Multi-model migration optimization method and system for aviation large model, computing device and storage medium

Abstract

The invention relates to the technical field of aviation large model optimization, provides a multi-model migration optimization method and system for an aviation large model, and aims to solve the problem that the stability and consistency of knowledge migration are difficult to maintain in a complex aerodynamic environment due to dependence on a static feature space in the prior art. The method comprises the following steps: constructing a flight mechanics feature library based on flight mechanics parameters and control response data of a source aircraft model and a target aircraft model; executing coupling simulation to generate a compensation parameter set; establishing a model knowledge migration channel, reconstructing the knowledge migration channel according to the compensation parameter set, and generating a target migration channel and a migration weight regulation and control function; and adjusting the target migration channel and monitoring the distribution offset of the aerodynamic characteristic knowledge component, updating the aerodynamic difference characteristics, and iteratively executing the operation from coupling simulation to adjustment of the transmission state register until the distribution offset enters a steady-state operation interval. According to the method, high-precision knowledge migration from a wide-body aircraft to a narrow-body aircraft is realized, and the aviation airworthiness safety standard is met.

Description

Technical Field

[0001] The present invention relates to the technical field of optimization of large aviation models, and particularly to a multi-aircraft type migration optimization method and system for large aviation models. Background Art

[0002] In the aviation field, with the accelerating speed of aircraft model updates and the increasing demand for multi-aircraft type collaborative operation, how to effectively migrate aerodynamic characteristic knowledge between different aircraft types has become a key technical challenge. Especially in the migration scenario where there are significant differences in aerodynamic characteristics, such as from wide-body aircraft to Airbus narrow-body aircraft, there are large differences in flight mechanics parameters and control responses, making it difficult for traditional model migration methods to meet the requirements of high-precision and low-error knowledge reuse.

[0003] Currently, existing research has proposed a migration learning framework based on unified feature space mapping for handling aerodynamic characteristic migration problems between different aircraft types. It constructs a general flight data representation space, projects the flight data of the source aircraft type and the target aircraft type into this space respectively, and uses domain adaptation technology to align the distribution differences between them. And a dynamic weighting mechanism is adopted to adjust the participation degree of source model knowledge in the training of the target model, so as to achieve knowledge migration optimization from one aircraft type to another. However, existing methods still have obvious limitations when facing aircraft type migrations with large differences in aerodynamic characteristics. Specifically, due to their reliance on the construction of a static feature space, it is easy to cause information distortion and deviation during the knowledge migration process; there is a lack of real-time monitoring and feedback adjustment mechanisms for the knowledge distribution state in the migration path, making it difficult to maintain the stability and consistency of knowledge migration in a complex aerodynamic environment, etc. Summary of the Invention

[0004] The present invention provides a multi-aircraft type migration optimization method and system for large aviation models to solve the problems in the prior art that rely on a static feature space, which is easy to cause information distortion and deviation during the knowledge migration process; and it is difficult to maintain the stability and consistency of knowledge migration in a complex aerodynamic environment, etc.

[0005] In a first aspect, the present invention provides a multi-aircraft type migration optimization method for large aviation models, including: Based on the obtained flight mechanics parameters and control response data of the source aircraft type and the target aircraft type, construct a flight mechanics feature library containing aerodynamic difference features; According to the flight mechanics feature library, perform a coupled simulation operation to generate a set of compensation parameters; Establish a knowledge migration channel from the pre-trained large model of the source aircraft type to the large model of the target aircraft type; According to the set of compensation parameters, reconstruct the transmission buffer area of the knowledge migration channel to obtain a target migration channel, and at the same time generate a migration weight regulation function; Using the migration weight regulation function, adjust the transmission status register of the target migration channel to obtain an adjusted migration channel; Monitor the distribution offset of the aerodynamic characteristic knowledge component in the adjusted migration channel during the migration process, and update the aerodynamic difference characteristics of the flight mechanics feature library according to the distribution offset, so as to iteratively execute the coupled simulation operation to the operation of adjusting the transmission status register until the distribution offset enters the steady-state operation interval.

[0006] Optionally, according to the flight mechanics feature library, perform a coupled simulation operation to generate a set of compensation parameters, including: Calculate the dynamic fluid pressure of the source aircraft model and the target aircraft model under the same flight state parameters according to the aerodynamic difference characteristics and flight state parameters of the flight mechanics feature library, so as to generate a solid deformation boundary constraint according to the dynamic fluid pressure; Synchronously excite the interaction process of the source aircraft model and the target aircraft model under the same flight state parameters to obtain the fluid pressure distribution responses corresponding to the source aircraft model and the target aircraft model respectively; Transfer the fluid pressure distribution response to the action area of the solid deformation boundary constraint to calculate the response quantity difference value of the source aircraft model and the target aircraft model under the same fluid pressure distribution response, and generate a set of compensation parameters according to the corresponding relationship between the response quantity difference value and the flight state parameters.

[0007] Optionally, transfer the fluid pressure distribution response to the action area of the solid deformation boundary constraint to calculate the response quantity difference value of the source aircraft model and the target aircraft model under the same fluid pressure distribution response, and generate a set of compensation parameters according to the corresponding relationship between the response quantity difference value and the flight state parameters, including: Transfer the fluid pressure distribution response to the action area of the solid deformation boundary constraint to obtain the solid deformation response quantities corresponding to the source aircraft model and the target aircraft model respectively under the action of the same fluid pressure distribution response, so as to calculate the response quantity difference value; Form a three-dimensional flight state coordinate according to the altitude layer identifier, speed interval code and flight attitude angle of the flight state parameters; Associate the response quantity difference value with the storage location corresponding to the three-dimensional flight state coordinate, and connect the response quantity difference values of adjacent storage locations to construct multiple difference value change trajectories; Convert all the difference value change trajectories into corresponding executable compensation instruction parameters, and combine all the executable compensation instruction parameters to generate a set of compensation parameters.

[0008] Optionally, establish a knowledge migration channel from the pre-trained source aircraft model large model to the target aircraft model large model, including: According to the distribution of hardware resources in the distributed computing node cluster, perform computing resource allocation and model loading processing on the pre-trained source model of the aircraft type to obtain an airworthiness certification model instance; According to the structural definition of the target model of the aircraft type, perform initialization processing on the target model of the aircraft type to obtain a model instance to be migrated; According to the airworthiness certification model instance and the model instance to be migrated, configure a bidirectional data route between the source model of the aircraft type and the target model of the aircraft type to form a basic migration channel; According to the need for aerodynamic characteristic knowledge migration, implant a feature extraction operator into the basic migration channel to generate a feature-enhanced migration channel; Synchronously load the encoding rules of flight state parameters into the feature-enhanced migration channel to obtain a synchronized migration channel; According to the aviation data security specification, perform security encapsulation processing on the synchronized migration channel to obtain a knowledge migration channel.

[0009] Optionally, according to the compensation parameter set, reconstruct the transmission buffer of the knowledge migration channel to obtain a target migration channel, and at the same time generate a migration weight regulation function, including: Generate aerodynamic difference compensation instruction parameters and load spectrum offset coefficients according to the compensation parameter set; Obtain the original airfoil feature vector according to the aerodynamic characteristic knowledge component transmitted in the knowledge migration channel; According to the pressure correction amount of the aerodynamic difference compensation instruction parameter, adjust the pressure distribution parameter of the original airfoil feature vector to reconstruct the transmission buffer of the knowledge migration channel to obtain a reconstructed migration channel; Calculate the change gradient value according to the flight state parameters, and calculate the weight factor in combination with the load spectrum offset coefficient; Compile the initial migration weight regulation function, load the migration weight regulation function into the reconstructed migration channel to obtain a target migration channel; Start the rudder effectiveness response validator of the target migration channel to obtain a verification result, and set the operation boundary value of the initial migration weight regulation function according to the verification result and the weight factor to obtain a migration weight regulation function.

