Flight simulator simulation model correction method and device based on data driving

Abstract

The invention relates to the technical field of data driving, provides a flight simulator simulation model correction method and device based on data driving, and solves the problems of poor consistency between a simulation model state and real data and low dynamic correction efficiency. The method comprises the following steps: identifying a current composite flight state of a flight target according to flight parameters of the flight target, wherein the flight parameters comprise reverse driving data; determining a deviation degree according to the key parameter and the secondary key parameter corresponding to the composite flight state; selecting a corresponding target state matching method according to the deviation degree and the composite flight state, and performing state matching according to the target state matching method; and after successful matching, calling a controller of a corresponding channel according to the composite flight state or the current flight stage, and correcting the state of the simulation model corresponding to the flight target through an existing control mode. According to the method, the consistency between the simulation model state and the real data and the dynamic correction efficiency are improved.

Description

Technical Field

[0001] This application relates to the field of data-driven technology, and in particular to a data-driven method and apparatus for correcting flight simulator simulation models. Background Technology

[0002] In the real-time simulation and state reproduction of dynamic systems, there is a clear need for high-precision, high-consistency model correction techniques. Especially in application scenarios that require reverse-engineering from real operating data to reproduce specific operating states, the simulation model must be able to quickly and automatically track changes in the real state and adjust in real time when deviations occur to ensure the validity and reliability of the simulation results.

[0003] Currently, a common targeted solution is to use a state monitoring and matching method based on fixed thresholds. This method sets allowable deviation ranges for key operating parameters based on predefined operating phase divisions. When the difference between the simulation output and the reverse-drive data exceeds the set threshold, a preset balancing strategy based on that operating phase is triggered to reset or correct the parameters of the simulation model so that it returns to the expected state.

[0004] However, these approaches exhibit limitations in their response methods and adjustment strategies when dealing with complex operating states or rapid dynamic transitions. Due to their reliance on predefined stages and fixed thresholds, the adjustment process may lag when handling multi-dimensional coupling, unsteady, or continuously changing states. Furthermore, multiple corrections can easily trigger state oscillations, affecting the consistency and reproducibility of the simulation process. Additionally, there is room for improvement in their adaptability to disturbances in the operating environment and errors inherent in the model itself. Summary of the Invention

[0005] This application provides a data-driven method, apparatus, electronic device, and storage medium for correcting flight simulator simulation models, in order to solve the problems of poor consistency between the simulation model state and real data and low dynamic correction efficiency in the prior art.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a data-driven method for calibrating a flight simulator model, comprising:

[0007] Identify the current composite flight state of the flight target based on its flight parameters, including anti-drive data;

[0008] The degree of deviation is determined based on the key and secondary key parameters corresponding to the composite flight state;

[0009] Select the corresponding target state matching method based on the degree of deviation and the composite flight state, and perform state matching according to the target state matching method;

[0010] After a successful match, based on the composite flight state or the current flight stage, the controller of the corresponding channel is invoked to correct the state of the simulation model corresponding to the flight target through the existing control mode.

[0011] Optionally, the step of selecting the corresponding target state matching method based on the degree of deviation and the composite flight state, and performing state matching according to the target state matching method, includes:

[0012] Based on the degree of deviation, determine whether the composite flight state is a stable state;

[0013] When the composite flight state is a stable state, state matching is performed based on the flight parameter state and the first trim method corresponding to the stable state, depending on the different composite flight states. Alternatively, when the composite flight state is an unstable state, state matching is performed according to the second trim method corresponding to the unstable state.

[0014] Secondly, this application provides a data-driven flight simulator model calibration device, comprising:

[0015] The identification module is used to identify the current composite flight state of the flight target based on the flight parameters of the flight target, including anti-drive data;

[0016] The determination module is used to determine the degree of deviation based on the key parameters and secondary key parameters corresponding to the composite flight state;

[0017] The selection module is used to select the corresponding target state matching method based on the degree of deviation and the composite flight state, and to perform state matching according to the target state matching method.

