Aircraft aerodynamics simulation modeling method, system, equipment and storage medium

Abstract

The present invention belongs to the technical field of aerodynamics, and specifically relates to a method, system, device and storage medium for aircraft aerodynamic simulation modeling, aiming to solve the problem that the existing technology lacks effective means for aerodynamic simulation modeling of flight training simulation equipment. The method proposed by the present invention includes: obtaining test flight data of a target aircraft; determining target data based on the test flight data; training a pre-built non-fully connected neural network based on the target data to obtain a relationship mapping model; constructing an aerodynamic model interpolation table using the relationship mapping model; and performing aerodynamic simulation modeling of the target aircraft based on the aerodynamic model interpolation table. This method constructs an aerodynamic model interpolation table based on the test flight data, and can then perform aerodynamic modeling of the target aircraft based on this, effectively solving the problem of no models available during the development of flight simulation training equipment.

Description

Technical Field

[0001] The present invention belongs to the technical field of aerodynamics, and in particular relates to an aircraft aerodynamics simulation modeling method, system, equipment and storage medium. Background Art

[0002] A flight simulator training device (FSTD) uses high-precision simulation technology to replicate realistic flight environments and aircraft behavior. It is used for pilot training, skill assessment, and emergency response drills. Its core goal is to simulate complex flight scenarios in a safe and controllable environment, improving pilots' operational capabilities and decision-making abilities. Therefore, developing high-quality aerodynamic simulation modeling for FSTDs is crucial for their analysis and improvement.

[0003] Currently, aerodynamic modeling during aircraft design and manufacturing is primarily achieved through a combination of wind tunnel test data and CFD finite element simulation, which can achieve good results. However, these methods are designed for aircraft that are already airworthy and are difficult to apply directly to flight training simulators. Specifically, these methods require accurate mechanical modeling data, which is not available from aircraft manufacturers for flight training simulators. Furthermore, wind tunnel testing is expensive, and CFD simulation analysis requires extensive experience and is difficult to implement. Furthermore, even if these deficiencies are overcome, the resulting model is still not accurate enough for direct application in flight simulation training equipment, requiring further optimization of the flight dynamics model using test flight data. Summary of the Invention

[0004] In order to solve the above-mentioned problem in the prior art, namely, the problem that the prior art lacks an effective means for aerodynamic simulation modeling of flight training simulation equipment, the present invention provides, in a first aspect, an aircraft aerodynamic simulation modeling method, the method comprising:

[0005] Obtain flight test data of the target aircraft;

[0006] Determining target data based on the test flight data, the target data including six-degree-of-freedom force / torque coefficients acting on the target aircraft and relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients, the relationship parameters being used to characterize factors affecting the six-degree-of-freedom force / torque coefficients;

[0007] Based on the target data, a pre-constructed non-fully connected neural network is trained to obtain a relationship mapping model, wherein the non-fully connected neural network includes a plurality of independent neuron groups, each neuron group is used to independently map the correspondence between at least one six-degree-of-freedom force / torque coefficient and a relationship parameter, and the non-fully connected neural network is trained using a pre-constructed hybrid loss function, wherein the hybrid loss function includes a data fitting term and a physical verification term, the data fitting term is used to minimize the deviation between the six-degree-of-freedom force / torque coefficient and the target data, and the physical verification term is used to verify the six-degree-of-freedom force / torque coefficient according to preset aerodynamic rules;

[0008] Using the relationship mapping model, constructing an aerodynamic model interpolation table, wherein the aerodynamic model interpolation table is used to indicate at least one six-degree-of-freedom force / torque coefficient corresponding to each relationship parameter;

[0009] According to the aerodynamic model interpolation table, aerodynamic simulation modeling is performed on the target aircraft, and the relationship mapping model is updated according to the modeling result.

[0010] In some preferred embodiments, determining target data based on the test flight data includes:

[0011] detecting the flight test data and determining missing values ​​and abnormal values ​​in the flight test data;

[0012] Cleaning the missing values ​​and outliers to obtain cleaned flight test data;

[0013] Performing inverse calculation on the cleaned test flight data to determine the six-degree-of-freedom forces / torques acting on the target aircraft;

[0014] The six-degree-of-freedom force / torque is dimensionally non-quantized to obtain six-degree-of-freedom force / torque coefficients acting on the target aircraft.