[0010] In a second aspect, the present invention provides a multi-aircraft-type migration optimization system for an aviation large model, including: A construction module for constructing a flight mechanics feature library containing aerodynamic difference features based on the obtained flight mechanics parameters and control response data of the source aircraft type and the target aircraft type; A simulation module for performing a coupled simulation operation according to the flight mechanics feature library to generate a compensation parameter set; A building module, configured to establish a knowledge transfer channel from a pre-trained large model of a source aircraft type to a large model of a target aircraft type; A reconstruction module, configured to reconstruct a transmission buffer of the knowledge transfer channel according to the set of compensation parameters to obtain a target transfer channel, and generate a transfer weight regulation function at the same time; An adjustment module, configured to use the transfer weight regulation function to adjust a transmission status register of the target transfer channel to obtain an adjusted transfer channel; An update module, configured to monitor a distribution offset of an aerodynamic characteristic knowledge component in the adjusted transfer channel during the transfer process, and update an aerodynamic difference characteristic of the flight mechanics feature library according to the distribution offset, so as to iteratively execute a coupled simulation operation to an operation of adjusting the transmission status register until the distribution offset enters a steady-state operation interval.

[0011] In a third aspect, an embodiment of the present invention provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a multi-aircraft type migration optimization method of an aviation large model as described in the first aspect above.

[0012] In a fourth aspect, an embodiment of the present invention provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it implements a multi-aircraft type migration optimization method of an aviation large model as described in the first aspect.

[0013] In the present invention, based on the obtained flight mechanics parameters and control response data of the source aircraft type and the target aircraft type, a flight mechanics feature library including aerodynamic difference characteristics is constructed; according to the flight mechanics feature library, a coupled simulation operation is performed to generate a set of compensation parameters; a knowledge transfer channel from a pre-trained large model of the source aircraft type to a large model of the target aircraft type is established; according to the set of compensation parameters, a transmission buffer of the knowledge transfer channel is reconstructed to obtain a target transfer channel, and a transfer weight regulation function is generated at the same time; the transfer weight regulation function is used to adjust a transmission status register of the target transfer channel to obtain an adjusted transfer channel; a distribution offset of an aerodynamic characteristic knowledge component in the adjusted transfer channel during the transfer process is monitored, and the aerodynamic difference characteristic of the flight mechanics feature library is updated according to the distribution offset, so as to iteratively execute a coupled simulation operation to an operation of adjusting the transmission status register until the distribution offset enters a steady-state operation interval.

[0014] Advantages of the present invention: The technical solution provided by the present invention solves the problem of fragmented feature space caused by scattered aircraft model data in traditional methods by constructing a flight mechanics feature library and unifying the aerodynamic parameters of heterogeneous aircraft models, providing a benchmark data source for cross-aircraft model migration; through performing fluid-solid coupling simulation, physically quantifying the aerodynamic differences, generating a set of compensation parameters based on the differences in solid deformation responses, replacing the static feature space mapping, and eliminating the distortion of aerodynamic characteristics (such as the differences in wing bending and torsion deformations); realizing private and secure knowledge transmission, avoiding the risk of sensitive flight data leakage through two-way routing and secure encapsulation under a distributed architecture, and meeting the compliance requirements of aviation data; dynamically correcting the offset of the feature space, adjusting the airfoil pressure distribution in real time based on the set of compensation parameters, and cracking the information distortion caused by the static feature space; synchronously generating a migration weight regulation function with safety constraints to prevent the divergence of the migration process; by adjusting the transmission status register, realizing adaptive migration intensity control, dynamically adjusting the knowledge transmission intensity in combination with the real-time change rate of flight state parameters, solving the problem of lagging response of traditional weighting mechanisms, and adapting to the aerodynamic mutation scenario in the transonic region; triggering the update of the feature library and simulation iteration through distributed offset monitoring, overcoming the defect of the lack of real-time feedback in existing methods, and maintaining the migration stability. Further, based on the set of compensation parameters, dynamically reconstruct the transmission buffer of the migration channel. Specifically, correct the pressure distribution parameters through the aerodynamic difference compensation instruction to reconstruct the channel feature space; compile the migration weight regulation function with the weight factors of the load spectrum offset coefficient and the change gradient value, and set the operation boundary value through the rudder effectiveness safety validator to realize the dual embedding of physical characteristics and safety constraints. By dynamically correcting the pressure distribution parameters, solve the problem of information distortion in the aerodynamic mutation scenario (such as transonic shock offset) of traditional unified feature space mapping, and ensure that narrow-body aircraft accurately inherit the aerodynamic characteristics of wide-body aircraft; the rudder effectiveness response validator converts the flight safety boundary (such as the limit of rudder surface deflection) into the operation boundary value of the weight regulation function, cracking the defect of insufficient stability of traditional dynamic weighting mechanisms in complex aerodynamic environments, and ensuring that the migration process meets the airworthiness standards.

[0015] These aspects or other aspects of the present invention will be more clearly understood in the following description of the embodiments. Brief Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 The flowchart of a multi-aircraft model migration optimization method for an aviation large model provided by the present invention is shown; Figure 2Shows a schematic structural diagram of a multi-aircraft model migration optimization system for an aviation large model provided by the present invention; Figure 3 Shows a schematic structural diagram of a computing device provided by the present invention. Detailed implementation manners

[0018] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0019] In some processes described in the specification, claims and above-mentioned drawings of the present invention, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0021] Aiming at the limitations faced by existing transfer learning methods in cross-aircraft aerodynamic characteristic transfer, such as the static feature space being difficult to adapt to complex fluid-structure coupling effects, and the lack of a dynamic monitoring and control mechanism for distribution shift during the knowledge transfer process, etc., the present invention takes the aerodynamic difference modeling between the source aircraft and the target aircraft as the core, constructs a flight mechanics feature library containing flight mechanics difference features, and combines fluid-solid coupling simulation means to generate a compensation parameter set to realize the dynamic reconstruction of the knowledge transfer channel and the adaptive adjustment of the transmission state. At the same time, a transfer weight regulation function and a distribution shift feedback mechanism are introduced to form a closed-loop iterative optimization structure, thereby effectively improving the stability and accuracy of the transfer process and breaking through the performance bottleneck of traditional solutions in high-difference transfer scenarios such as from wide-body aircraft to narrow-body aircraft. Figure 1 For the embodiments of the present invention, a flowchart of a multi-aircraft model migration optimization method for an aviation large model is provided, as Figure 1 shown, the method includes: Step 101: Based on the flight mechanics parameters and control response data of the source aircraft model and the target aircraft model obtained, construct a flight mechanics feature library including aerodynamic difference features; In this step, the source aircraft model refers to the aircraft model that provides the basis for knowledge transfer (such as a certain wide-body airliner model), and its aerodynamic model serves as the starting point for transfer. The target aircraft model refers to the new model that receives knowledge transfer (such as a certain narrow-body airliner model), which needs to adapt to the aerodynamic characteristics of the source model. Flight mechanics parameters refer to physical quantities that reflect aerodynamic performance, including airfoil bending and torsion stiffness, Reynolds number sensitivity coefficient, etc. Control response data refers to data containing rudder deflection efficiency, roll response time delay, etc., which are used to quantify flight control characteristics. Aerodynamic difference features refer to difference quantification indicators such as the lift curve slope difference and pressure center offset between aircraft models. The flight mechanics feature library refers to a structured database that stores the aerodynamic difference feature matrix according to flight states.

[0022] In an embodiment of the present invention, based on the flight mechanics parameters (including airfoil bending and torsion stiffness, lift coefficient curve) and control response data (such as rudder deflection rate, elevator effectiveness) of the source aircraft model (such as a certain wide-body airliner model) and the target aircraft model (such as a certain narrow-body airliner model), aerodynamic difference features (such as wing pressure center offset) are extracted through feature fusion technology, and a structured flight mechanics feature library is constructed. This library stores the aerodynamic characteristic differences between aircraft models in matrix form, providing input for subsequent simulations.