[0018] The calling module is used to call the controller of the corresponding channel after a successful match, based on the composite flight state or the current flight stage, and correct the state of the simulation model corresponding to the flight target through the existing control mode.

[0019] Thirdly, this application provides an electronic device, comprising:

[0020] Memory, used to store computer programs;

[0021] A processor, configured to execute the computer program to implement the steps of the data-driven flight simulator simulation model correction method as described in the first aspect above.

[0022] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the data-driven flight simulator simulation model correction method described in the first aspect above.

[0023] The technical solution provided in this application has the following beneficial effects:

[0024] This application achieves refined and multi-dimensional identification of the target's operational state, providing accurate state data for subsequent targeted corrections. A hierarchical deviation discrimination mechanism is established to improve the accuracy of state assessment and the ability to distinguish the impact of different parameters. Intelligent adaptation and dynamic selection of correction strategies are implemented, enhancing the effectiveness and flexibility of the method under different states. A smooth transition and rapid convergence of the correction process are ensured, improving the overall system's response speed and the consistency of state reproduction.

[0025] Furthermore, based on the determined degree of deviation, this application first determines whether the composite operating state is stable or unstable. If it is stable, matching is performed by setting parameters and combining them with a first-type balancing method specifically designed for stable states, according to the specific state type and real-time parameters. If it is unstable, matching is performed using a second-type balancing method suitable for unsteady states, thereby achieving differentiated and precise correction under different states. Moreover, this method can adaptively select the most suitable balancing matching strategy for different state characteristics, such as stable and unstable states, effectively improving the accuracy and adaptability of state correction and ensuring the response quality and reproducibility consistency of the simulation model under various dynamic conditions.

[0026] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A flowchart illustrating a data-driven flight simulator simulation model calibration method provided in this application embodiment;

[0029] Figure 2 A schematic diagram of a real-time calibration technology module for a data-driven flight simulator calibration method provided in this application embodiment;

[0030] Figure 3A flowchart illustrating the target state matching process of a data-driven flight simulator simulation model correction method provided in this application embodiment;

[0031] Figure 4 This is a schematic diagram of a data-driven flight simulator model calibration device provided in an embodiment of this application. Detailed Implementation

[0032] In the field of dynamic system simulation and state reproduction, existing correction methods mostly rely on state monitoring and matching strategies based on fixed thresholds. While this method is simple in structure, it has significant shortcomings when dealing with complex, continuously changing, or rapidly transitioning operating states: the adjustment process is prone to lag, multiple corrections may trigger state oscillations, and its adaptability to model errors and external disturbances is limited, affecting the accuracy and continuity of simulation reproduction.

[0033] To address the aforementioned limitations, this application proposes a data-driven flight simulator model calibration method. Its core lies in identifying the target's complex operational states based on real-time parameters and adaptively selecting a matching strategy according to state characteristics and deviation levels. Specifically, this method distinguishes between stable and unstable states and employs targeted trimming and matching methods for each, achieving rapid and accurate calibration of the simulation state. This scheme effectively overcomes the insufficient adaptability and response lag of fixed threshold strategies, improves the tracking accuracy and reproducibility of the simulation model under complex dynamic operating conditions, and ensures the smoothness and reliability of the calibration process.

[0034] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] The core of this application is to provide a data-driven method for correcting flight simulator simulation models, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0036] Step 101: Identify the current composite flight state of the flight target based on its flight parameters, including anti-dash data.

[0037] In step 101, flight parameters refer to a set of data reflecting the motion and state of the flight target acquired in real time, mainly including position, velocity, attitude, acceleration, and anti-drive data from external systems. Anti-drive data refers to a data sequence obtained through reverse drive that reflects the external forces or control commands acting on the target in the actual operating environment, such as maneuver inputs and environmental disturbances. Composite flight state refers to a complex operating mode composed of multiple basic operating states (such as horizontal movement, ascent, descent, and turning) superimposed in time and space, and its identification depends on the collaborative analysis of multi-dimensional flight parameters.