[0015] In some preferred embodiments, determining the target data further includes:

[0016] Acquiring configuration data of the target aircraft;

[0017] determining configuration parameters of the target aircraft according to the configuration data, wherein the configuration parameters are data in the configuration data related to the flight state of the target aircraft;

[0018] determining, based on the configuration parameters, influencing factors of six-degree-of-freedom forces / torques on the target aircraft;

[0019] The influencing factors are dimensionally processed to determine the relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients.

[0020] In some preferred embodiments, the dimensionless coefficients of the forces and moments of the six degrees of freedom of the target aircraft include a lift coefficient, a drag coefficient, a side force coefficient, a pitch moment coefficient, a rolling moment coefficient, and a yaw moment coefficient;

[0021] The relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients include the angle of attack, flap configuration, landing gear configuration, engine thrust coefficient, angle of attack change rate, and pitch angle change rate of the target aircraft.

[0022] In some preferred embodiments, the lift coefficient includes a static stability coefficient, a first dynamic stability coefficient, and a second dynamic stability coefficient of the target aircraft, wherein the first dynamic stability coefficient is affected by the angle of attack of the target aircraft, and the second dynamic stability coefficient is affected by the pitch angular velocity of the target aircraft.

[0023] In some preferred embodiments, constructing an aerodynamic model interpolation table using the relationship mapping model includes:

[0024] Based on aerodynamic principles, the relationship parameters are divided into a plurality of parameter groups, each parameter group corresponds to at least one neuron group in the relationship mapping model, and each parameter group generates at least one of the six-degree-of-freedom force / torque coefficients through the neuron group;

[0025] Determining a target parameter group and a target neuron group, wherein the target parameter group is any one of a plurality of parameter groups, and the target neuron group is a neuron group corresponding to the target parameter group;

[0026] Determining discrete target interpolation points according to the target parameter group, and using the target interpolation points as inputs to the target neuron group;

[0027] Determine the output result of the N-1th layer of neurons in the target neuron group, where N is the total number of layers of the target neuron group;

[0028] According to the output results corresponding to each target parameter group, the aerodynamic model interpolation table is determined.

[0029] In some preferred embodiments, the hybrid loss function satisfies:

[0030] ;

[0031] Where N is the number of training samples, is the prediction coefficient vector of the i-th sample, is the true value coefficient vector of the i-th sample, is the strength factor of the aerodynamic rule constraint, K is the number of aerodynamic rule constraints, is the weight of the K-th physical rule, is the kth physical constraint function.

[0032] In a second aspect, the present invention further provides an aircraft aerodynamic simulation modeling system, comprising:

[0033] A data acquisition module, used to acquire the flight test data of the target aircraft;

[0034] a parameter calculation module, configured to determine target data based on the test flight data, the target data including six-degree-of-freedom force / torque coefficients acting on the target aircraft and relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients, the relationship parameters being used to characterize factors influencing the six-degree-of-freedom force / torque coefficients;

[0035] a model training module, configured to train a pre-constructed non-fully connected neural network based on the target data to obtain a relationship mapping model, wherein the non-fully connected neural network includes a plurality of independent neuron groups, each neuron group being configured to independently map a correspondence between at least one six-degree-of-freedom force / torque coefficient and a relationship parameter, and wherein the non-fully connected neural network is trained using a pre-constructed hybrid loss function, wherein the hybrid loss function includes a data fitting term and a physical verification term, wherein the data fitting term is configured to minimize a deviation between the six-degree-of-freedom force / torque coefficient and the target data, and the physical verification term is configured to verify the six-degree-of-freedom force / torque coefficient according to preset aerodynamic rules;

[0036] an interpolation construction module, configured to construct an aerodynamic model interpolation table using the relationship mapping model, wherein the aerodynamic model interpolation table is configured to indicate at least one six-degree-of-freedom force / torque coefficient corresponding to each discrete interpolation point;

[0037] The simulation modeling module is used to perform aerodynamic simulation modeling on the target aircraft according to the aerodynamic model interpolation table, and update the relationship mapping model according to the modeling results.

[0038] In a third aspect, the present invention further provides an electronic device, comprising:

[0039] at least one processor; and

[0040] a memory communicatively connected to at least one of the processors; wherein,

[0041] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the method according to the first aspect.

[0042] In a fourth aspect, the present invention further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the method as described in the first aspect.

[0043] Beneficial effects of the present invention:

[0044] (1) The present invention proposes a method for independently performing aerodynamic modeling based on test flight data when the aircraft manufacturer cannot provide an aircraft dynamics model. This method solves the current situation in which no model is available during the development of flight simulation training equipment and fills a gap in the industry.

[0045] (2) In view of the characteristics of the dynamic model of flight simulation training equipment, the present invention modifies the fully connected neural network algorithm commonly used in this field and replaces it with a non-fully connected neural network algorithm, thereby avoiding the situation where an extremely large interpolation table is obtained when the discrete influencing factor interpolation points are introduced into the six-degree-of-freedom influencing factor and six-degree-of-freedom force / torque coefficient model. At the same time, it can also effectively avoid the coupling of irrelevant factors, further improving the accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0047] Figure 1 1 is a flow chart of an aircraft aerodynamic simulation modeling method proposed in an embodiment of the present invention;

[0048] Figure 2 is a schematic diagram of a fully connected neural network proposed in an embodiment of the present invention;

[0049] Figure 3 is a schematic diagram of a non-fully connected neural network proposed in an embodiment of the present invention;

[0050] Figure 4 Schematic diagram of constructing an interpolation table of key parameters of an aerodynamic model proposed in an embodiment of the present invention;

[0051] Figure 5 It is a structural diagram of a computer system proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The present application will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the relevant invention are shown in the accompanying drawings.

[0053] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0054] Please refer to Figure 1 The first embodiment of the present application provides a method for extracting features of faint celestial targets based on deep learning, the method comprising:

[0055] Step S10, obtaining the flight test data of the target aircraft;

[0056] Step S20: determining target data based on the test flight data, wherein the target data includes six-degree-of-freedom force / torque coefficients acting on the target aircraft and relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients, wherein the relationship parameters are used to characterize factors affecting the six-degree-of-freedom force / torque coefficients.

[0057] In this embodiment, the target aircraft actually refers to a test flight aircraft or flight simulation training equipment used for training, which is generally used in a comprehensive training system integrating modules such as human-computer interaction and environmental simulation to conduct operational training and assessment of pilots.

[0058] Taking flight simulation training equipment as an example, its test flight data specifically refers to the time series set of the above-mentioned relevant aircraft actual motion state, control input and environmental parameters, which can characterize the characteristics of the flight simulation training equipment during simulated flight.

[0059] Those skilled in the art will understand that the six degrees of freedom refer to the three translation directions of the target aircraft (respectively: X, Y, and Z axes), and the three rotation directions (respectively: roll, pitch, and yaw). Based on this, the forces and moments of the six degrees of freedom of the target aircraft include but are not limited to the drag, side force, lift of the target aircraft, and the corresponding rolling moment, pitching moment, yaw moment, etc.

[0060] The six-degree-of-freedom force / torque coefficients of this embodiment refer to dimensionless coefficients based on the six-degree-of-freedom force / torque, that is, parameters obtained by standardizing physical dimensions, and are used to describe the mathematical relationship between aerodynamic forces, torques, and flight conditions. Their dimension is 1 (i.e., they are considered to be unitless).

[0061] In this embodiment, by performing dimensionless processing, the unit dependence of physical quantities can be effectively eliminated, so that the aerodynamic characteristics under different sizes, speeds, and environments can be directly compared and universally modeled. At the same time, it is beneficial to reduce the dimensional differences of input data in subsequent steps and improve the convergence speed of the target model.

[0062] Those skilled in the art will understand that the configuration of the target aircraft (such as aerodynamic shape, mass distribution, control surface configuration, etc.) directly determines the influencing factors of its six-degree-of-freedom forces / torques and their mathematical expressions. For example, the wing configuration of the target aircraft will affect its lift and rolling moment, the tail configuration of the target aircraft will affect its pitching moment and yaw moment, the fuselage configuration of the target aircraft will affect its drag and side force, and the winglets of the target aircraft will affect its dynamic lift and pitch damping. Based on this understanding, the above-mentioned relationship parameters can be obtained by sorting the related configuration influencing factors according to the dimensionless coefficients of each force and moment according to category.