[0023] Step 102: According to the flight mechanics feature library, perform a coupled simulation operation to generate a set of compensation parameters; In this step, the set of compensation parameters refers to the set of all parameters that can execute compensation instructions and is used for channel reconstruction.

[0024] In an embodiment of the present invention, according to the aerodynamic difference features of the flight mechanics feature library, the Navier-Stokes equation is solved by computational fluid dynamics to calculate the dynamic fluid pressure. Based on the dynamic fluid pressure, a solid deformation boundary constraint is generated. Under the same flight state parameters, the interaction process between the source aircraft model and the target aircraft model is synchronously excited, and the fluid pressure distribution responses corresponding to the source aircraft model and the target aircraft model are obtained respectively; the fluid pressure distribution response is transmitted to the action area of the solid deformation boundary constraint to calculate the difference value of the response amounts of the source aircraft model and the target aircraft model under the same fluid pressure distribution response. According to its corresponding relationship with the flight state parameters, a set of compensation parameters is generated.

[0025] Step 103: Establish a knowledge transfer channel from the pre-trained large model of the source aircraft model to the large model of the target aircraft model; In this step, the large model of the source aircraft model refers to a pre-trained neural network model that learns the aerodynamic characteristics of the source aircraft model. The large model of the target aircraft model refers to the initial model framework to be transferred, which inherits the knowledge structure of the source model. The knowledge transfer channel refers to the data transmission path under a distributed architecture, including two-way data routing and feature extraction operators.

[0026] In the embodiment of the present invention, for the hardware resource distribution of the distributed computing node cluster, calculate resource allocation and model loading processing for the pre-trained source model of the aircraft type, and initialize the target model of the aircraft type. According to the results of the above steps, establish a basic channel by configuring a bidirectional data route, implant an aerodynamic feature extraction operator (such as an airfoil lift-drag ratio capturer) into it, load the encoding rules of flight state parameters, and finally generate a knowledge transfer channel through secure encapsulation.

[0027] Step 104: According to the set of compensation parameters, reconstruct the transmission buffer of the knowledge transfer channel to obtain a target transfer channel, and simultaneously generate a transfer weight regulation function; In this step, the transmission buffer refers to the temporary storage area of the aerodynamic characteristic knowledge component in the memory. The target transfer channel refers to the transfer path after buffer reconstruction. The transfer weight regulation function refers to the machine code function that dynamically controls the knowledge transfer intensity (such as the bandwidth allocation ratio).

[0028] In the embodiment of the present invention, generate aerodynamic difference compensation instruction parameters and load spectrum offset coefficients according to the set of compensation parameters; obtain the original airfoil feature vector according to the aerodynamic characteristic knowledge component transmitted in the knowledge transfer channel; adjust the pressure distribution parameters of the original airfoil feature vector according to the pressure correction amount of the aerodynamic difference compensation instruction parameters to reconstruct the transmission buffer of the knowledge transfer channel and obtain the reconstructed transfer channel; calculate the change gradient value according to the flight state parameters, establish a weight factor between the load spectrum offset coefficient and the change gradient value, compile the initial transfer weight regulation function, and load it into the reconstructed transfer channel to obtain the target transfer channel; further set the operation boundary value of the initial transfer weight regulation function to obtain the transfer weight regulation function.

[0029] Step 105: Use the transfer weight regulation function to adjust the transmission status register of the target transfer channel to obtain an adjusted transfer channel; In this step, the transmission status register refers to the hardware register that stores the channel bandwidth and priority configuration. The adjusted transfer channel refers to the final transfer path after weight regulation, and the status register value has been updated.

[0030] In the embodiment of the present invention, interpret the executable code segment of the transfer weight regulation function to separate the transmission intensity control parameters; obtain the boundary constraint value of the rudder effectiveness response validator, combine the real-time change rate of the flight state parameters and the transmission intensity control parameters, calculate the transmission intensity correction amount; input it into the transmission scheduler of the target transfer channel to trigger the bandwidth reallocation operation of the transmission scheduler, and adjust the transmission status register of the target transfer channel according to the updated bandwidth configuration to obtain the adjusted transfer channel.

[0031] Step 106: Monitor the distribution offset of the aerodynamic characteristic knowledge components in the adjusted migration channel during the migration process, and update the aerodynamic difference characteristics of the flight mechanics feature library according to the distribution offset, so as to iteratively execute the coupled simulation operation to the operation of adjusting the transmission status register until the distribution offset enters the steady-state operation interval; In this step, the aerodynamic characteristic knowledge components refer to the core migration data such as the airfoil pressure distribution vector and the rudder effectiveness gain coefficient. The distribution offset refers to the difference (standard deviation difference) between the knowledge component distribution and the benchmark during the migration process. The steady-state operation interval refers to the allowable fluctuation range (-0.15 to +0.15) of the distribution offset, which is determined by the airworthiness standard.

[0032] In the embodiment of the present invention, monitor the distribution offset of the aerodynamic characteristic knowledge components (such as the airfoil pressure distribution vector) in the adjusted migration channel. The distribution offset = the current distribution standard deviation - the benchmark distribution standard deviation. When the distribution offset exceeds the steady-state operation interval (±0.15), update the aerodynamic difference characteristics of the flight mechanics feature library (such as increasing the leading-edge pressure coefficient threshold by 10%), and re-trigger the fluid-structure interaction simulation to the adjustment operation of the transmission status register until the offset converges.

[0033] The embodiment of the present invention physically quantifies the aerodynamic differences between aircraft models through fluid-structure interaction simulation to generate dynamic compensation parameters; uses buffer reconstruction to correct the problem of feature space distortion and breaks through the limitation of static mapping; combines the weight control mechanism of rudder effectiveness safety verification to solve the problem of migration instability under complex working conditions; the closed-loop iterative mechanism monitors the offset in real time to optimize the migration path, and finally realizes high-precision knowledge migration from wide-body aircraft to narrow-body aircraft, with the error convergence speed improved and meeting the airworthiness safety standards.

[0034] The present invention provides a specific embodiment. In step 102, according to the flight mechanics feature library, perform a coupled simulation operation to generate a set of compensation parameters, which specifically includes the following steps: Step 201: Calculate the dynamic fluid pressure of the source aircraft model and the target aircraft model under the same flight state parameters according to the aerodynamic difference characteristics and flight state parameters of the flight mechanics feature library, so as to generate a solid deformation boundary constraint according to the dynamic fluid pressure; In this step, the flight state parameters refer to the three-dimensional combination of flight altitude (unit: km), Mach number (unit: 1), and angle of attack (unit: °), which are used to uniquely identify a specific flight condition. The dynamic fluid pressure refers to the time-varying pressure value (unit: kPa) of the airflow acting on the aircraft surface, which is calculated by multiplying the air density by the square of the speed and then multiplying by the pressure coefficient. The solid deformation boundary constraint refers to the structural displacement limit value (such as wing bending ≤ 6 mm) derived from the fluid pressure and is calculated based on the material flexibility matrix (deformation = pressure × flexibility).

[0035] In the embodiments of the present invention, according to the aerodynamic difference features (such as the wing bending-torsion stiffness difference) in the flight mechanics feature library and the flight state parameters (such as altitude 10 km / Mach 0.8), the Navier-Stokes equations are solved by computational fluid dynamics to calculate the dynamic fluid pressure of the source aircraft model and the target aircraft model under the same working conditions. The dynamic fluid pressure = air density × velocity squared × pressure coefficient; a solid deformation boundary constraint is generated based on the dynamic fluid pressure. Specifically, the fluid pressure value is multiplied by the material flexibility matrix (which reflects the material stiffness characteristics) to obtain the displacement constraint value (such as the maximum displacement of the wing leading edge ≤ 120 mm), which is used as the solid domain simulation boundary constraint.

[0036] Step 202: Synchronously excite the interaction process of the source aircraft model and the target aircraft model under the same flight state parameters to obtain the fluid pressure distribution responses corresponding to the source aircraft model and the target aircraft model respectively; In this step, the fluid pressure distribution response refers to the set of pressure values at discrete points on the aircraft surface (such as 30 kPa at the leading edge, 50 kPa at the maximum thickness point, and 20 kPa at the trailing edge), which is obtained by solving the computational fluid dynamics flow field.

[0037] In the embodiments of the present invention, under the same flight state parameters (such as altitude 10 km, Mach 0.8), the distributed simulation platform synchronously excites the fluid-structure interaction process of the source aircraft model and the target aircraft model. Specifically, the finite volume method is used to discretize the fluid domain, and the surface pressure distribution is calculated. For example, the surface pressure distribution on the upper surface of the wing = inlet total pressure - dynamic fluid pressure × loss coefficient; it is mapped to the solid grid nodes (such as through radial basis function interpolation) to obtain the fluid pressure distribution responses of the source aircraft model and the target aircraft model respectively. Among them, the row vector represents the chord position, and the column vector represents the surface pressure distribution.

[0038] Step 203: Transmit the fluid pressure distribution response to the action area of the solid deformation boundary constraint to calculate the difference value of the response quantities of the source aircraft model and the target aircraft model under the same fluid pressure distribution response, and generate a set of compensation parameters according to the corresponding relationship between the response quantity difference value and the flight state parameters; In this step, the response quantity difference value refers to the absolute difference between the solid deformation amounts of the two aircraft models under the same working conditions.

[0039] In the embodiments of the present invention, the fluid pressure distribution response is transmitted to the action area of the solid deformation boundary constraint (such as the leading edge 20% chord length area), the solid equilibrium equation is solved based on the solid deformation boundary constraint to obtain the solid deformation response quantity; the difference value of the response quantities of the source model and the target model under the same pressure distribution is calculated, and the difference value of the response quantities = |source wing deformation quantity - target wing deformation quantity|; according to the altitude layer identifier, speed interval code, and flight attitude angle of the flight state parameters, a three-dimensional flight state coordinate is formed; the difference value of the response quantities is associated with the storage location corresponding to the three-dimensional flight state coordinate, and the difference values of the response quantities at adjacent storage locations are connected to construct a plurality of difference value change trajectories; all the difference value change trajectories are converted into corresponding executable compensation instruction parameters, and all the executable compensation instruction parameters are combined to generate a compensation parameter set.

[0040] In the embodiments of the present invention, the aerodynamic structure coupling differences between models are accurately quantified through physical simulation. Specifically, the dynamic fluid pressure calculation breaks through the limitations of empirical formulas and reflects the influence of real transonic shock waves; the two-way fluid-structure coupling solves the errors of traditional one-way simulation and captures the feedback effect of wing deformation on the airflow; the three-dimensional trajectory compensation parameters eliminate the discrete errors of the static look-up table method and achieve continuous adaptation of the entire flight envelope.

[0041] The present invention provides a specific embodiment. In step 203, the fluid pressure distribution response is transmitted to the action area of the solid deformation boundary constraint to calculate the difference value of the response quantities of the source model and the target model under the same fluid pressure distribution response, and a compensation parameter set is generated according to the corresponding relationship between the difference value of the response quantities and the flight state parameters. The specific steps are as follows: Step 211: Transmit the fluid pressure distribution response to the action area of the solid deformation boundary constraint to obtain the solid deformation response quantities corresponding to the source model and the target model respectively under the action of the same fluid pressure distribution response, so as to calculate the difference value of the response quantities; In this step, the solid deformation response quantity refers to the physical deformation quantity (unit: mm) of the aircraft structure under the action of fluid pressure, including wing deflection and fuselage bending.

[0042] In the embodiments of the present invention, the fluid pressure distribution response (such as the sequence of pressure values on the upper surface of the wing) is transmitted to the action area of the solid deformation boundary constraint (such as the connection point of the front beam and the skin) through the finite element coupling interface, and the finite element analysis is used to solve the structural stress distribution to obtain the solid deformation response quantities of the source model (a certain wide-body airliner model) and the target model (a certain narrow-body airliner model) under the same pressure action. For example, the wing deflection value = structural stress distribution × flexibility. Calculate the absolute difference between the two, that is, the difference value of the response quantities = |wing deflection value of the source model - wing deflection value of the target model|.

[0043] Step 212: Form a three-dimensional flight state coordinate based on the altitude layer identifier, speed interval code, and flight attitude angle of the flight state parameters; In this step, the altitude layer identifier refers to the digital encoding of the flight altitude (an integer from 0 to 15), where 0 corresponds to sea level and 15 corresponds to an altitude of 15 km, divided according to the international standard flight altitude layer. The speed interval code refers to the discretized encoding of the Mach number (an integer from 0 to 12), where 0 corresponds to stationary and 12 corresponds to Mach 1.2, covering the subsonic to supersonic flight envelope. The flight attitude angle refers to the range of the aircraft's angle of attack (-10° to +40°), with negative values indicating a diving attitude and positive values indicating a climbing attitude. The three-dimensional flight state coordinate refers to an integer triple composed of the altitude layer identifier, speed interval code, and flight attitude angle (such as [10, 8, 5]), which uniquely identifies a specific flight condition.

[0044] In the embodiment of the present invention, when parsing the flight state parameters, specifically, for the altitude layer identifier, it includes 0 - 15, corresponding to 0 - 15 km. For example, an altitude of 10 km is encoded as 10; for the speed interval code, it includes 0 - 12, corresponding to 0 - 1.2 Mach. For example, Mach 0.8 is encoded as 8; the range of the flight attitude angle is -10° to +40°, such as an angle of attack of 5°. Combining the three forms a three-dimensional flight state coordinate (such as [10, 8, 5]), which serves as the spatial positioning reference for aerodynamic differences.

[0045] Step 213: Associate the response quantity difference value with the storage location corresponding to the three-dimensional flight state coordinate, and connect the response quantity difference values of adjacent storage locations to construct multiple difference value change trajectories; In this step, the storage location refers to the physical memory address. The difference value change trajectory refers to an interpolation continuous curve of the response quantity difference value between adjacent three-dimensional coordinate points, reflecting the gradual change law of aerodynamic differences with the flight state.

[0046] In the embodiment of the present invention, associate the response quantity difference value (such as 1.7 mm) with the storage location corresponding to the three-dimensional coordinate, such as memory address = base address + altitude × 13 × 51 + speed × 51 + angle; traverse adjacent coordinate units, connect the response quantity difference values through linear interpolation, where interpolation = previous response quantity difference value + (next response quantity difference value - previous response quantity difference value) × step size ratio, and construct a difference value change trajectory, such as a continuous curve of the deflection difference increasing from 1.7 mm to 1.9 mm when the angle of attack changes from 5° to 6°.

[0047] Step 214: Convert all difference value change trajectories into corresponding executable compensation instruction parameters, and combine all executable compensation instruction parameters to generate a compensation parameter set; In this step, the executable compensation instruction parameter refers to a compensation command in machine code format that can be directly executed by the flight control computer.

[0048] In the embodiments of the present invention, the variation trajectory of the difference value is converted into machine-executable compensation instruction parameters. Specifically, key points in the variation trajectory of the difference value are extracted, such as the difference value of the response amount corresponding to an angle of attack of 5° or 6°, and compiled into the format of avionics instructions; all executable compensation instruction parameters are combined to generate a compensation parameter set.

[0049] In the embodiments of the present invention, through finite element analysis of stress deformation conversion, the error of traditional empirical estimation is solved; the continuous flight state is discretized into addressable storage units, breaking through the limitation of two-dimensional look-up tables; linear interpolation is used to generate a continuous compensation curve to eliminate the adaptation blind area of aerodynamic mutations in the transonic region.

[0050] The present invention provides a specific embodiment. Step 103: Establish a knowledge migration channel from the pre-trained large model of the source aircraft type to the large model of the target aircraft type, which specifically includes the following steps: Step 301: According to the hardware resource distribution of the distributed computing node cluster, perform computing resource allocation and model loading processing on the pre-trained large model of the source aircraft type to obtain an airworthiness certification model instance; In this step, the distributed computing node cluster refers to the integrated modular avionics architecture of the avionics system, which includes partitioned computers such as flight control, navigation, and engine control, and is interconnected through a backplane bus. The hardware resource distribution refers to the physical allocation scheme of CPU cores, memory blocks, and I / O ports. The airworthiness certification model instance refers to the neural network operation instance that has passed the DO-178C Level A certification, including model weights, operating environment, and digital signature, and can be deployed on avionics hardware.

[0051] In the embodiments of the present invention, according to the hardware resource distribution of the distributed computing node cluster (such as the integrated modular avionics module of the avionics system), for example, CPU cores 0-3 are allocated to the flight control partition, and resource allocation is performed on the pre-trained large model of the source aircraft type through memory partition management technology: a 128MB memory space is divided, and the model weight file is loaded; after being compiled by the DO-178C Level A certification tool chain, an airworthiness certification model instance is generated.

[0052] Step 302: According to the structure definition of the large model of the target aircraft type, perform initialization processing on the large model of the target aircraft type to obtain a model instance to be migrated; In this step, the model instance to be migrated refers to the initialization entity of the neural network of the target aircraft type, including an untrained weight matrix and an empty data buffer, waiting for knowledge injection.

[0053] In the embodiments of the present invention, according to the structure definition of the large model of the target aircraft type, such as the number of neural network layers and nodes of a certain narrow-body airliner type, the model initialization function is called to allocate random initial values for the weight matrix, and the allocated data is written into the memory address of the target node to generate a model instance to be migrated.

[0054] Step 303: Configure a bidirectional data route between the large model of the source aircraft type and the large model of the target aircraft type according to the airworthiness certification model instance and the model instance to be migrated, so as to form a basic migration channel; In this step, the bidirectional data route refers to two independent virtual links, which realize data transmission from the source model to the target model (forward) and from the target back to the source (reverse). The basic migration channel refers to the minimum communication path that only contains the bidirectional route and does not embed data processing functions.

[0055] In the embodiment of the present invention, according to the airworthiness certification model instance (such as the address is 0x5000) and the model instance to be migrated (such as the address is 0x8000), virtual links are configured through the aviation data exchange network switch. Among them, the virtual link number from the large model of the source aircraft type to the large model of the target aircraft type is 101, and the bandwidth is 64 kilobits per second; the virtual link number from the large model of the target aircraft type to the large model of the source aircraft type is 102, and the bandwidth is 32 kilobits per second, so as to build a basic migration channel, and store the routing table information of this channel at the address 0x3000.

[0056] Step 304: According to the aerodynamic characteristic knowledge migration requirement, implant a feature extraction operator into the basic migration channel to generate a feature-enhanced migration channel; In this step, the aerodynamic characteristic knowledge migration requirement refers to the core aerodynamic parameters that need to be retained during the migration process, such as the lift coefficient derivative and the position of the center of pressure. The feature extraction operator refers to a hardware-accelerated mathematical operation unit, such as a convolution kernel implemented by a field-programmable gate array, which is used to extract features such as the wing airfoil pressure gradient in real time. The feature-enhanced migration channel refers to a data path that adds feature extraction capabilities to the basic migration channel and can process aerodynamic knowledge components online.

[0057] In the embodiment of the present invention, according to the aerodynamic characteristic knowledge migration requirement, a feature extraction operator is injected into the basic migration channel. Specifically, the VL101 data packet processing function is modified, and a convolution kernel (which is a 3×3 Gaussian filter) is added to extract the pressure distribution feature; the operator machine code (such as 0x90F2) is written into the channel coprocessor to generate a feature-enhanced migration channel, so as to reduce the feature extraction delay.

[0058] Step 305: Synchronously load the encoding rule of the flight state parameter to the feature-enhanced migration channel to obtain a synchronized migration channel; In this step, the encoding rule refers to the digital mapping rule of the flight state parameter. The synchronized migration channel refers to a channel that supports time-triggered communication, and each node synchronously transmits according to the global clock.

[0059] In the embodiment of the present invention, the encoding rules of flight state parameters (including that heights from 0 to 15 are represented by 4-bit binary numbers, the Mach number multiplied by 10 is converted into an 8-bit integer value, and the angle of attack plus 10 is used as the offset value) are burned into the field-programmable gate array register of the feature-enhanced migration channel; the clocks of each node are synchronized through the time-triggered Ethernet bus to obtain a synchronized migration channel, and its state synchronization period is 5 ms.

[0060] Step 306: According to the avionics data security specification, perform security encapsulation processing on the synchronized migration channel to obtain a knowledge migration channel; In this step, the avionics data security specification refers to the requirements of data encryption, access control, and integrity verification defined according to the standard.

[0061] In the embodiment of the present invention, according to the requirements of the avionics data security specification, the following security measures are implemented on the synchronized migration channel, including encrypting the payload data of the VL101 and VL102 virtual links using the Advanced Encryption Standard 256-bit algorithm to ensure the confidentiality of the transmitted content; restricting access to the network only by authorized devices by setting a Media Access Control address white list (for example: 00-0C-29-XX) to achieve effective access control; at the same time, adding a 32-bit cyclic redundancy check code to each frame of data to ensure the integrity of data transmission. Through the above measures, a knowledge migration channel that meets the security level requirements of the avionics system is constructed.

[0062] In the embodiment of the present invention, by allocating model resources, interference of critical flight control tasks by the migration process is prevented; feature extraction operators are implanted to accurately extract the aerodynamic characteristics of the airfoil to avoid distortion of knowledge during transmission; time-triggered Ethernet is used to achieve clock synchronization to ensure the determinism of cross-node data transmission; unauthorized access is prevented to meet the security requirements of the avionics system security level.

[0063] The present invention provides a specific embodiment. In step 104, according to the set of compensation parameters, reconstruct the transmission buffer of the knowledge migration channel to obtain a target migration channel, and at the same time generate a migration weight regulation function, which specifically includes the following steps: Step 401: Generate aerodynamic difference compensation instruction parameters and load spectrum offset coefficients according to the set of compensation parameters; In this step, the aerodynamic difference compensation instruction parameters refer to the physical correction amounts parsed from the machine instructions in the set of compensation parameters, which are used to adjust the airfoil characteristics. The load spectrum offset coefficient refers to the structural load difference value between the source aircraft model and the target aircraft model under the same working conditions, which reflects the aerodynamic stiffness difference.

[0064] In an embodiment of the present invention, machine instructions in the parsing compensation parameter set are parsed, and pneumatic difference compensation instruction parameters are extracted, such as a pressure correction amount of +1.2 kPa, and a load spectrum offset coefficient, such as a bending moment difference of 0.8 kN·m. Among them, an instruction decoder based on the x86 instruction set is used to separate the operation code from the parameters.

[0065] Step 402: Obtain an original airfoil feature vector according to the pneumatic characteristic knowledge component transmitted in the knowledge migration channel; In this step, the pneumatic characteristic knowledge component refers to the core data unit transmitted in the migration channel (such as an airfoil pressure coefficient sequence), which contains key information such as lift-drag characteristics. The original airfoil feature vector refers to a set of uncorrected wing section pressure distribution parameters, stored in the order of [leading edge, maximum thickness point, trailing edge].

[0066] In an embodiment of the present invention, a direct memory access controller is used to read the pneumatic characteristic knowledge component in the transmission buffer of the knowledge migration channel, such as a pressure distribution sequence [0.8, 0.5, 0.3], and extract the unprocessed original airfoil feature vector (including the pressure coefficients of the leading edge, maximum thickness, and trailing edge) from it; subsequently, a 3×3 Gaussian kernel is used as a feature selection filter to perform noise reduction processing on it to obtain a purified original airfoil feature vector.

[0067] Step 403: Adjust the pressure distribution parameters of the original airfoil feature vector according to the pressure correction amount of the pneumatic difference compensation instruction parameter to reconstruct the transmission buffer of the knowledge migration channel and obtain a reconstructed migration channel; In this step, the pressure correction amount refers to the specific correction value in the pneumatic difference compensation instruction (unit: kPa), which is superimposed on the original parameter through floating-point operation. The pressure distribution parameter refers to a set of pressure coefficient values of discrete points on the airfoil surface, which is the core variable determining the aerodynamic performance. The reconstructed migration channel refers to the migration path in which the pressure parameters in the transmission buffer are updated and pass through a cyclic redundancy check.

[0068] In an embodiment of the present invention, according to the pressure correction amount in the pneumatic difference compensation instruction parameter (, a floating-point addition operation is performed on the pressure distribution parameters in the original airfoil feature vector. For example, the adjusted trailing edge pressure value = the original trailing edge pressure value + the correction amount. The updated compensation parameter set is rewritten into the transmission buffer, and a reconstructed migration channel is generated through a 32-bit cyclic redundancy check mechanism. At the same time, the 7th bit of the status register is set to 1 to activate this channel.

[0069] Step 404: Calculate a change gradient value according to the flight state parameters, and calculate a weight factor in combination with the load spectrum offset coefficient; In this step, the change gradient value refers to the rate of change of flight state parameters over time, which is obtained by differential calculation: (current value - previous value) ÷ time difference. The weight factor refers to the mathematical mapping between the load spectrum offset coefficient and the change gradient value, which determines the regulation intensity.

[0070] In the embodiment of the present invention, the change gradient value is calculated based on the current flight state parameters (such as altitude 10 km, Mach number 0.8, angle of attack 5 degrees). Specifically, the change gradient value = altitude gradient + Mach gradient, where the altitude gradient = (current altitude - previous altitude) ÷ time interval; the Mach gradient = (current Mach number - previous Mach number) × quantization coefficient, and the quantization coefficient is a constant, such as 100. Subsequently, the weight factor is calculated as weight factor = load spectrum offset coefficient ÷ change gradient value.

[0071] Step 405: Compile the initial migration weight regulation function, and load the migration weight regulation function into the reconfigured migration channel to obtain the target migration channel; In this step, the initial migration weight regulation function refers to a machine code function without a set safety boundary, which implements the basic bandwidth allocation logic.

[0072] In the embodiment of the present invention, the LLVM compiler is used to generate the machine code of the initial migration weight regulation function, such as opcode 0xB2 plus parameter 0x2A. Subsequently, the machine code is loaded into the instruction queue of the co-processor of the reconfigured migration channel, and finally the target migration channel is generated, and the channel status word is updated to active.

[0073] Step 406: Start the rudder effect response validator of the target migration channel to obtain a verification result. According to the verification result and the weight factor, set the operation boundary value of the initial migration weight regulation function to obtain the migration weight regulation function; In this step, the rudder effect response validator refers to a flight control system hardware module that verifies whether the rudder surface deflection efficiency meets the standard through a look-up table method. The verification result refers to the physical safety boundary output by the rudder effect verification, such as the maximum load of the rudder 3.2 kN·m, with the unit of force / moment. The operation boundary value refers to the maximum output value allowed by the migration weight regulation function, such as the bandwidth upper limit 2.14, which is calculated based on the verification result × weight factor.

[0074] In the embodiment of the present invention, the rudder effect response validator of the target migration channel is started, and verification parameters are input, such as the rudder deflection limit ±25 degrees. By comparing with the rudder effect truth table, a verification result is obtained, for example, the safety boundary is 3.2 kN·m. Based on this result, the operation boundary value of the initial migration weight regulation function is set as operation boundary value = verification result × weight factor, and this boundary value is burned into the 12th to 15th bits of the function register to complete the safety boundary configuration.

[0075] In the embodiments of the present invention, the distortion problem of the static feature space is solved by adjusting the pressure parameters in real time; the load spectrum-state gradient coupling relationship enables the weight distribution to match the change of the flight state; the rudder effectiveness verification converts the aircraft operation limit into a mathematical boundary to eliminate the risk of out-of-control during the migration process.

[0076] The present invention provides a specific embodiment. In step 403, according to the pressure correction amount of the pneumatic difference compensation instruction parameter, the pressure distribution parameter of the original airfoil feature vector is adjusted to reconstruct the transmission buffer of the knowledge migration channel, and the reconstructed migration channel is obtained, which specifically includes the following steps: Step 411: Locate the pressure distribution parameter storage area to which the original airfoil feature vector belongs to obtain the pressure distribution parameter; In this step, the pressure distribution parameter storage area refers to the physical address space in the memory of the original airfoil feature vector (such as 0x5000 - 0x500B), and 32-bit floating-point pressure values are stored in the order of the leading edge, the maximum thickness, and the trailing edge.

[0077] In the embodiments of the present invention, the memory management unit is used to locate the pressure distribution parameter storage area in the original airfoil feature vector. The physical address range of this area is from 0x5000 to 0x500B, where the leading edge pressure value is stored at address 0x5000, the maximum thickness point pressure value is stored at address 0x5004, and the trailing edge pressure value is stored at address 0x5008. Each parameter occupies 4 bytes. The current pressure distribution parameter (for example, [0.75, 0.52, 0.28]) is read using the direct memory access method and loaded into the central processor register for subsequent processing.

[0078] Step 412: Obtain the pressure correction amount of the pneumatic difference compensation instruction parameter, and superimpose the pressure correction amount on the pressure distribution parameter to generate an updated pressure distribution parameter set; In this step, the updated pressure distribution parameter set refers to the set of corrected pressure distribution parameters, which is generated by floating-point addition of the pressure distribution parameter and the pressure correction amount and stored in the temporary buffer (0x7000).

[0079] In the embodiments of the present invention, the operation code in the pneumatic difference compensation instruction parameter is parsed to extract the corresponding pressure correction amount, such as an increase in the trailing edge pressure of +0.1 kPa. The scalar addition operation is performed by the floating-point arithmetic unit, and the calculation logic is that the updated pressure distribution parameter = pressure distribution parameter + pressure correction amount. Finally, an updated set of pressure distribution parameters is generated and temporarily stored in the temporary buffer.

[0080] Step 413: Add the updated pressure distribution parameter set to the pressure distribution parameter storage area to update the fields of the original airfoil feature vector to obtain a reconstructed airfoil feature vector; In this step, the reconstructed airfoil feature vector refers to a complete data structure that includes the updated pressure distribution parameters and the new version identifier.

[0081] In the embodiments of the present invention, the updated set of pressure distribution parameters is written into the original pressure distribution parameter storage area to overwrite the original data; subsequently, a field update function is called to update the version identifier of the airfoil feature vector from the original 0x01 to 0x02, generating a reconstructed airfoil feature vector, and updating the memory status word to updated. At the same time, a hardware interrupt signal is triggered as a notification signal indicating that the parameter update is complete.

[0082] Step 414: In response to the signal indicating the completion of updating the original airfoil feature vector, lock the write permission of the transmission buffer of the knowledge migration channel; In this step, the write permission refers to the write enable flag bit controlled by the memory protection unit, where 0 = writable and 1 = locked.

[0083] In the embodiments of the present invention, in response to the interrupt signal indicating the completion of the update, the memory protection unit sets the write protection flag bit (the 0th bit is set to 1) for the transmission buffer, restricting the write operation to this area to only allow privileged mode access, thereby ensuring data consistency and security during the reconstruction process.

[0084] Step 415: In the locked state, write the reconstructed airfoil feature vector into the aerodynamic characteristic data segment of the transmission buffer to activate the data consistency check mechanism to perform a check on the transmission buffer. According to the signal indicating that the check has passed, reset the sequence number identifier of the transmission buffer to generate a reconstructed migration channel; In this step, the aerodynamic characteristic data segment refers to the partition in the transmission buffer that specifically stores the airfoil pressure parameters, with a length of 12 bytes. The data consistency check mechanism refers to the hardware process of calculating the 32-bit checksum of the cyclic redundancy check in a dual-core lockstep manner. Core A and Core B independently calculate and compare the results, with an error tolerance of ≤ 1 clock cycle. The sequence number identifier refers to the version counter of the transmission buffer, which increments each time a reconstruction occurs to prevent data version confusion.

[0085] In the embodiments of the present invention, in the state where write protection is enabled (i.e., the locked state), the reconstructed airfoil feature vector is copied to the aerodynamic characteristic data segment of the transmission buffer. At this time, the data consistency check mechanism is activated, and in a dual-core lockstep manner, the two processing cores synchronously calculate the 32-bit checksum of the cyclic redundancy check. If the calculation results of the two cores are consistent, a signal indicating that the check has passed is generated, and then the write protection flag bit is cleared (the 0th bit is restored to 0); after receiving the signal indicating that the check has passed, the sequence number identifier in the transmission buffer is incremented to mark this reconstruction as a new version; at the same time, the reconstruction completion bit in the transmission status register is updated, and finally a reconstructed migration channel is generated.

[0086] In the embodiment of the present invention, through the mechanisms of write protection locking, writing, and unlocking, the risk of data competition that may occur during the reconstruction process is eliminated, ensuring the security and consistency of data operations. At the same time, a dual-redundancy check mechanism is introduced, and the dual-core parallel execution of cyclic redundancy check is used to double-guarantee the integrity of the pressure parameters, meeting the requirements for the protection of critical data in airworthiness standards.

[0087] The present invention provides a specific embodiment. In step 105, the transfer status register of the target transfer channel is adjusted by using the migration weight regulation function to obtain an adjusted transfer channel, which specifically includes the following steps: Step 501: Decode the executable code segment of the migration weight regulation function to separate the transmission intensity control parameter; In this step, the executable code segment refers to the machine instruction sequence after the compilation of the migration weight regulation function, which is stored at the memory address 0x9000 - 0x90FF and directly executed by the CPU. The transmission intensity control parameter refers to the proportional coefficient (a floating point number from 0.0 to 1.0) inside the regulation function, which determines the weight of the basic bandwidth allocation and is extracted by operand decoding.

[0088] In the embodiment of the present invention, the executable code segment (machine code such as 0xB2 0x18) of the migration weight regulation function is parsed by the central processing unit instruction decoder to extract the operation code (such as 0xB2) and the operand (such as 0x18) therein, and the transmission intensity control parameter (for example, the bandwidth allocation proportional coefficient is 0.75) is separated therefrom. The specific implementation method is to load the machine code into the instruction pipeline, and after the decoding unit divides and processes its operation code and operand, the extracted transmission intensity control parameter is stored in the register R1.

[0089] Step 502: Obtain the boundary constraint value of the rudder effectiveness response validator, and calculate the transmission intensity correction amount in combination with the real-time change rate of the flight state parameters and the transmission intensity control parameter; In this step, the boundary constraint value refers to the physical safety threshold (unit: kN·m) output by the rudder effectiveness validator, which reflects the ultimate bearing capacity of the aircraft control surface. The transmission intensity correction amount refers to the final bandwidth adjustment coefficient, which is used to dynamically adapt to the change of the flight state.

[0090] In the embodiments of the present invention, the boundary constraint value output by the rudder effectiveness response validator is read. For example, the maximum load of the rudder is 3.2 kN·m. Combining with the real-time change rate in the flight state parameters, such as the climb rate of 2.5 meters per second, and the obtained transmission intensity control parameter, such as 0.75, the transmission intensity correction amount is calculated through the floating-point arithmetic unit. The transmission intensity correction amount = transmission intensity control parameter × boundary constraint value ÷ real-time change rate. For example, 0.75 × 3.2 ÷ 2.5 = 0.96. Finally, the calculated transmission intensity correction amount (such as 0.96) is output and stored in the temporary storage area.

[0091] Step 503: Input the transmission intensity correction amount into the transmission scheduler of the target migration channel to trigger the bandwidth reallocation operation of the transmission scheduler, and obtain the updated bandwidth configuration; In this step, the transmission scheduler refers to the hardware control module of the avionics network switch, which realizes the virtual link bandwidth allocation. The updated bandwidth configuration refers to the new bandwidth value (unit: Kbps) that takes effect after the reallocation operation and is written into the switch configuration register.

[0092] In the embodiments of the present invention, an aviation full-duplex switched Ethernet switch chip is adopted. The transmission intensity correction amount is input into the transmission scheduler of the target migration channel. By sending a write command to the switch configuration register address, the bandwidth reallocation operation is triggered. The updated bandwidth = original bandwidth × transmission intensity correction amount, and the updated bandwidth configuration is updated to the corresponding field in the virtual link configuration table, such as the bandwidth configuration item of VL101, to generate the updated bandwidth configuration.

[0093] Step 504: Adjust the transmission status register of the target migration channel according to the updated bandwidth configuration to obtain the adjusted migration channel; In this step, the transmission status register refers to the hardware register that stores the real-time configuration of the channel.

[0094] In the embodiments of the present invention, according to the updated bandwidth configuration, the bandwidth control bits (bits 0 to 7) of the transmission status register of the target migration channel are written with new values (for example, the converted hexadecimal value 0x3D), and the parity bit (such as bit 8) is updated at the same time. After the writing is completed, the register latch signal is activated to generate the adjusted migration channel.

[0095] In the embodiments of the present invention, the transmission intensity is dynamically corrected through the real-time change rate of the flight state parameters, solving the problem of response lag of the traditional fixed weight mechanism under variable speed conditions such as climbing or diving; the boundary constraint value converts the physical limit of the aircraft into a mathematical boundary, eliminating the risk of system overload caused by over-allocation of bandwidth; the atomic writing of the register ensures the consistency of configuration switching, meeting the millisecond-level real-time requirement of the avionics system.

[0096] Figure 2The following is a schematic structural diagram of a multi-aircraft model migration optimization system for an aviation large model, as Figure 2 shown. The system includes: A construction module 21, configured to construct a flight mechanics feature library containing aerodynamic difference features based on the obtained flight mechanics parameters and control response data of the source aircraft model and the target aircraft model; A simulation module 22, configured to perform a coupled simulation operation according to the flight mechanics feature library to generate a set of compensation parameters; An establishment module 23, configured to establish a knowledge migration channel from a pre-trained source aircraft model large model to a target aircraft model large model; A reconstruction module 24, configured to reconstruct the transmission buffer area of the knowledge migration channel according to the set of compensation parameters to obtain a target migration channel, and simultaneously generate a migration weight regulation function; An adjustment module 25, configured to use the migration weight regulation function to adjust the transmission status register of the target migration channel to obtain an adjusted migration channel; An update module 26, configured to monitor the distribution offset of the aerodynamic characteristic knowledge component in the adjusted migration channel during the migration process, and update the aerodynamic difference features of the flight mechanics feature library according to the distribution offset, so as to iteratively execute the operations of the coupled simulation operation to the operation of adjusting the transmission status register until the distribution offset enters the steady-state operation interval.

[0097] Figure 2 The multi-aircraft model migration optimization system for an aviation large model can execute Figure 1 the multi-aircraft model migration optimization method for an aviation large model described in the embodiments shown. The implementation principle and technical effects will not be elaborated here. For the multi-aircraft model migration optimization system for an aviation large model in the above embodiments, the specific manners in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0098] In a possible design, Figure 2 the multi-aircraft model migration optimization system for an aviation large model in the embodiments shown can be implemented as a computing device, as Figure 3 shown. The computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, where the one or more computer instructions are called and executed by the processing component 32.

[0099] The processing component 32 is used for the multi-aircraft model migration optimization method for an aviation large model in the above Figure 1 embodiments.

[0100] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above methods. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above methods.

[0101] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0102] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0103] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0104] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0105] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above processing component, storage component, etc. may be basic server resources leased or purchased from the cloud computing platform.

[0106] The embodiment of the present invention also provides a computer storage medium storing a computer program, and when the computer program is executed by the computer, it can implement the above Figure 1 shown embodiment of a multi-aircraft model migration optimization method for an aviation large model.

[0107] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0108] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative work.

[0109] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-aircraft model migration optimization method for an aviation large model, characterized in that, Including: Based on the obtained flight mechanics parameters and control response data of the source aircraft model and the target aircraft model, construct a flight mechanics feature library containing aerodynamic difference features; According to the flight mechanics feature library, perform a coupled simulation operation to generate a set of compensation parameters; Establish a knowledge transfer channel from the pre-trained large model of the source aircraft model to the large model of the target aircraft model; According to the set of compensation parameters, reconstruct the transmission buffer of the knowledge transfer channel to obtain a target transfer channel, and generate a transfer weight adjustment function; Use the transfer weight adjustment function to adjust the transmission status register of the target transfer channel to obtain an adjusted transfer channel; Monitor the distribution offset of the aerodynamic characteristic knowledge component in the adjusted transfer channel during the transfer process, and update the aerodynamic difference features of the flight mechanics feature library according to the distribution offset. Iteratively execute the operations from the coupled simulation operation to the adjustment of the transmission status register until the distribution offset enters the steady-state operation interval.

2. The method according to claim 1, characterized in that According to the flight mechanics feature library, perform a coupled simulation operation to generate a set of compensation parameters, including: According to the aerodynamic difference features and flight state parameters of the flight mechanics feature library, calculate the dynamic fluid pressure of the source aircraft model and the target aircraft model under the same flight state parameters, and generate a solid deformation boundary constraint according to the dynamic fluid pressure; Synchronously stimulate the interaction process of the source aircraft model and the target aircraft model under the same flight state parameters, and obtain the fluid pressure distribution responses corresponding to the source aircraft model and the target aircraft model respectively; Transfer the fluid pressure distribution response to the action area of the solid deformation boundary constraint, calculate the response quantity difference value of the source aircraft model and the target aircraft model under the same fluid pressure distribution response, and generate a set of compensation parameters according to the corresponding relationship between the response quantity difference value and the flight state parameters.

3. The method according to claim 2, wherein Transfer the fluid pressure distribution response to the action area of the solid deformation boundary constraint, calculate the response quantity difference value of the source aircraft model and the target aircraft model under the same fluid pressure distribution response, and generate a set of compensation parameters, including: Transfer the fluid pressure distribution response to the action area of the solid deformation boundary constraint, obtain the solid deformation response quantities corresponding to the source aircraft model and the target aircraft model respectively under the action of the same fluid pressure distribution response, and calculate the response quantity difference value; Form a three-dimensional flight state coordinate according to the altitude layer identifier, speed interval code, and flight attitude angle of the flight state parameters; Associate the response quantity difference value with the storage location corresponding to the three-dimensional flight state coordinate, and connect the response quantity difference values of adjacent storage locations to construct multiple difference value change trajectories; Convert all difference value change trajectories into corresponding executable compensation instruction parameters, and combine all executable compensation instruction parameters to generate a set of compensation parameters.

4. The method according to claim 1, characterized in that, Establish a knowledge transfer channel from the pre-trained large model of the source aircraft model to the large model of the target aircraft model, including: According to the hardware resource distribution of the distributed computing node cluster, perform computing resource allocation and model loading processing on the pre-trained large model of the source aircraft model to obtain an airworthiness certification model instance; Initialize the target aircraft model according to its structural definition to obtain the model instance to be migrated; Configure the bidirectional data routing between the source aircraft model and the target aircraft model based on the airworthiness certification model instance and the model instance to be migrated to form a basic migration channel; Implant a feature extraction operator into the basic migration channel according to the aerodynamic characteristic knowledge migration requirement to generate a feature-enhanced migration channel; Synchronously load the encoding rule of the flight state parameters into the feature-enhanced migration channel to obtain a synchronized migration channel; Perform a security encapsulation process on the synchronized migration channel according to the aviation data security specification to obtain a knowledge migration channel.

5. The method according to claim 1, wherein Reconstruct the transmission buffer of the knowledge migration channel according to the compensation parameter set to obtain a target migration channel and generate a migration weight regulation function, including: Generate an aerodynamic difference compensation instruction parameter and a load spectrum offset coefficient according to the compensation parameter set; Obtain the original airfoil feature vector according to the aerodynamic characteristic knowledge component transmitted in the knowledge migration channel; Adjust the pressure distribution parameter of the original airfoil feature vector according to the pressure correction amount of the aerodynamic difference compensation instruction parameter, reconstruct the transmission buffer of the knowledge migration channel to obtain a reconstructed migration channel; Calculate the change gradient value according to the flight state parameters, and calculate the weight factor in combination with the load spectrum offset coefficient; Compile the initial migration weight regulation function, load the migration weight regulation function into the reconstructed migration channel to obtain a target migration channel; Start the rudder effect response validator of the target migration channel to obtain a verification result, and set the operation boundary value of the initial migration weight regulation function according to the verification result and the weight factor to obtain a migration weight regulation function.

6. The method according to claim 5, characterized in that Adjust the pressure distribution parameter of the original airfoil feature vector according to the pressure correction amount of the aerodynamic difference compensation instruction parameter, reconstruct the transmission buffer of the knowledge migration channel to obtain a reconstructed migration channel, including: Locate the pressure distribution parameter storage area where the original airfoil feature vector belongs and obtain the pressure distribution parameter; Obtain the pressure correction amount of the aerodynamic difference compensation instruction parameter, superimpose the pressure correction amount on the pressure distribution parameter to generate an updated pressure distribution parameter set; Add the updated pressure distribution parameter set to the pressure distribution parameter storage area, update the field of the original airfoil feature vector to obtain a reconstructed airfoil feature vector; Respond to the signal indicating the completion of the update of the original airfoil feature vector, and lock the write permission of the transmission buffer of the knowledge migration channel; In the locked state, write the reconstructed airfoil feature vector into the aerodynamic characteristic data segment of the transmission buffer, activate the data consistency check mechanism to check the transmission buffer, and reset the sequence number identifier of the transmission buffer according to the signal indicating successful verification to generate a reconstructed migration channel.

7. The method according to claim 1, characterized in that Use the migration weight regulation function to adjust the transmission status register of the target migration channel to obtain an adjusted migration channel, including: Interpret the executable code segment of the migration weight regulation function and separate the transmission intensity control parameter; Obtain the boundary constraint values of the rudder effect response validator, and calculate the transmission intensity correction amount in combination with the real-time change rate of the flight state parameters and the transmission intensity control parameters; Input the transmission intensity correction amount into the transmission scheduler of the target migration channel, trigger the bandwidth reallocation operation of the transmission scheduler, and obtain the updated bandwidth configuration; According to the updated bandwidth configuration, adjust the transmission status register of the target migration channel to obtain the adjusted migration channel.

8. A multi-aircraft model migration optimization system for an aviation large model, characterized in that, It includes: A construction module for constructing a flight mechanics feature library containing aerodynamic difference features based on the obtained flight mechanics parameters and manipulation response data of the source aircraft model and the target aircraft model; A simulation module for performing a coupled simulation operation according to the flight mechanics feature library to generate a set of compensation parameters; An establishment module for establishing a knowledge migration channel from the pre-trained source aircraft model large model to the target aircraft model large model; A reconstruction module for reconstructing the transmission buffer of the knowledge migration channel according to the set of compensation parameters to obtain the target migration channel and generate a migration weight regulation function; An adjustment module for using the migration weight regulation function to adjust the transmission status register of the target migration channel to obtain the adjusted migration channel; An update module for monitoring the distribution offset of the aerodynamic characteristic knowledge component in the adjusted migration channel during the migration process, and updating the aerodynamic difference characteristics of the flight mechanics feature library according to the distribution offset, and iteratively executing the operations from the coupled simulation operation to the adjustment of the transmission status register until the distribution offset enters the steady-state operation interval.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a multi-aircraft model migration optimization method for an aviation large model as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by the computer, it implements a multi-aircraft model migration optimization method for an aviation large model as described in any one of claims 1 to 7.

Images