[0038] In this embodiment, the system first collects various flight parameters of the target in real time, including position, speed, attitude angle, angular velocity, and anti-flight data. Then, using these parameters, the system performs a preliminary classification of the current operating state through a state recognition algorithm (such as a rule-based state machine or a lightweight temporal pattern recognition method), identifying basic states such as level flight, climb, descent, and turning. Next, based on the temporal continuity of these basic states and the coupling relationship between parameters, the system further integrates and judges them to form a description of a composite flight state, such as "climb and turn" or "accelerate descent". Finally, the system outputs the composite flight state result identified at the current moment, providing a state basis for subsequent steps.

[0039] For example, suppose a flight target A is performing a mission. The system acquires its altitude, speed, heading angle, and control commands from the anti-ship data in real time. Analysis reveals that the current altitude is continuously increasing, the speed is relatively stable, the heading angle is changing slowly, and there is significant pitch control input in the anti-ship data. Based on this, the system identifies the current state as "climbing." Simultaneously, the continuous change in heading, combined with rudder control input, indicates a "turning" behavior. The system ultimately merges these two indicators, determining the current composite flight state as "climb and turn." This identification result provides a clear state label for subsequent steps.

[0040] Step 102: Determine the degree of deviation based on the key parameters and secondary key parameters corresponding to the composite flight state.

[0041] In step 102, key parameters refer to the core operational indicators that have a decisive impact on the current complex flight state. For example, in the "climb and turn" state, the rate of change of altitude, speed, and roll angle can be considered key parameters. Secondary key parameters refer to other parameters that have an auxiliary impact on state maintenance or transition, such as angular velocity and acceleration. Deviation degree refers to the comprehensive measure of the differences between the simulation model output state and the actual flight target state across various parameter dimensions, used to quantify the degree of mismatch in the current simulation.

[0042] In this embodiment of the application, the system obtains the set of key parameters and secondary key parameters of the composite flight state identified in step 101 from the predefined configuration; calculates the values ​​of these parameters in the simulation model in real time and compares them with the corresponding parameters of the real flight target; calculates the deviation values ​​of the key parameters and secondary key parameters respectively according to the difference tolerance range set in advance for each type of parameter; finally, integrates these deviation values ​​and generates an overall deviation index by weighting or logical combination to characterize the overall difference level between the current simulation state and the real state.

[0043] For example, continuing from the previous example, after identifying the "climb and turn" state, the system retrieves the key parameters (altitude change rate, airspeed, roll angle) and secondary key parameters (pitch rate, yaw rate) for this state. Assume the real target's altitude change rate is 10 meters per second, airspeed is 100 meters per second, and roll angle is 15 degrees; while the simulation model outputs 8 meters per second, 105 meters per second, and 12 degrees, respectively. The system calculates the differences in each parameter and determines whether the key parameters deviate based on preset thresholds. If the difference in altitude change rate exceeds the threshold, the deviation is considered high, and correction needs to be initiated.

[0044] Step 103: Select the corresponding target state matching method based on the degree of deviation and the composite flight state, and perform state matching according to the target state matching method.

[0045] In step 103, the target state matching method refers to a set of specific strategies or algorithms used to adjust the state of the simulation model to match the actual target state, including parameter resetting, model trimming, and control law adjustment. The selection of this method depends on the current degree of deviation and the type of complex flight state.

[0046] In this embodiment, the system receives the deviation degree output in step 102 and the composite flight state result in step 101. If the deviation degree is low, a mild correction strategy is selected, such as fine-tuning the model parameters. If the deviation degree is high, a matching method is further selected based on the composite flight state: for stable states (such as level flight), a parameter trim method based on equilibrium conditions is adopted; for unstable states (such as maneuvering turns), a dynamic tracking or predictive control method is adopted. Finally, the selected matching method is executed to adjust the internal parameters or control input of the simulation model so that its output state approaches the real target state.

