novel wolf optimization algorithm
CN110308649B (en) PID parameter optimization method based on PSO-SOA fusion algorithm and applied to industrial process control
CN110195660B (en) Aero-engine control device based on deep Q learning
CN104635741B (en) Control System for Reusable Launch Vehicle reenters attitude control method
CN107506821A (en) A kind of improved particle group optimizing method
CN111898201B (en) High-precision autonomous attack guiding method for fighter in air combat simulation environment
Bazazzadeh et al. Gas turbine engine control design using fuzzy logic and neural networks
CN112670984A (en) Power system economic load distribution method based on improved whale algorithm
CN116123124A (en) Deep reinforcement learning-based active surge control method and system for gas compressor
CN105807610A (en) Self-adaptive weight matrix weighing pseudo-inverse thrust distribution and saturation handling method
CN115576353A (en) Aircraft formation control method based on deep reinforcement learning
CN115342812A (en) Unmanned aerial vehicle three-dimensional flight path planning method based on improved butterfly optimization algorithm
CN114519292A (en) Air-to-air missile over-shoulder launching guidance law design method based on deep reinforcement learning
Dong et al. Trial input method and own-aircraft state prediction in autonomous air combat
Krishnakumar et al. Solving large parameter optimization problems using genetic algorithms
CN116301028B (en) Multi-constraint on-line flight trajectory planning middle section guiding method based on air suction hypersonic speed platform
Ansari et al. Hybrid genetic algorithm fuzzy rule based guidance and control for launch vehicle
CN112660426A (en) Rocket soft landing guidance method
Qazi et al. Rapid trajectory optimization using computational intelligence for guidance and conceptual design of multistage space launch vehicles
Gui et al. Reaction control system optimization for maneuverable reentry vehicles based on particle swarm optimization
Shengnan et al. Trajectory optimization of a reentry vehicle based on artificial emotion memory optimization
Tanil et al. External configuration optimization of missiles in conceptual design
Montazeri-Gh et al. Fuzzy-based gas turbine engine fuel controller design using particle swarm optimization
Silva et al. Genetic algorithms for satellite launcher attitude controller design

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant