A recoverable rocket control effectiveness evaluation system and method

Abstract

The application discloses a recoverable rocket control performance evaluation system and method, the system comprises a multi-granularity analysis module, a hybrid twin correction module, a simulation working condition setting module and a control performance evaluation module; the method comprises the following steps: configuring a rocket model granularity and a simulation environment granularity according to a task requirement; modeling components, dynamics and an environment according to a granularity configuration result; if hybrid twin correction is to be performed, calculating a residual error by using real data and simulation data, training a neural network by using the residual error, correcting a model by using the neural network, configuring a working condition according to a task requirement; selecting a working condition, completing recoverable flight control simulation, and performing control performance evaluation. The application adopts a multi-granularity mode to construct a rocket and a model of each executing mechanism of the rocket, can set different simulation granularities according to requirements, and can flexibly evaluate recoverable flight control performance of the rocket.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aerospace technology, and more particularly to a recoverable rocket control effectiveness evaluation system and method. BACKGROUND

[0002] With the increasing demand for deep space exploration, satellite launch and other space missions, the recoverable launch vehicle with the characteristics of low cost and reusability will become the main carrier. For the recoverable launch vehicle, the guidance method in the vertical recovery process is the key technology to reduce fuel consumption, ensure recovery accuracy and ultimately realize the reusable rocket.

[0003] Rocket recovery flight control generally uses the first stage engine of the rocket itself, four orthogonally installed grid fins and yaw attitude thrusters to control the orbit and attitude of the body. In the ideal case, all actuators work normally, and the rocket recovery flight control can be equivalent to an optimal control problem. The rocket can complete the vertex soft landing by flying according to the nominal optimal trajectory. However, due to the complex and variable internal and external environment of the actual flight process, various errors and even failures of the actuators may occur, resulting in abnormal recovery of the rocket during the recovery flight process and failure of the recovery. Therefore, it is necessary to study the fault-tolerant control considering various errors and failures of the actuators and evaluate the control effectiveness in order to reasonably deal with different abnormal situations.

[0004] In order to realize accurate control effectiveness evaluation, it is necessary to build an accurate control effectiveness evaluation system and establish a perfect evaluation process. The key to whether the effectiveness evaluation system is accurate lies in whether it can comprehensively and accurately depict various error models and failure models of the actuators. Digital twin technology is a dynamic simulation technology widely used today, which fully utilizes physical models, sensor updates, operation history and other data, integrates multi-disciplinary, multi-physical, multi-scale and multi-probability simulation processes, and completes mapping in virtual space, thereby reflecting the whole life cycle process of the corresponding entity equipment. The concept of digital twin originated from the field of industrial manufacturing. By virtually constructing a digital model of the product, simulation testing and verification are performed. During production and manufacturing, the operation of the equipment and the changes brought by parameter adjustment can be simulated, thereby improving the reliability and usability of the product and reducing the risk of product research and development and manufacturing. So far, digital twin technology has shown great potential and has been used in the design and manufacturing, flight simulation and fault diagnosis of aircraft, satellites and rockets.

[0005] A typical drawback of digital twins is that the twin model is built from a real physical model, representing a white-box simulation of the real model. This requires a precise understanding and characterization of all the physical properties of the model. When the real model itself contains complex uncertainties or currently incomprehensible characteristics (such as unknown error sources or unknown failure modes), digital twin technology cannot construct a virtual model that is consistent with the real model. Hybrid twins, building upon digital twins, compensate for these shortcomings by introducing real data to correct the model. However, the development of hybrid twins is still in its early stages, with relatively few publicly available related technologies.

[0006] Therefore, how to provide a system and method for evaluating the fault-tolerant control performance of reusable rockets using hybrid twin technology is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides a system and method for evaluating the control performance of reusable rockets. It utilizes hybrid twin technology to address the typical shortcomings of digital twin technology, thereby evaluating the fault-tolerant control performance of reusable rockets. First, a multi-field coupled flight environment model of the reusable rocket is built, and actual environmental data is continuously incorporated to correct the model. Then, component-level multi-physics coupled modeling is performed on the reusable rocket and its installed main engine, grid fins, and attitude control thrusters, with actual actuator data continuously incorporated to correct the model. Finally, a large number of different flight conditions are designed to evaluate the performance of the fault-tolerant control law of the reusable rocket.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A reusable rocket control performance evaluation system includes: a multi-granularity analysis module, a hybrid twin correction module, a simulation condition setting module, and a control performance evaluation module;

[0010] The multi-granularity analysis module is used to model the reusable rocket based on its geometric shape and component installation positions to obtain a rocket twin model. The rocket twin model is divided into three granularities: coarse-grained rocket model, medium-grained rocket model, and fine-grained rocket model. It is also used to construct a flight environment model, which is divided into two granularities: coarse-grained environment model and fine-grained environment model.

[0011] The hybrid twin correction module is used to obtain a residual by subtracting the real data from the simulation data of the rocket twin model, use the residual to train a deep neural network, obtain the prediction residual of the reusable rocket through the trained deep neural network, and correct the rocket twin model through the prediction residual to obtain a hybrid twin model.

[0012] The simulation condition setting module is configured to determine a simulation granularity, set simulation initial parameters, and configure a simulation condition.

[0013] The control effectiveness evaluation module is configured to set a weighted and formal control effectiveness evaluation index according to a task requirement, select a simulation condition to complete a recovery flight control simulation, and complete control effectiveness evaluation according to the control effectiveness evaluation index.

[0014] Preferably, for the coarse-grained rocket model:

[0015] Only rigid body dynamics of the whole rocket is considered, and no components and flexibility are considered, and a state equation of the rocket is expressed as:

[0016]

[0017] wherein x = [r, v, q, ω, m] is a state quantity of the rocket, including a three-axis position vector, a three-axis velocity vector and a mass of the rocket at a current time; u = [f a ,m a ,f t ,m t ,θ] is a control quantity, f a is a thruster thrust, m a is a moment generated by the thruster, f t is an engine thrust, m t is a moment generated by the engine, θ is a grid rudder rotation angle; and g is an interference received by the whole rocket.

[0018] Preferably, for the medium-grained rocket model:

[0019] Rigid body dynamics of the rocket and an ideal component model are considered, wherein the ideal component model includes an ideal thruster model, an ideal grid rudder model and an ideal engine model:

[0020] The ideal thruster model:

[0021] Let a thrust size and direction be constant, and a deviation be subject to a normal distribution; let a thruster installation position be r, and a nominal thrust size be F, and a thrust vector generated when the thruster is ignited is:

[0022] f a =r·F+f r

[0023] wherein f r is a thrust size deviation subject to a normal distribution;

[0024] A moment vector is:

[0025] m a =r×f a

[0026] The ideal grid fin model:

[0027] The ideal grid fin model is equivalent to a flat wing surface, and the normal aerodynamic force generated by deflection is taken as the control force f s ; the reference area S area of the grid fin is set a , the axial force coefficient C n , the normal force coefficient C p , and the pressure center coefficient C s do not change with the flight speed v of the rocket, but only change with the grid fin deflection angle θ; the generated control force is:

[0028] f area =f(S a ,C n ,C p ,θ,v)

[0029] The ideal engine model:

[0030] The ideal engine model is consistent with the ideal thruster model except for the thrust size and installation direction.

[0031] Preferably, for the fine-grained rocket model:

[0032] The rocket flexible dynamics and the actuator model accurate to the components are considered simultaneously, including: the rocket flexible dynamics model, the thruster fine-grained model, the engine fine-grained model, and the fine-grained grid fin model:

[0033] The rocket flexible dynamics model is used to describe the liquid sloshing of the internal oxidizer and fuel storage tanks; the dry mass of the rocket body is set as m rocket , and the liquid mass stored in the oxidizer and fuel storage tanks is set as m fuel , then the acceleration generated by the engine thrust F is:

[0034] a=F / (m rocket +m fuel )

[0035] The motion equation of the pendulum ball is:

[0036]

[0037] Wherein, the equivalent pendulum length is l k , the pendulum ball swing angular acceleration is γ k , the rocket body angular acceleration is the pendulum ball position is L k , the pendulum ball mass is m k , and the pendulum rod tension is F k ; the pendulum rod tension and the liquid sloshing disturbance force on the rocket body are action and reaction forces.

[0038] The thruster fine-grained model:

[0039] (a) Determine the average time delay from the ignition command to the thrust not being zero from the thruster experimental data as the thruster's opening time delay t on ;

[0040] (b) Estimate the thruster curve from 0 thrust to nominal thrust from the real combustion chamber combustion data as the thruster's nominal thrust opening curve;

[0041] (c) Determine the average time delay from the shutdown command to the thrust starting to decrease from the thruster experimental data as the thruster's closing time delay t off ;

[0042] (d) Estimate the thruster curve from nominal thrust to 0 thrust from the real combustion chamber combustion data as the thruster's nominal thrust closing curve;

[0043] (e) According to the thruster curves obtained from (a)-(d), input the current time to determine the actual output thrust of the thruster;

[0044] (f) Calculate the torque generated by the thrust according to the ideal thruster model;

[0045] The engine fine-grained model:

[0046] (1) Determine the average time delay from the ignition command to the thrust not being zero from the engine experimental data as the engine's opening time delay t on ;

[0047] (2) Estimate the thruster curve from 0 thrust to nominal thrust from the real combustion chamber combustion data as the engine's nominal thrust opening curve;

[0048] (4) Determine the average time delay from the shutdown command to the thrust being zero from the engine experimental data as the thruster's closing time delay t off ;

[0049] (4) Estimate the thruster curve from nominal thrust to 0 thrust from the real combustion chamber combustion data as the engine's nominal thrust closing curve;

[0050] (5) According to the thruster curves obtained from (1)-(4), input the current time to determine the actual output thrust of the engine;

[0051] (6) Calculate the torque generated by the thrust according to the ideal engine model;

[0052] The fine-grained grid rudder model:

[0053] The aerodynamic characteristics are determined by finite element simulation or wind tunnel experiment, and then modeling is performed.

[0054] Preferably, the coarse-grained environment model comprises a coarse-grained gravity model and the uniform atmosphere model:

[0055] The Earth is taken as a standard sphere, and the atmospheric density distribution is uniform, so that the gravity acceleration received by the rocket is the coarse-grained gravity model:

[0056] g=GM earth / r 2

[0057] Wherein, G is the gravitational constant, M earth is the mass of the Earth, and r is the rocket coordinate;

[0058] The uniform atmosphere model approximates the atmospheric density distribution with an exponential function:

[0059]

[0060] Wherein, ρ0 is the atmospheric density on the reference sphere r=r0, H is the density altitude, and r is the rocket coordinate;

[0061] Considering that the atmospheric density decreases with the increase of altitude, the density altitude is:

[0062] H=H0+μ(r-r0)

[0063] Wherein, μ is a constant parameter, and H is the altitude constant;

[0064] The fine-grained environment model comprises a fine-grained Earth gravity field model and a fine-grained atmosphere model:

[0065] The fine-grained Earth gravity field model directly adopts GEM series model, GRIM series model, JGM series model or EGM series model;

[0066] The fine-grained atmosphere model directly adopts the International Reference Atmosphere model, Jacchia series model or DTM series model.

[0067] Preferably, the hybrid twin correction module comprises a residual calculation unit, a deep neural network and a residual correction unit;

[0068] The residual calculation unit is configured to introduce real data to calculate an uncertainty residual, wherein the real data D real is derived from test data of components or real flight data of the rocket, and the real data is subtracted from the simulation data D digital of the rocket twin model to obtain a residual:

[0069] ΔD=D real -D digital

[0070] where, when a coarse-grained rocket model is adopted, the overall flight state X of the recoverable rocket at the next moment t and the flight state X at the previous moment t-1 , the self-control U and the environmental disturbance G are related:

[0071] X t = f(X t-1 , U, G, t)

[0072] For any moment t, the residual error between the flight state X simulated by the coarse-grained rocket model and the real flight state X is taken as the overall residual error of the recoverable rocket:

[0073]

[0074] When a medium-grained or fine-grained rocket model is adopted, for any moment t, the residual error between the simulation quantity C of the control force / torque output of an actuator and the real output quantity C of the actuator is taken as the residual error of the actuator model: digital real

[0075]

[0076] The deep neural network is a multi-layer deep neural network f, and gradient descent is used to complete the training;

[0077] The residual error correction unit is configured to, after the training converges, obtain a predicted residual error through the trained deep neural network; and add the predicted residual error to the simulation result of the rocket twin model to complete residual error correction.

[0078] The flight state X of the rocket at the next moment

[0079] X t = f(X t-1 , U t , G, t) + f φ (S)

[0080] where

[0081] S = [X t-1 , U t , G]

[0082] X t-1 is the flight state of the rocket at the previous moment, U t is the control instruction at the current moment, G is the environmental disturbance at the current moment, φ is a trainable hyperparameter in the neural network, S is the current flight state of the rocket, f φ (S) is the predicted residual error corresponding to S. ​​​​

[0083] Preferably, the MSE loss function is used as a training evaluation index during the training of the deep neural network.

[0084]

[0085] Wherein, N is the number of samples, the network input composed of environmental parameters, current control instructions, and multi-physical field parameters.

[0086] Preferably, the simulation working condition setting module includes a simulation granularity determination unit, a simulation initial parameter setting unit, and a simulation working condition configuration unit.

[0087] The simulation granularity determination unit is configured to determine the simulation granularity, wherein the simulation granularity includes coarse granularity, medium granularity, fine granularity, and mixed granularity.

[0088] Under coarse-grained simulation, only coarse-grained environmental models and coarse-grained rocket models are considered.

[0089] Under medium-grained simulation, only medium-grained environmental models and medium-grained rocket models are considered.

[0090] Under fine-grained simulation, fine-grained environmental models and fine-grained rocket models are considered.

[0091] Under mixed-grained simulation, different-grained environmental models and rocket models are considered according to actual requirements.

[0092] The simulation initial parameter setting unit is configured to set simulation initial parameters, which specifically include selecting simulation initial time and simulation step size.

[0093] The simulation working condition configuration unit is configured to configure simulation working conditions, which specifically include setting the landing point coordinates R s and the initial pose interval of the rocket flight according to task requirements, and obtaining N initial flight poses of the rocket through random sampling based on the initial pose interval of the rocket flight.

[0094] Preferably, the control effectiveness evaluation module includes a control effectiveness evaluation index setting unit, a simulation unit, and a control effectiveness evaluation unit.

[0095] The control effectiveness evaluation index setting unit is configured to set the control effectiveness evaluation index in the form of weighted sum according to task requirements, wherein the control effectiveness evaluation index includes landing accuracy, landing impact size, landing lateral velocity, and fuel consumption.

[0096] The landing accuracy is:

[0097] P a = ΔR = |R r -R s |

[0098] wherein R r is the coordinate of the rocket landing moment, R s is the landing field coordinate;

[0099] The landing impact size is:

[0100] P s = |V v |

[0101] wherein V v is the normal velocity component of the rocket at the landing moment relative to the landing field ground;

[0102] The landing lateral velocity is:

[0103] P v = |V h |

[0104] wherein V h is the lateral velocity component of the rocket at the landing moment relative to the landing field ground;

[0105] The fuel consumption is defined as

[0106] P m = |m i -m f |

[0107] wherein m i is the total mass of the rocket at the initial moment, and m f is the total mass of the rocket at the landing moment;

[0108] The control performance evaluation index is:

[0109] P = C1P a + C2P s + C3P v + C4P m

[0110] wherein the weighting coefficients C1, C2, C3, and C4 are determined according to the task requirements;

[0111] The simulation unit is configured to select a simulation working condition to complete the recovery flight control simulation.

[0112] The control performance evaluation unit is configured to complete the control performance evaluation according to the control performance evaluation index.

[0113] A control performance evaluation method for a recoverable rocket, comprising the following steps:

[0114] S1. Performing granularity configuration according to task requirements, including rocket model granularity configuration and simulation environment granularity configuration;

[0115] S2. Model components, dynamics and environment according to the granularity configuration result;

[0116] S3. Determine whether to perform hybrid twin correction according to task requirements and whether there is available real data; if hybrid twin correction is to be performed, proceed to S4, and if hybrid twin correction is not to be performed, proceed to S5;

[0117] S4. If hybrid twin correction is to be performed, calculate the residual using real data and simulation data, train a neural network using the residual, correct the model using the neural network, and proceed to S5;

[0118] S5: Configure working conditions according to task requirements;

[0119] S6: Select a working condition, complete a recovery flight control simulation, and perform control effectiveness evaluation;

[0120] S7: Determine whether to change to the next working condition; if so, return to S6, and if not, proceed to S8;

[0121] S8: Determine whether to adjust granularity; if so, return to S1, and if not, complete control effectiveness evaluation and end.

[0122] According to the technical solution described above, compared with the prior art, the present disclosure provides a recoverable rocket control effectiveness evaluation system and method, which is used to solve the design, analysis and verification of fault-tolerant control law of new recoverable launch vehicles, adopts a multi-granularity hybrid twin technology, establishes a multi-granularity multi-field coupling hybrid twin flight environment model, a rocket model and an actuator model, approximates the error and failure of real components, and the system and method can be applied to various fault-tolerant control laws of recoverable rockets and support evaluation of control effectiveness under various flight conditions. Compared with existing systems and processes, the system and process proposed by the present disclosure have the advantages of multi-granularity, multi-field coupling, virtual-real data fusion, etc., and can flexibly cover various recoverable rockets, various actuators and various flight conditions. BRIEF DESCRIPTION OF DRAWINGS

[0123] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creative labor.

[0124] Figure 1 A system structure schematic diagram of a recoverable rocket control effectiveness evaluation system provided by the present application;

[0125] Figure 2A schematic diagram of the structure of a reusable rocket provided in an embodiment of the present invention;

[0126] Figure 3 A schematic diagram of an equivalent pendulum model for liquid sloshing provided in an embodiment of the present invention;

[0127] Figure 4 A schematic diagram of the command curve and actual thrust curve of the thruster provided in an embodiment of the present invention;

[0128] Figure 5 This is a flowchart illustrating a method for evaluating the control effectiveness of a reusable rocket, as provided by the present invention. Detailed Implementation

[0129] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0130] This invention discloses a reusable rocket control performance evaluation system, such as... Figure 1 As shown, it includes: a multi-granularity analysis module, a hybrid twin correction module, a simulation condition setting module, and a control performance evaluation module;

[0131] The multi-granularity analysis module is used to model the reusable rocket based on its geometry and component installation locations to obtain a rocket twin model. The rocket twin model is divided into three granularities: coarse-grained rocket model, medium-grained rocket model, and fine-grained rocket model. It is also used to construct a flight environment model, which is divided into two granularities: coarse-grained environment model and fine-grained environment model.

[0132] The hybrid twin correction module is used to obtain residuals by subtracting real data from simulation data of the rocket twin model. The residuals are used to train a deep neural network. The trained deep neural network is used to obtain the prediction residuals of the reusable rocket. The rocket twin model is then corrected using the prediction residuals to obtain the hybrid twin model.

[0133] The simulation condition setting module is used to determine the simulation granularity, set the initial simulation parameters, and configure the simulation conditions.

[0134] The control performance evaluation module is used to set weighted sum-type control performance evaluation indicators according to mission requirements, select simulation conditions to complete the recovery flight control simulation, and complete the control performance evaluation based on the control performance evaluation indicators.

[0135] It should be noted that:

[0136] Reusable rockets are generally long and slender cylinders with a first-stage rocket engine mounted at the center of the bottom. Four grid fins are orthogonally mounted on the upper part of the rocket. Some reusable rockets also have pairs of attitude control thrusters additionally installed on the rocket body. One possible reusable rocket configuration is as follows: Figure 2 As shown.

[0137] The rocket's cylindrical geometric parameters (length, width, height, cylinder cross-sectional radius, centroid, fuel tank dimensions, etc.), mass characteristics (rocket dry mass, fuel mass, oxidizer mass), component installation information (installation location, installation attitude), and component nominal parameters (engine thrust, grid fin equivalent cross-sectional area, grid fin maximum rotational speed, etc.) are all set by the configuration file and can be flexibly adjusted according to the requirements of the mission conditions.

[0138] To further implement the above technical solution, for the coarse-grained rocket model:

[0139] Considering only the rigid body dynamics of the rocket as a whole, without considering its components and flexibility, the rocket's equation of state can be expressed as:

[0140]

[0141] Where x = [r, v, q, ω, m] represents the rocket's state variables, including the rocket's current three-axis position vector, three-axis velocity vector, and mass; u = [f a ,m a ,f t ,m t ,],f a For thruster thrust, m a The torque generated by the thruster, f t For engine thrust, m t θ represents the torque generated by the engine, θ represents the grid fin rotation angle, and g represents the overall disturbance received by the rocket.

[0142] To further implement the above technical solution, for the medium-grained rocket model:

[0143] Consider rocket rigid body dynamics and ideal component models, where the ideal component models include ideal thruster models, ideal grid fin models, and ideal engine models:

[0144] Ideal thruster model:

[0145] Assuming the magnitude and direction of the thrust are constant, and the deviation follows a normal distribution; and assuming the thruster's installation position is r and the nominal thrust magnitude is F, then the thrust vector generated during thruster ignition is:

[0146] f a =r·F+f r

[0147] where f r is the thrust magnitude deviation obeying normal distribution;

[0148] The moment vector is:

[0149] m a = r x f a

[0150] Ideal grid fin model:

[0151] The ideal grid fin model is equivalent to a flat wing surface, and the normal aerodynamic force generated by deflection is taken as the control force f s ; the reference area S area , the axial force coefficient C a , the normal force coefficient C n , and the pressure center coefficient C p of the grid fin are assumed to be constant with respect to the flight speed v of the rocket and only vary with the grid fin deflection angle θ; the generated control force is:

[0152] f s = f(S area , C a , C n , C p , θ, v)

[0153] Ideal engine model:

[0154] The ideal engine model is consistent with the ideal thruster model except for the thrust magnitude and installation orientation.

[0155] To further implement the above technical solutions, for the fine-grained rocket model:

[0156] The rocket flexible dynamics and the actuator model accurate to the components are considered simultaneously, including: the rocket flexible dynamics model, the thruster fine-grained model, the engine fine-grained model, and the fine-grained grid fin model:

[0157] The rocket flexible dynamics model is used to describe the liquid sloshing of the internal oxidizer and fuel tanks; the liquid sloshing can be equivalent to an equivalent pendulum model, as shown in Figure 3 . The equivalent approximation conditions include: the momentum of the equivalent model is approximately equal to the momentum of the liquid, the moment of momentum of the equivalent model relative to the total center of mass is approximately equal to the moment of momentum of the liquid relative to the total center of mass, and the dynamic equation of the equivalent model is approximately equal to the dynamic equation of the liquid sloshing.

[0158] When the recoverable rocket is descending in free fall, the rocket body is almost in free fall state. When the grid fins are opened, the rocket body will generate a reverse resistance, but the resistance will not cause too much liquid sloshing. Until the engine is ignited, the whole rocket body can be regarded as receiving an equivalent upward acceleration, and then the action of the grid fins and the attitude control thrusters will cause the liquid to slosh more obviously.

[0159] Let the dry mass of the rocket body be m rocket , and the liquid mass stored in the oxidizer and fuel tanks be m fuel , then the acceleration generated by the engine thrust F is:

[0160] a = F / (m rocket +m fuel )

[0161] The motion equation of the pendulum ball is:

[0162]

[0163] where the equivalent pendulum length is l k , the pendulum ball angular acceleration is γ k , the rocket body angular acceleration is , the pendulum ball position is L k , the pendulum ball mass is m k , and the pendulum rod tension is F k ; the pendulum rod tension and the liquid sloshing disturbance force on the rocket body are action and reaction forces;

[0164] The thruster is composed of an electromagnetic valve, a combustion chamber, and a nozzle. The on-off state of the electromagnetic valve is instantaneous switching, but there is a process for the oxidizer and fuel to flow out of the tank and burn in the combustion chamber, so there is also a process for the thruster thrust to be generated and disappeared. The fine-grained model of the thruster is introduced:

[0165] (a) According to the thruster experimental data, the average time delay from the ignition command to the thrust not being zero is determined as the opening time delay t on of the thruster;

[0166] (b) According to the real combustion chamber combustion data, the thruster curve from 0 thrust to the nominal thrust of the thruster is estimated as the nominal thrust opening curve of the thruster;

[0167] (c) According to the thruster experimental data, the average time delay from the closing command to the thrust beginning to decrease is determined as the closing time delay t off of the thruster;

[0168] (d) According to the real combustion chamber combustion data, the thruster curve from the nominal thrust to 0 thrust of the thruster is estimated as the nominal thrust closing curve of the thruster;

[0169] (e) According to the thrust curve obtained from (a)-(d), input the current time, determine the actual output thrust of the thruster;

[0170] (f) Calculate the torque generated by the thrust according to the ideal thruster model;

[0171] The final nominal thrust curve is shown in Figure 4 ;

[0172] Engine fine-grained model:

[0173] (1) According to the engine experimental data, determine the average time delay from the ignition command to the thrust not being zero as the engine opening time delay t on ;

[0174] (2) According to the real combustion chamber combustion data, estimate the thruster curve from 0 thrust to nominal thrust as the nominal thrust opening curve of the engine;

[0175] (5) According to the engine experimental data, determine the average time delay from the closing command to the thrust being zero as the thruster closing time delay t off ;

[0176] (4) According to the real combustion chamber combustion data, estimate the thruster curve from the nominal thrust to 0 thrust as the nominal thrust closing curve of the engine;

[0177] (5) According to the thrust curve obtained from (1)-(4), input the current time, determine the actual output thrust of the engine;

[0178] (6) Calculate the torque generated by the thrust according to the ideal engine model;

[0179] Fine-grained grid rudder model:

[0180] Determine the aerodynamic characteristics through finite element simulation or wind tunnel experiment, and then model.

[0181] In order to further implement the above technical solutions, the coarse-grained environment model includes a coarse-grained gravity model and a uniform atmosphere model:

[0182] Assuming that the Earth is a standard sphere and the atmospheric density is uniformly distributed, the gravitational acceleration received by the rocket is the coarse-grained gravity model:

[0183] g=GM earth / r 2

[0184] Where G is the gravitational constant, M earth is the mass of the Earth, and r is the rocket coordinate;

[0185] The uniform atmosphere model approximates the atmospheric density distribution with an exponential function:

[0186]

[0187] wherein ρ0 is the atmospheric density on the reference sphere r=r0, H is the density altitude, and r is the rocket coordinate;

[0188] Considering that the atmospheric density decreases with the increase of altitude, the density altitude is:

[0189] H=H0+μ(r-r0)

[0190] wherein μ is a constant parameter, and H is the altitude constant;

[0191] The fine-grained environment model includes a fine-grained earth gravity field model and a fine-grained atmospheric model:

[0192] The fine-grained earth gravity field model directly adopts a GEM series model, a GRIM series model, a JGM series model, or an EGM series model;

[0193] The fine-grained atmospheric model directly adopts an international reference atmospheric model, a Jacchia series model, or a DTM series model.

[0194] In order to further implement the above technical solutions, the hybrid twin correction module includes a residual calculation unit, a deep neural network, and a residual correction unit;

[0195] The residual calculation unit is configured to calculate an uncertainty residual by introducing real data, wherein the real data D real is derived from test data of components or real flight data of the rocket, the real data is compared with simulation data D digital of the rocket twin model, and a difference is obtained as a residual:

[0196] ΔD=D real -D digital

[0197] wherein when a coarse-grained rocket model is adopted, the overall flight state X t of the recoverable rocket at the next moment is related to the flight state X t-1 at the previous moment, self-control U, and environmental disturbance G:

[0198] X t =f(X t-1 ,U,G,t)

[0199] For any moment t, the residual between the simulation flight state X of the coarse-grained rocket model and the real flight state X is taken as the overall residual of the recoverable rocket:

[0200]

[0201] When the medium-fine-grained rocket model is adopted, for any time t, the simulation quantity C of the control force / torque output of the certain actuator digital The residual between the real output quantity telemetry value C real is taken as the actuator model residual:

[0202]

[0203] The deep neural network is a multi-layer deep neural network f, and the gradient descent is adopted to complete the training;

[0204] The residual correction unit is configured to, after the training converges, acquire the predicted residual through the trained deep neural network; and add the predicted residual to the simulation result of the rocket twin model to complete the residual correction.

[0205] The rocket flight state at the next time

[0206] X t = f(X t-1 , U t , G, t) + f φ (S)

[0207] Wherein

[0208] S = [X t-1 , U t , G]

[0209] X t-1 is the rocket flight state at the last time, U t is the control instruction at the current time, G is the environmental disturbance at the current time, φ is a trainable hyperparameter in the neural network, S is the current flight state of the rocket, f φ (S) is the predicted residual corresponding to S.

[0210] In order to further implement the above technical solutions, the MSE loss function is adopted as the training evaluation index in the training process of the deep neural network:

[0211]

[0212] Wherein, N is the sample number, the network input composed of the environmental parameters, the current control instruction and the multi-physical field parameters.

[0213] In order to further implement the above technical solutions, the simulation condition setting module includes a simulation granularity determination unit, a simulation initial parameter setting unit and a simulation condition configuration unit.

[0214] The simulation granularity determination unit is configured to determine the simulation granularity, wherein the simulation granularity includes coarse granularity, medium granularity, fine granularity and mixed granularity.

[0215] Under coarse-grained simulation, only coarse-grained environment model and coarse-grained rocket model are considered;

[0216] Under medium-grained simulation, only medium-grained environment model and medium-grained rocket model are considered;

[0217] Under fine-grained simulation, fine-grained environment model and fine-grained rocket model are considered;

[0218] Under hybrid-grained simulation, according to actual requirements, environment model and rocket model of different granularity are considered respectively;

[0219] The simulation initial parameter setting unit is used for setting simulation initial parameters, which specifically include: selecting simulation initial time and simulation step length;

[0220] The simulation working condition configuration unit is used for configuring simulation working conditions, which specifically include: setting landing point coordinates R s and initial pose interval of rocket flight according to task requirements; and obtaining N rocket initial flight poses through random sampling according to the initial pose interval of rocket flight.

[0221] It should be noted that:

[0222] In actual application process, under coarse-grained simulation, simulation step length is generally not less than 1 second;

[0223] Under medium-grained simulation, simulation step length is generally not less than 0.1 second;

[0224] Under fine-grained simulation, simulation step length is generally not less than 0.01 second;

[0225] Under hybrid-grained simulation, simulation step length is selected according to the simulation step length of the finest-grained model considered.

[0226] In order to further implement the above technical solutions, the control performance evaluation module includes: a control performance evaluation index setting unit, a simulation unit and a control performance evaluation unit;

[0227] The control performance evaluation index setting unit is used for setting weighted and formal control performance evaluation indexes according to task requirements, and the control performance evaluation indexes include landing accuracy, landing impact size, landing lateral speed and fuel consumption; wherein,

[0228] The landing accuracy is:

[0229] P a =ΔR=|R r -R s |

[0230] Wherein R r is the coordinate of the moment when the rocket lands, and R s is the landing field coordinate;

[0231] The landing impact size is:

[0232] P s = |V v |

[0233] wherein V v is the normal velocity component of the rocket at the landing moment relative to the landing site ground surface;

[0234] The landing lateral velocity is:

[0235] P v = |V h |

[0236] wherein V h is the lateral velocity component of the rocket at the landing moment relative to the landing site ground surface;

[0237] The fuel consumption is defined as

[0238] P m = |m i -m f |

[0239] wherein m i is the total mass of the rocket at the initial moment, and m f is the total mass of the rocket at the landing moment;

[0240] The control performance evaluation index is

[0241] P = C1P a + C2P s + C3P v + C4P m

[0242] wherein the weighting coefficients C1, C2, C3 and C4 are determined according to the task requirements;

[0243] The simulation unit is configured to select a simulation working condition to complete the recovery flight control simulation;

[0244] The control performance evaluation unit is configured to complete the control performance evaluation according to the control performance evaluation index.

[0245] A control performance evaluation method for a recoverable rocket, as shown in Figure 5 , comprises the following steps:

[0246] S1. Performing granularity configuration according to task requirements, including rocket model granularity configuration and simulation environment granularity configuration;

[0247] S2. Modeling the components, dynamics and environment according to the granularity configuration results;

[0248] S3. According to the task requirement and whether there is available real data, it is determined whether to carry out hybrid twin correction; if the hybrid twin correction is to be carried out, S4 is carried out, and if the hybrid twin correction is not to be carried out, S5 is carried out;

[0249] S4. If the hybrid twin correction is to be carried out, residual error is calculated using real data and simulation data, a neural network is trained using the residual error, the model is corrected using the neural network, and S5 is entered;

[0250] S5: According to the task requirement, a working condition is configured;

[0251] S6: One working condition is selected, recovery flight control simulation is completed, and control effectiveness evaluation is carried out;

[0252] S7: It is judged whether to replace the next working condition; if yes, S6 is returned; and if no, S8 is entered;

[0253] S8: It is judged whether to adjust the granularity; if yes, S1 is returned; and if no, control effectiveness evaluation is completed, and the method ends.

[0254] It should be noted that:

[0255] In S4, the residual error is the difference between the real data and the simulation data; when the first hybrid twin correction is carried out, the simulation data is obtained by pre-simulation; in the iterative process, the simulation data is obtained by simulation in S6.

[0256] The method can support various recoverable rocket control laws; in the method, the landing flight control law design of the recoverable rocket can also be included without hybrid twin correction.

[0257] The application provides a recoverable rocket control effectiveness evaluation system and process, which are used for solving the problems of incomplete and inflexible effectiveness evaluation in the landing flight control process of the recoverable rocket, and have the following advantages.

[0258] 1) The rocket and each actuator model are constructed in a multi-granularity manner, and different simulation granularities can be set according to needs. In the initial stage of rocket design and recovery scheme design, a coarse-grained model is used for rapid iteration; in the detailed design and fault-tolerant control law evaluation stage, a fine-grained model is used for detailed analysis and evaluation, so that the rocket recovery flight control effectiveness is comprehensively evaluated; according to the specific flight control task requirement, a model combination with different granularities can also be set to form a hybrid-granularity simulation working condition, and flexible rocket recovery flight control effectiveness evaluation can be carried out;

[0259] 2) The application adopts a hybrid twin technology to construct a virtual model. The physical model part that can be modeled is modeled in a digital twin manner; for the part that cannot be modeled and cannot be cognized, a neural network model is trained by introducing real data as samples for quantitative prediction. The above virtual model construction process can consider the existing real test or experimental data as much as possible on the basis of ensuring the correctness of the model physical characteristics and mechanism, and approximate the state and output characteristics of the real components and models to realize the error and failure conditions of the real components;

[0260] On the basis of constructing a multi-granularity hybrid twin model, the application introduces a physical field environment coupled by force, heat, light and electricity, can build a very realistic recyclable rocket flight control scene, and thus establishes a complete closed loop from sensor state measurement to rocket computer executing control law, to actuator actuating according to control command;

[0261] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0262] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A recoverable rocket control effectiveness evaluation system, characterized by, Comprise: Multi-granularity analysis module, hybrid twin correction module, simulation working condition setting module and control performance evaluation module; The multi-granularity analysis module is used for modeling the recoverable rocket according to the geometric shape and component installation position of the recoverable rocket to obtain a rocket twin model, the rocket twin model is divided into a coarse-grained rocket model, a medium-grained rocket model and a fine-grained rocket model according to granularity, the coarse-grained rocket model only considers rigid body dynamics of the whole rocket and does not consider components and flexibility, the medium-grained rocket model considers rigid body dynamics of the rocket and an ideal component model, wherein the ideal component model comprises an ideal thruster model, an ideal grid rudder model and an ideal engine model, and the fine-grained rocket model simultaneously considers rocket flexible dynamics and an actuator model accurate to components, comprising a rocket flexible dynamics model, a fine-grained thruster model, a fine-grained engine model and a fine-grained grid rudder model; And a flight environment model is constructed simultaneously, the flight environment model is divided into a coarse-grained environment model and a fine-grained environment model according to granularity, the coarse-grained environment model comprises a coarse-grained gravity model and a uniform atmosphere model, and the fine-grained environment model comprises a fine-grained earth gravity field model and a fine-grained atmosphere model, the fine-grained earth gravity field model directly adopts a GEM series model, a GRIM series model, a JGM series model or an EGM series model, and the fine-grained atmosphere model directly adopts an international reference atmosphere model, a Jacchia series model or a DTM series model; The hybrid twin correction module is used for introducing a residual error obtained by subtracting simulation data of the rocket twin model from real data, training a deep neural network using the residual error, obtaining a predicted residual error of the recoverable rocket through the trained deep neural network, and correcting the rocket twin model through the predicted residual error to obtain a hybrid twin model; The simulation working condition setting module is used for determining a simulation granularity, setting simulation initial parameters and configuring a simulation working condition, and the simulation working condition setting module comprises a simulation granularity determination unit, the simulation granularity determination unit is used for determining a simulation granularity, wherein the simulation granularity comprises a coarse-grained simulation, a medium-grained simulation, a fine-grained simulation and a hybrid-grained simulation; Under coarse-grained simulation, only a coarse-grained environment model and a coarse-grained rocket model are considered; Under medium-grained simulation, only a medium-grained environment model and a medium-grained rocket model are considered; Under fine-grained simulation, a fine-grained environment model and a fine-grained rocket model are considered; Under hybrid-grained simulation, different granularities of environment models and rocket models are considered according to actual requirements; The control performance evaluation module is used for setting a weighted and formal control performance evaluation index according to task requirements, selecting a simulation working condition to complete recoverable flight control simulation, and completing control performance evaluation according to the control performance evaluation index.

2. The recoverable rocket control effectiveness evaluation system according to claim 1, wherein The state equation of the rocket is represented as: wherein, is the state of the rocket, including the three-axis position vector, three-axis velocity vector and mass of the rocket at the current time instant; , is the thrust of the thruster, is the moment of the thruster, is the engine thrust, is the moment of the engine, The ideal thruster model: is the grid fin angle of attack; is the disturbance received by the rocket as a whole.

3. The recoverable rocket control effectiveness evaluation system according to claim 1, wherein The torque vector is: Let the thrust magnitude and direction be constant, and the deviation obey the normal distribution; let the thruster installation position be r , and the nominal thrust magnitude be F , then the thrust vector generated when the thruster is ignited is: wherein f r is a normally distributed thrust magnitude deviation; The ideal grid rudder model: The ideal engine model: The ideal grid rudder model is equivalent to a flat airfoil, and the normal aerodynamic force generated by deflection is used as the control force. ; Set the reference area of ​​the grid rudder Axial force coefficient Normal force coefficient and core pressure coefficient Not affected by rocket flight speed It changes with the grid rudder angle only. The torque vector is: Changes occur; the resulting control force is: The motion equation of the pendulum ball is: Let the thrust magnitude and direction be constant, and the deviation obey the normal distribution; let the engine installation position be r , and the nominal thrust magnitude be F , then the thrust vector generated when the engine ignites is: wherein f r is a normally distributed thrust magnitude deviation; The fine-grained thruster model: 。 4. The recoverable rocket control effectiveness evaluation system according to claim 1, wherein, The rocket flexible dynamics model is used to describe the liquid sloshing of the internal oxidizer and fuel tanks; the dry mass of the rocket body is m rocket the liquid mass stored in the oxidizer and fuel tanks is m fuel the engine thrust is F The generated acceleration is: ​ wherein the equivalent pendulum length is l k , the pendulum bob angular acceleration is ​ k , the arrow angular acceleration is , the pendulum bob position is L k , the pendulum bob mass is m k , the pendulum rod tension is F k ; the pendulum rod tension and the liquid sloshing disturbance force on the arrow are action and reaction forces. ​ (a) determining from thruster experimental data an average time delay from the firing command to the time when the thrust is not zero as the opening time delay of the thruster t on ; (b) estimating the thruster curve of the thruster from 0 thrust to nominal thrust according to the real combustion chamber combustion data as the nominal thrust opening curve of the thruster; (c) determining from thruster experimental data an average time delay from the issue of a shut down command to the start of thrust reduction as a shut down time delay of the thruster t off ; (d) estimating the thruster curve of the thruster from nominal thrust to 0 thrust according to the real combustion chamber combustion data as the nominal thrust closing curve of the thruster; (e) inputting the current time into the thruster curves obtained according to (a)-(d) to determine the actual output thrust of the thruster; (f) calculating the torque generated by the thrust according to the ideal thruster model; The engine fine-grained model: (1) The average time delay from the ignition command to the thrust not being zero is determined according to the engine experimental data as the opening time delay of the engine t on ; (2) estimating the thruster curve of the thruster from 0 thrust to nominal thrust according to the real combustion chamber combustion data as the nominal thrust opening curve of the engine; (3) The average time delay from the issuance of the shutdown command to the thrust being zero is determined from engine test data as the thruster's shutdown time delay t off ; (4) estimating the thruster curve of the thruster from nominal thrust to 0 thrust according to the real combustion chamber combustion data as the nominal thrust closing curve of the engine; (5) inputting the current time into the thruster curves obtained according to (1)-(4) to determine the actual output thrust of the engine; (6) calculating the torque generated by the thrust according to the ideal engine model; The fine-grained grid rudder model: Determine the aerodynamic characteristics through finite element simulation or wind tunnel experiment, and then model.

5. The recoverable rocket control effectiveness evaluation system according to claim 1, wherein Assuming that the earth is a standard sphere and the atmospheric density is uniform, the gravitational acceleration received by the rocket is the coarse-grained gravity model: wherein, G is the gravitational constant, M is the mass of the earth, R is the rocket coordinate; The uniform atmosphere model approximates the atmospheric density distribution with an exponential function: where is the atmospheric density on the reference sphere is the density altitude, is the density altitude, is the rocket coordinates; Considering that the atmospheric density decreases with the increase of altitude, the density altitude is: wherein is a constant parameter, H is an elevation constant.

6. The recoverable rocket control effectiveness evaluation system according to claim 1, wherein The hybrid twin correction module includes a residual calculation unit, a deep neural network, and a residual correction unit. The residual calculation unit is configured to introduce a true data calculation uncertainty residual, wherein the true data The true data is derived from test data of components or real flight data of the rocket, and the true data is subtracted from simulation data of a twin model of the rocket to obtain a residual: wherein, when a coarse-grained rocket model is employed, the overall flight state of the recoverable rocket at the next time instant is recoverable X t related to the flight state at the previous time instant X t-1 , the self-control U and the environmental disturbance G ​ For any time instant t the residual between the coarse-grained rocket model simulation flight state and the real flight state is the overall residual of the recoverable rocket: When using a medium- to fine-grained rocket model, for any given time... t Simulated quantity of control force / torque output of a certain actuator C digital Compared with its actual output telemetry value C real The residuals between them are used as the residuals of the actuator model: The deep neural network is a multi-layer deep neural network f and the training is completed in a gradient descent manner. The residual correction unit is configured to, after training convergence, obtain a predicted residual through the trained deep neural network, and accumulate the predicted residual to a simulation result of the rocket twin model to complete residual correction. Next time rocket flight state wherein is the rocket flight state at the previous time instant, is the control command at the current time instant, is the environmental disturbance at the current time instant, φ is the trainable hyperparameter in the neural network, S is the current flight state of the rocket, f φ S is the prediction residual corresponding to S.​ 7. The recoverable rocket control effectiveness evaluation system according to claim 6, wherein In the training process of the deep neural network, the MSE loss function is used as the training evaluation index: wherein, N is the number of samples, environmental parameters, current control instructions, and multi-physical field parameters constitute a network input.

8. The recoverable rocket control effectiveness evaluation system according to claim 1, wherein The simulation working condition setting module includes a simulation initial parameter setting unit and a simulation working condition configuration unit. The simulation initial parameter setting unit is configured to set simulation initial parameters, specifically including: selecting simulation initial time and simulation step length. The simulation working condition configuration unit is configured to configure a simulation working condition, and specifically includes: setting a landing point coordinate according to a task requirement and an initial pose interval of the rocket flight; and obtaining N initial flight poses of the rocket according to the initial pose interval of the rocket flight through a random sampling manner.

9. The recoverable rocket control effectiveness evaluation system according to claim 1, wherein The control effectiveness evaluation module includes a control effectiveness evaluation index setting unit, a simulation unit, and a control effectiveness evaluation unit. The control effectiveness evaluation index setting unit is configured to set the control effectiveness evaluation index in the form of weighting according to task requirements, and the control effectiveness evaluation index includes landing accuracy, landing impact size, landing lateral velocity, and fuel consumption; wherein The landing accuracy is: wherein is the coordinate of the moment of rocket landing, is the landing field coordinate; The landing impact size is: wherein Vn is the normal velocity component of the rocket at the moment of landing relative to the landing site surface; The landing lateral velocity is: wherein Vtis the lateral velocity component of the rocket at the moment of landing relative to the landing site surface; The fuel consumption is defined as wherein M0is the total mass of the rocket at the initial time, M1is the total mass of the rocket at the landing time; The control effectiveness evaluation index is where the weighting coefficients determined according to the task requirements; The simulation unit is configured to select a simulation working condition to complete recovery flight control simulation. The control effectiveness evaluation unit is configured to complete control effectiveness evaluation according to the control effectiveness evaluation index. 10.A recoverable rocket control effectiveness evaluation method, applied to the recoverable rocket control effectiveness evaluation system according to any one of claims 1-9, characterized in that, The method comprises the following steps: S1. Configure the granularity according to the task requirements, including rocket model granularity configuration and simulation environment granularity configuration; S2. Model components, dynamics and environment according to the granularity configuration result; S3. Determine whether to perform hybrid twin correction according to task requirements and whether there is available real data, if yes, go to S4, if not, go to S5; S4. If hybrid twin correction is to be performed, calculate the residual error using real data and simulation data, train the neural network using the residual error, correct the model using the neural network, and go to S5; S5. Configure the working condition according to the task requirements; S6. Select a working condition and complete the recovery flight control simulation to evaluate the control effectiveness; S7. Determine whether to change to the next working condition, if yes, return to S6, if not, go to S8; S8. Determine whether to adjust the granularity, if yes, return to S1, otherwise, complete the control effectiveness evaluation and end.

Images