[0047] For example, continuing the previous example, the system determines that the current deviation is high and the composite flight state is "climb and turn" (an unstable state), so the dynamic tracking matching method is selected. This method generates control commands for the simulation model based on the target's altitude change rate, airspeed, and roll angle. The internal controller then adjusts the simulation model's pitch, thrust, and roll control surfaces, gradually bringing the output altitude change rate closer to 10 meters per second, airspeed closer to 100 meters per second, and roll angle closer to 15 degrees, thus completing the state matching.

[0048] Step 104: After successful matching, based on the composite flight state or the current flight stage, call the controller of the corresponding channel and correct the state of the simulation model corresponding to the flight target through the existing control mode.

[0049] In step 104, the channel controller refers to a control module independently designed for different motion dimensions of the simulation model (such as pitch, roll, yaw, thrust, etc.). The control mode refers to the specific adjustment law adopted by the controller, such as proportional-integral-derivative control, fuzzy control, model tracking control, etc., to ensure that the model state is stable and accurately tracks the target.

[0050] In this embodiment of the application, after state matching is completed in step 103, the system activates the corresponding channel controller (such as pitch controller, roll controller) according to the current composite flight state or flight phase (such as climb, cruise, landing); each controller continuously calculates the error between the simulation model state and the target state according to its built-in control mode, and generates control commands; these commands are applied to the simulation model to dynamically adjust its control surfaces or power output, so as to achieve fine maintenance and continuous correction of the model state and ensure that its output is consistent with the real flight target state.

[0051] For example, after matching is complete, the system confirms that it is still in a "climb and turn" state, so it calls the pitch and roll controllers. The pitch controller calculates the elevator deflection angle based on the altitude change rate error, and the roll controller calculates the aileron deflection angle based on the roll angle error. These control commands are applied to the simulation model in real time, keeping the altitude change rate stable at 10 meters per second and the roll angle stable at 15 degrees during the climb and turn, thus achieving continuous consistency between the simulation state and the real target state.

[0052] This method accurately determines the degree of deviation between the simulation model and the real state by identifying the complex operational states of the flight target in real time, and adaptively selects matching and correction strategies accordingly. Finally, it achieves continuous and stable tracking of the simulation state through multi-channel control. This method improves the response speed, state consistency, and adaptive capability of the simulation model in complex operating scenarios, effectively ensuring the accuracy and reliability of simulation reproduction.

[0053] To address the challenge of adaptively selecting the most suitable matching and correction method based on the degree of deviation between the simulated state and the real target, as well as the type of operational state, to improve the accuracy and efficiency of state reproduction, in some embodiments, step 103: selecting the corresponding target state matching method based on the degree of deviation and the composite flight state, and performing state matching according to the target state matching method, includes:

[0054] Step 201: Determine whether the composite flight state is a stable state based on the degree of deviation.

[0055] In step 201, a stable state refers to an operational condition in which the flight target's operating parameters change smoothly, the forces are balanced, and there is no significant acceleration or drastic attitude change, such as uniform horizontal flight or stable climb. The judgment process depends on the overall degree of deviation determined in step 102 and the time-varying characteristics of each key parameter in the current composite flight state.

[0056] In this embodiment, the system receives the deviation value and current composite flight state information from step 102. First, the system checks whether the deviation is lower than a threshold value set for a stable state. If it is lower, it initially indicates that the system may be in a stable state. Next, the system further analyzes the changes of various key parameters constituting the current composite flight state over a recent period, such as calculating their rate of change or fluctuation amplitude. If the changes of these parameters remain within a small range, the current composite flight state is ultimately determined to be a stable state; otherwise, it is determined to be an unstable state. This determination result will be directly used to guide the selection of the matching method in the subsequent step 202.

[0057] Step 202: When the composite flight state is a stable state, for different composite flight states, state matching is performed based on the flight parameter state by setting parameters and the first trim method corresponding to the stable state; or, when the composite flight state is an unstable state, state matching is performed according to the second trim method corresponding to the unstable state.

[0058] In step 202, the first balancing method is a set of adjustment strategies designed specifically for steady-state conditions. Its core is to enable the model to quickly enter and maintain a steady-state operation by directly setting the core parameters of the simulation model (such as position, attitude, and velocity) and supplementing them with balance calculations. The second balancing method is a strategy designed to deal with unstable states (such as acceleration and sharp turns). It focuses more on the dynamic tracking and compensation of the force or motion trend of the model, rather than pursuing instantaneous static balance.

[0059] In this embodiment, the system performs a branch operation based on the stable state judgment result made in step 201. If the state is determined to be stable, the system extracts target parameter values ​​from real data according to the current specific composite flight state (such as "level flight" or "stable climb"), and adopts the first trim method: first, these target parameter values ​​are directly assigned to the simulation model, and then the balance calculation program for the stable state is started to fine-tune the control variables inside the model so that the model output quickly converges and stabilizes at the target state. If the state is determined to be unstable in step 201, the second trim method is activated: the system no longer seeks static parameter settings, but calculates the differences in motion trends and forces between the simulation model and the real target in real time based on the current flight parameter state, and compensates for these differences by dynamically adjusting the control commands, so that the state change process of the simulation model is as close as possible to the real target.

[0060] Here is a specific example: Continuing from the previous embodiment, flight target A is continuously in a climb-turn state. After the system determines that the deviation is high and the composite flight state is unstable, it enters the target state matching method selection and execution process. The system first determines whether the composite flight state is stable based on the degree of deviation. Since the difference in altitude change rate exceeds a preset threshold and the parameters change drastically, it is determined to be an unstable state. Subsequently, the system performs state matching according to the second trim method corresponding to the unstable state. This method performs dynamic tracking compensation based on the current flight parameter state. The system acquires the real target A's altitude change rate of 10 meters per second, airspeed of 100 meters per second, and roll angle of 15 degrees as target values ​​in real time, and continuously collects the corresponding output values ​​of the simulation model: altitude change rate of 8 meters per second, airspeed of 105 meters per second, and roll angle of 12 degrees. The system calculates the instantaneous errors of each parameter, where the altitude change rate error is 2 meters per second, the airspeed error is 5 meters per second, and the roll angle error is 3 degrees. These error values ​​are input to a dynamic compensation controller, which generates control commands based on the error ratio. The core calculation is as follows: the control command adjustment equals the error value multiplied by a proportional coefficient, where the proportional coefficient is pre-set according to the parameter importance. For example, the altitude change rate proportional coefficient is 0.5, the airspeed proportional coefficient is 0.2, and the roll angle proportional coefficient is 0.3. Therefore, the altitude change rate control adjustment is 2 meters per second multiplied by 0.5 equals 1 meter per second; the airspeed control adjustment is 5 meters per second multiplied by 0.2 equals 1 meter per second; and the roll angle control adjustment is 3 degrees multiplied by 0.3 equals 0.9 degrees. Based on this, the system generates additional thrust control commands and pitch and roll control commands, applying them to the simulation model to dynamically adjust its internal parameters. After several control cycles, the altitude change rate output by the simulation model gradually approaches 10 meters per second, the airspeed approaches 100 meters per second, and the roll angle approaches 15 degrees, thus completing the matching under unstable conditions and laying the foundation for subsequent continuous correction by calling the channel controller.

[0061] In the embodiments of this application, the above complete steps intelligently determine the stability of the operating state and select distinctly different but highly targeted matching strategies accordingly. This ensures that the simulation model can obtain fast, accurate and stable correction results regardless of whether the operating conditions are stable or drastically changing, thereby enhancing the adaptability of the whole method to different operating scenarios and the overall correction efficiency.

[0062] To address the challenge of adaptively selecting and executing precise trim strategies based on the stability of the composite flight state to efficiently achieve state matching in the simulation model, in some embodiments, step 202 involves: when the composite flight state is stable, performing state matching based on flight parameter states and a first trim method corresponding to the stable state for different composite flight states; or, when the composite flight state is unstable, performing state matching according to a second trim method corresponding to the unstable state, including:

[0063] Step 301: When the composite flight state is a stable state, set the position parameters, attitude parameters and velocity parameters of the simulation model corresponding to the flight target, and use the position parameters, attitude parameters and velocity parameters as the initial conditions for state matching.

[0064] In step 301, position parameters refer to the coordinate data describing the specific location of the flight target in three-dimensional space. Attitude parameters refer to the angular data describing the flight target's orientation or tilt. Velocity parameters refer to the vector data describing the speed and direction of the flight target's motion. The initial conditions for state matching refer to a set of initial model state values ​​preset for subsequent trim operations; it is the starting point and basis for matching.

[0065] In this embodiment, when the system determines that the composite flight state is a stable state, it first extracts the position, attitude, and velocity data of the current real flight target from the real-time flight parameter state. Subsequently, the system directly sets these data as the corresponding parameter values ​​of the simulation model, thereby enabling the simulation model to instantly enter an initial operating state very close to the real target, preparing conditions for subsequent fine trim.

[0066] Step 302: Based on the flight parameter state and the initial conditions, select the first trim method that matches the composite flight state from a variety of preset trim methods.

[0067] In step 302, the preset multiple trim methods refer to a library of pre-designed model balance adjustment strategies to achieve different stable flight states.

[0068] In this embodiment, the system queries a preset trim strategy mapping table based on the current specific composite flight state type and pre-set initial conditions. This table defines the correspondence between different composite flight states and the optimal trim method. For example, one trim method is selected for level flight, and another for a stable climb. By looking up the table, the system automatically selects the first trim method that best matches the current state.

[0069] Step 303: Using the selected first balancing method, perform a balancing operation on the simulation model under the initial conditions to complete state matching.

[0070] In step 303, the balancing operation refers to applying the calculation rules and adjustment procedures defined by the selected balancing method to calculate and set the internal control variables of the simulation model in order to eliminate residual unbalanced forces or torques and stabilize the model output state at the target value.

[0071] In this embodiment, the system initiates the calculation program corresponding to the selected first trim mode. This program takes the currently set initial conditions as input and calculates the precise deflection angles or thrust magnitudes of the various control surfaces required by the simulation model to maintain this stable state. After the calculation is complete, the system assigns these control quantities to the simulation model, which then adjusts its internal state, reaching a stable equilibrium after a brief transition, thereby achieving precise matching with the actual flight target state.

[0072] Step 304: When the composite flight state is unstable, based on the flight parameter state, use a long short-term memory network to predict the dynamic change trend of the unstable axis system, and based on the dynamic change trend, determine the final force state under the corresponding axis system.

[0073] In step 304, an unstable axis system refers to a specific axis system that, under the current flight condition, experiences force imbalance or a rate of change in motion exceeding a stability threshold, and whose state cannot be stabilized through conventional trim operations. It does not specifically refer to a single fixed axis system among the longitudinal, lateral, or normal axes, but rather to any one or more axes systems identified as unstable under the current specific flight condition due to force imbalance or drastic changes in motion; its specific orientation depends on real-time flight conditions and dynamic characteristics. The corresponding axis system refers to the reference coordinate system determined based on the aircraft's attitude and direction of motion, derived from the analysis of the aircraft's force state and motion attitude in the air. Specifically, it includes the longitudinal, lateral, and normal axes systems, corresponding to the aircraft's pitch, roll, and yaw motion dimensions, respectively. The dynamic change trend refers to the predicted changes in the motion parameters of these axes over a short period in the future. The final force state refers to the target force state used for trim calculations, determined by comprehensively considering the current measured forces and predicted future trends.