[0063] Step S30: Using the target data, a pre-constructed non-fully connected neural network is trained to obtain a relationship mapping model, wherein the non-fully connected neural network includes multiple independent neuron groups, each neuron group is used to independently map the correspondence between at least one six-degree-of-freedom force / torque coefficient and a relationship parameter, and the non-fully connected neural network is trained using a pre-constructed hybrid loss function, wherein the hybrid loss function includes a data fitting term and a physical verification term, wherein the data fitting term is used to minimize the deviation between the six-degree-of-freedom force / torque coefficient and the target data, and the physical verification term is used to verify the six-degree-of-freedom force / torque coefficient according to preset aerodynamic rules;

[0064] It is understood by those skilled in the art that common neural network algorithm models are generally fully connected, such as Figure 2 As shown in the figure, the neurons in the input layer, intermediate layer, and output layer are all connected, and the neurons in the intermediate layer have no physical meaning. Fully connected neural network models are more suitable for situations where the relationship between various parameters is unclear. However, for flight dynamics models, since the relationship between various parameters is very clear and the neurons in the intermediate layer have specific physical meanings, a non-fully connected neural network algorithm is required for flight dynamics simulation.

[0065] At the same time, if a fully connected neural network is used as the model framework, it is easy to obtain an extremely large interpolation table when the discrete influencing factor interpolation points are imported into the target model in the subsequent steps. However, this embodiment uses a non-fully connected neural network as the model framework. While avoiding obtaining an extremely large interpolation table, it can also effectively avoid the coupling of irrelevant factors and improve the accuracy of the model.

[0066] In this embodiment, decoupling is performed during modeling using a non-fully connected neural network. This reduces the degree of model coupling, enabling higher-precision simulations. It also avoids the problem of excessively dimensional interpolation tables being generated when obtaining the aerodynamic model interpolation table based on the model after modeling, which can be inconvenient for debugging. This approach is based on the knowledge of the influencing factors for each component, specifically the factors affecting the static stability coefficient, the dynamic stability coefficient due to the rate of attack angle change, and the dynamic stability coefficient due to the pitch angular velocity.

[0067] Furthermore, the hybrid loss function satisfies:

[0068] ;

[0069] Where N is the number of training samples, is the prediction coefficient vector of the i-th sample, is the true value coefficient vector of the i-th sample, is the strength factor of the aerodynamic rule constraint, K is the number of aerodynamic rule constraints, is the weight of the K-th physical rule, is the kth physical constraint function.

[0070] Step S40: determining discrete interpolation points based on the relationship parameters, and importing the discrete interpolation points into the relationship mapping model to determine an aerodynamic model interpolation table, wherein the aerodynamic model interpolation table is used to indicate at least one six-degree-of-freedom force / torque coefficient corresponding to each discrete interpolation point;

[0071] Step S50 , performing aerodynamic simulation modeling on the target aircraft according to the aerodynamic model interpolation table, and updating the relationship mapping model according to the modeling result.

[0072] It's easy to understand that an interpolation table is a tool used to estimate the value of unknown data points between known data points, allowing for the estimation of continuous data. Specifically, a continuous function is interpolated based on given discrete data, so that this continuous curve passes through all given discrete data points. This curve can then be used to estimate the value of each continuous point.

[0073] In this embodiment, the target model is used to process the mapping relationship between various dimensionless parameters and their corresponding relational parameters, so that the values ​​of all dimensionless coefficients can be approximately obtained based on the discrete relational parameters, that is, the one-to-one mapping relationship between dimensionless parameters and relational parameters is determined, and then based on this, the forces and moments of the aircraft can be calculated in real time during the simulation process, the six-degree-of-freedom dynamic simulation of the target aircraft can be completed, and finally the aerodynamic simulation modeling of the target aircraft can be completed.

[0074] Furthermore, in this embodiment, obtaining target data includes:

[0075] The invention also provides a method for obtaining test flight data of the target aircraft; preprocessing the test flight data to remove missing values ​​and outliers in the test flight data; performing inverse calculation on the preprocessed test flight data to obtain forces and moments acting on the six degrees of freedom of the target aircraft; and performing dimensionless processing on the six-degree-of-freedom forces / moments to obtain dimensionless coefficients of the six-degree-of-freedom forces and moments of the target aircraft.

[0076] In this embodiment, the flight test data of the target aircraft can be determined by the flight test plan for the target aircraft. The purpose of preprocessing the flight test data is to ensure data accuracy and meet the requirements of model training in subsequent steps. Outliers in the flight test data can be cleaned based on the 3σ criterion, that is, data that deviates by three standard deviations from the mean is eliminated. Furthermore, in this embodiment, the inverse solution of the flight test data can be performed based on the Newton-Euler equation. This embodiment does not limit the specific process of the inverse solution and dimensionless processing; those skilled in the art may select any feasible method.

[0077] Furthermore, in this embodiment, the acquiring of target data further includes: determining the configuration of the target aircraft; determining, based on the configuration of the target aircraft, influencing factors of the forces and moments of the six degrees of freedom of the target aircraft; and determining, based on each influencing factor, relationship parameters corresponding to the dimensionless coefficients.

[0078] It should be noted that the configuration of the target aircraft is defined by key elements such as its aerodynamic shape, mass distribution, and control surface configuration, such as its wings, tail, fuselage, control surfaces, propulsion system, mass distribution, etc. Based on its configuration, the mathematical expression of each influencing factor can be further determined to obtain the above-mentioned relationship parameters.

[0079] Furthermore, the dimensionless coefficients of the forces and moments of the six degrees of freedom of the target aircraft include lift coefficient, drag coefficient, side force coefficient, pitch moment coefficient, roll moment coefficient and yaw moment coefficient;

[0080] The relationship parameters corresponding to the dimensionless coefficients include the angle of attack, flap configuration, landing gear configuration, engine thrust coefficient, angle of attack change rate, and pitch angle change rate of the target aircraft.

[0081] Furthermore, the lift coefficient includes a static stability coefficient, a first dynamic stability coefficient, and a second dynamic stability coefficient of the target aircraft, wherein the first dynamic stability coefficient is affected by the angle of attack of the target aircraft, and the second dynamic stability coefficient is affected by the pitch angular velocity of the target aircraft.

[0082] For more details, please refer to Figure 3 , Figure 3 A schematic diagram of a non-fully connected neural network is shown, where the input represents dimensionless six-degree-of-freedom force / torque influencing factors, including the aircraft's angle of attack, flap configuration, landing gear configuration, engine thrust coefficient, angle of attack rate of change, and pitch angle rate of change. The output represents the aircraft's six-degree-of-freedom force / torque coefficients, which include the lift coefficient, drag coefficient, side force coefficient, pitch moment coefficient, rolling moment coefficient, and yaw moment coefficient. Each modeling operation specifically targets one of the six-degree-of-freedom force / torque coefficients. For ease of understanding, this embodiment focuses on the lift coefficient. Accordingly, the output here specifically represents the aircraft's lift coefficient.

[0083] Figure 3 The neural network consists of three components: the static stability coefficient, the dynamic stability coefficient due to angle of attack, and the dynamic stability coefficient due to pitch angular velocity. The sum of these three parameters gives the aircraft's lift coefficient. Its inputs are also divided into three parts: the static stability coefficient inputs include factors affecting the static stability coefficient, such as the aircraft's angle of attack, flap configuration, landing gear configuration, and engine thrust coefficient; the dynamic stability coefficient due to the rate of attack includes factors affecting the dynamic stability coefficient due to the rate of attack, flap configuration, and other factors; and the dynamic stability coefficient due to pitch angular velocity includes factors affecting the dynamic stability coefficient due to pitch angular velocity, flap configuration, and other factors. The activation function of the intermediate neural network algorithm uses the tan function, and the convergence function uses the mse function. Once the operation converges, the dynamic model of the aircraft's lift coefficient is obtained. It should be noted that this model is a black box and cannot be directly used in the development of flight simulation training equipment.

[0084] Furthermore, the step of determining discrete interpolation points based on the relationship parameters and importing the discrete interpolation points into the relationship mapping model to determine an aerodynamic model interpolation table includes:

[0085] Based on the principles of aerodynamics, the relationship parameters are divided into multiple parameter groups, each parameter group corresponds to at least one neuron group in the relationship mapping model, and each parameter group generates at least one of the six-degree-of-freedom force / torque coefficients through the neuron group; a target parameter group and a target neuron group are determined, the target parameter group is any one of the multiple parameter groups, and the target neuron group is the neuron group corresponding to the target parameter group; according to the target parameter group, discrete target interpolation points are determined, and the target interpolation points are used as inputs of the target neuron group; the output results of the N-1th layer of neurons in the target neuron group are determined, where N is the total number of layers of the target neuron group; and an aerodynamic model interpolation table is determined according to the output results corresponding to each target parameter group.

[0086] The following describes the above process using an aircraft whose control surfaces include elevators, ailerons, rudders, flaps, and whose landing gear cannot be retracted as an example:

[0087] Specifically, a non-fully connected neural network algorithm is used to split the relationship parameters (influencing parameters) according to known theoretical basis (aerodynamic principles). The elevator, flaps, angle of attack, and engine thrust coefficient are used as the relationship parameters of the first part (static stability coefficient); the flaps, elevator, and angle of attack change rate are used as the relationship parameters of the second part (dynamic stability coefficient caused by angle of attack change rate); the flaps, elevator, and pitch angular velocity are used as the relationship parameters of the third part (dynamic stability coefficient caused by pitch angular velocity). These three parts do not cross in the previous layers. In the penultimate layer, the three parts converge to one neuron output. In the last layer, all parameter values ​​of the penultimate layer are added together to obtain the final output. After training, three interpolation tables are generated: a static stability coefficient table with four variables (elevator, flaps, angle of attack, and engine thrust coefficient) as input and one variable as output; a dynamic stability coefficient table caused by the angle of attack rate with three variables (flaps, elevator, and angle of attack rate) as input and one variable as output; and a dynamic stability coefficient table caused by the pitch rate with three variables (flaps, elevator, and pitch rate) as input and one variable as output. This reduces the coupling of the model and facilitates debugging.

[0088] For details, please refer to Figure 4 , Figure 4 The process of importing the discrete six-degree-of-freedom influencing factor interpolation points into the above-mentioned target model to obtain the interpolation table of the key parameters of the aerodynamic model is shown. Taking the above-mentioned lift coefficient as an example, the discrete angle of attack values, discrete flap configurations, discrete landing gear configurations, and discrete engine thrust coefficients are input into the aircraft lift coefficient dynamics model black box according to all permutations and combinations, and the static stability coefficient in the second-to-last layer of the black box model is selected as the output to obtain the interpolation table of the aircraft static stability coefficient; the discrete flap configurations and discrete angle of attack change rates are input into the aircraft lift coefficient dynamics model black box according to all permutations and combinations, and the dynamic stability coefficient caused by the angle of attack change rate in the second-to-last layer of the black box model is selected as the output to obtain the interpolation table of the dynamic stability coefficient caused by the angle of attack change rate; the discrete flap configurations and discrete pitch angular velocity change rates are input into the aircraft lift coefficient dynamics model black box according to all permutations and combinations, and the dynamic stability coefficient caused by the pitch angular velocity in the second-to-last layer of the black box model is selected as the output to obtain the interpolation table of the dynamic stability coefficient caused by the pitch angular velocity. Finally, after processing various relationship parameters, a discretized dynamic model interpolation table of lift coefficient is obtained, which can be directly applied to the development of flight simulation training equipment.

[0089] A second embodiment of the present application provides an aircraft aerodynamics simulation modeling system, comprising:

[0090] A data acquisition module, used to acquire the flight test data of the target aircraft;

[0091] a parameter calculation module, configured to determine target data based on the test flight data, the target data including six-degree-of-freedom force / torque coefficients acting on the target aircraft and relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients, the relationship parameters being used to characterize factors influencing the six-degree-of-freedom force / torque coefficients;

[0092] a model training module, configured to train a pre-constructed non-fully connected neural network based on the target data to obtain a relationship mapping model, wherein the non-fully connected neural network includes a plurality of independent neuron groups, each neuron group being configured to independently map a correspondence between at least one six-degree-of-freedom force / torque coefficient and a relationship parameter, and wherein the non-fully connected neural network is trained using a pre-constructed hybrid loss function, wherein the hybrid loss function includes a data fitting term and a physical verification term, wherein the data fitting term is configured to minimize a deviation between the six-degree-of-freedom force / torque coefficient and the target data, and the physical verification term is configured to verify the six-degree-of-freedom force / torque coefficient according to preset aerodynamic rules;

[0093] an interpolation construction module, configured to construct an aerodynamic model interpolation table using the relationship mapping model, wherein the aerodynamic model interpolation table is configured to indicate at least one six-degree-of-freedom force / torque coefficient corresponding to each relationship parameter point;

[0094] The simulation modeling module is used to perform aerodynamic simulation modeling on the target aircraft according to the aerodynamic model interpolation table, and update the relationship mapping model according to the modeling results.

[0095] The third embodiment of the present application further provides an electronic device, including:

[0096] At least one processor; and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to implement the method as described in the first embodiment.

[0097] The fourth embodiment of the present application further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the method described in the first embodiment.

[0098] Reference below Figure 5 , which shows a structural diagram of a server computer system suitable for implementing the method, system, and device embodiments of the present application. Figure 5The server shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0099] like Figure 5 As shown, the computer system includes a central processing unit 301, which can perform various appropriate actions and processes according to programs stored in a read-only memory 302 or programs loaded from a storage unit 308 into a random access memory 303. Various programs and data required for system operation are also stored in the random access memory 303. The central processing unit 301, the read-only memory 302, and the random access memory 303 are connected to each other via a bus 304. An I / O interface 305 is also connected to the bus 304.

[0100] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, and the like; an output section 307 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 308 including a hard disk and the like; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. Removable media 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like, is installed in the drive 310 as needed so that computer programs read therefrom can be installed into the storage section 308 as needed.

[0101] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 309, and installed with or from the removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, the above-mentioned functions defined in the method of the present application are executed. It should be noted that the computer-readable medium mentioned above in the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above.

[0102] More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this application, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, an electromagnetic signal, an optical signal, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0103] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0104] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and the AND or flow chart, as well as the combination of the boxes in the block diagram and the AND or flow chart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0105] The terms "first", "second", etc. are used to distinguish similar objects, rather than to describe or indicate a particular order or sequence.

[0106] The term "comprise" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed or inherent to such process, method, article, apparatus, or apparatus.

[0107] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings.

[0108] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the present application.

Claims

1. An aircraft aerodynamic simulation modeling method, characterized in that: The method comprises: Obtain flight test data of the target aircraft; Determining target data based on the test flight data, the target data including six-degree-of-freedom force / torque coefficients acting on the target aircraft and relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients, the relationship parameters being used to characterize factors affecting the six-degree-of-freedom force / torque coefficients; Based on the target data, a pre-constructed non-fully connected neural network is trained to obtain a relationship mapping model, wherein the non-fully connected neural network includes a plurality of independent neuron groups, each neuron group is used to independently map the correspondence between at least one six-degree-of-freedom force / torque coefficient and a relationship parameter, and the non-fully connected neural network is trained using a pre-constructed hybrid loss function, wherein the hybrid loss function includes a data fitting term and a physical verification term, the data fitting term is used to minimize the deviation between the six-degree-of-freedom force / torque coefficient and the target data, and the physical verification term is used to verify the six-degree-of-freedom force / torque coefficient according to preset aerodynamic rules; Using the relationship mapping model, constructing an aerodynamic model interpolation table, wherein the aerodynamic model interpolation table is used to indicate at least one six-degree-of-freedom force / torque coefficient corresponding to each relationship parameter; Performing aerodynamic simulation modeling on the target aircraft according to the aerodynamic model interpolation table, and updating the relationship mapping model according to the modeling results; The method of constructing an aerodynamic model interpolation table by using the relationship mapping model includes: Based on aerodynamic principles, the relationship parameters are divided into a plurality of parameter groups, each parameter group corresponds to at least one neuron group in the relationship mapping model, and each parameter group generates at least one of the six-degree-of-freedom force / torque coefficients through the neuron group; Determining a target parameter group and a target neuron group, wherein the target parameter group is any one of a plurality of parameter groups, and the target neuron group is a neuron group corresponding to the target parameter group; Determining discrete target interpolation points according to the target parameter group, and using the target interpolation points as inputs to the target neuron group; Determine the output result of the N-1th layer of neurons in the target neuron group, where N is the total number of layers of the target neuron group; According to the output results corresponding to each target parameter group, the aerodynamic model interpolation table is determined.

2. The aircraft aerodynamic simulation modeling method according to claim 1, characterized in that: Determining target data based on the flight test data includes: detecting the flight test data and determining missing values ​​and abnormal values ​​in the flight test data; Cleaning the missing values ​​and outliers to obtain cleaned flight test data; Performing inverse calculation on the cleaned test flight data to determine the six-degree-of-freedom forces / torques acting on the target aircraft; The six-degree-of-freedom force / torque is dimensionally non-quantized to obtain six-degree-of-freedom force / torque coefficients acting on the target aircraft.

3. The aircraft aerodynamic simulation modeling method according to claim 2, characterized in that: The determining of target data further includes: Acquiring configuration data of the target aircraft; determining configuration parameters of the target aircraft according to the configuration data, wherein the configuration parameters are data in the configuration data related to the flight state of the target aircraft; determining, based on the configuration parameters, influencing factors of six-degree-of-freedom forces / torques on the target aircraft; The influencing factors are dimensionally processed to determine the relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients.

4. The aircraft aerodynamic simulation modeling method according to claim 1, characterized in that: The six-degree-of-freedom force / moment coefficients include the lift coefficient, drag coefficient, side force coefficient, pitch moment coefficient, roll moment coefficient, and yaw moment coefficient of the target aircraft; The relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients include angle of attack parameters, flap configuration parameters, landing gear configuration parameters, engine thrust parameters, angle of attack change rate parameters, and pitch angle change rate parameters of the target aircraft.

5. The aircraft aerodynamic simulation modeling method according to claim 4, characterized in that: The lift coefficient includes a static stability coefficient, a first dynamic stability coefficient, and a second dynamic stability coefficient of the target aircraft. The first dynamic stability coefficient is a dynamic stability coefficient affected by the target aircraft's angle of attack, and the second dynamic stability coefficient is a dynamic stability coefficient affected by the target aircraft's pitch angular velocity.

6. The aircraft aerodynamic simulation modeling method according to claim 1, characterized in that: The hybrid loss function satisfies: ； Where N is the number of training samples, is the prediction coefficient vector of the i-th sample, is the true value coefficient vector of the i-th sample, is the strength factor of the aerodynamic rule constraint, K is the number of aerodynamic rules, is the weight of the K-th physical rule, is the kth physical constraint function.

7. An aircraft aerodynamic simulation modeling system, applied to the method according to any one of claims 1 to 6, characterized in that: The system comprises: A data acquisition module, used to acquire the flight test data of the target aircraft; a parameter calculation module, configured to determine target data based on the test flight data, the target data including six-degree-of-freedom force / torque coefficients acting on the target aircraft and relationship parameters corresponding to the six-degree-of-freedom force / torque coefficients, the relationship parameters being used to characterize factors influencing the six-degree-of-freedom force / torque coefficients; a model training module, configured to train a pre-constructed non-fully connected neural network based on the target data to obtain a relationship mapping model, wherein the non-fully connected neural network includes a plurality of independent neuron groups, each neuron group being configured to independently map a correspondence between at least one six-degree-of-freedom force / torque coefficient and a relationship parameter, and wherein the non-fully connected neural network is trained using a pre-constructed hybrid loss function, wherein the hybrid loss function includes a data fitting term and a physical verification term, wherein the data fitting term is configured to minimize a deviation between the six-degree-of-freedom force / torque coefficient and the target data, and the physical verification term is configured to verify the six-degree-of-freedom force / torque coefficient according to preset aerodynamic rules; an interpolation construction module, configured to construct an aerodynamic model interpolation table using the relationship mapping model, wherein the aerodynamic model interpolation table is configured to indicate at least one six-degree-of-freedom force / torque coefficient corresponding to each discrete interpolation point; The simulation modeling module is used to perform aerodynamic simulation modeling on the target aircraft according to the aerodynamic model interpolation table, and update the relationship mapping model according to the modeling results.

8. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are configured to be executed by the computer to implement the method according to any one of claims 1 to 6.

Images