[0074] In this embodiment, when the system is in an unstable state, it inputs the flight parameter states over a recent time series into a pre-trained Long Short-Term Memory (LSTM) network model. This network model analyzes the time dependencies of the parameters and outputs a predicted trend of the unstable axis motion changes in the near future. The system then combines this predicted trend with the currently measured flight parameter states, and through a data fusion algorithm, calculates a more accurate and forward-looking final force state on each corresponding axis, serving as the target basis for trim.

[0075] Step 305: Based on the final stress state, select a second balancing method that does not perform balancing operations on the unstable shaft system.

[0076] In this embodiment of the application, the system analyzes the final stress state obtained in step 304 and identifies which shaft system is unstable. Subsequently, the system selects a strategy from the library of preset balancing methods for unstable states. This strategy explicitly indicates that conventional balancing calculations should not be performed on the identified unstable shaft system, but rather a special adjustment method should be adopted to bypass or compensate for the instability. This method is then determined as the second balancing method to be used.

[0077] Step 306: Perform a balancing operation on the simulation model using the selected second balancing method to complete state matching.

[0078] In step 306, the balancing operation here refers to the execution of a specific calculation and adjustment process defined by the second balancing method. The process may differ from the balancing in a steady state, and it focuses more on dynamic tracking and compensation.

[0079] In this embodiment, the system operates the selected second balancing method. This method uses the previously determined final force state as the main input to calculate a set of control commands. These commands may include adjustment commands for other stable shaft systems, or directly generate compensation signals for dynamically tracking the changing trends of unstable shaft systems. The system applies these control commands to the simulation model, and the model adjusts its behavior according to the commands, thereby effectively converging the overall output state towards the true target state without directly forcing the unstable shaft system to balance, thus completing the matching.

[0080] Here is a specific example: Continuing from the previous embodiment, flight target A is continuously in an unstable composite flight state of climb and turn, and the system has completed preliminary matching based on dynamic compensation, making its altitude change rate approach 10 meters per second, airspeed approach 100 meters per second, and roll angle approach 15 degrees. To further achieve accurate state matching that adapts to future changes, the system executes a deep trim process under unstable conditions. The system first uses a long short-term memory network to predict the dynamic change trend of the unstable axis system based on the current flight parameter state. Specifically, it inputs recent time series data of altitude change rate, airspeed, and roll angle [8.2 meters per second, 99 meters per second, 14.8 degrees], [8.5 meters per second, 99.5 meters per second, 14.9 degrees], and [8.8 meters per second, 100 meters per second, 15 degrees] into the network. The network outputs a prediction that the roll axis system angular velocity will increase from the current 1.5 degrees per second to 2.1 degrees per second within the next 2 seconds, indicating that the roll axis system is unstable and has a tendency to intensify instability. Based on this dynamic trend, the system determines the final stress state of the corresponding shaft system by weighted fusion of the current measured force and the predicted trend, where the target torque value of the rolling shaft system is... From the formula The calculation shows that, among which Represents the final target torque, measured in Newton-meters. The weight representing the current measured value can be 0.6. This represents the current measured rolling torque, which can be taken as 150 Newton-meters. The weight representing the predicted value can be 0.4. The expected torque, calculated based on the predicted change in angular velocity, can be taken as 160 Nm. Substitute this value into the calculation: Based on this final stress state, the system selects a second trim method that does not directly trim the unstable roll axis system, i.e., a strategy that focuses on indirect compensation by coordinating the motion of other axes. Finally, the system uses the selected second trim method to trim the simulation model. This method calculates, based on the target moment value of 154 Nm for the roll axis system, that the aileron deflection angle of the simulation model needs to be increased by 2 degrees while the rudder deflection angle is decreased by 1 degree to generate a coordinated yaw moment for compensation, rather than directly adjusting the roll control surfaces. After these commands are applied to the model, the roll moment output is stabilized at around 154 Nm within several control cycles, thus achieving high-precision state matching without directly interfering with the unstable axis system, laying a solid foundation for subsequent control.

[0081] In the embodiments of this application, the above complete steps intelligently distinguish between stable and unstable states, and use different strategies combining static presets and dynamic predictions for precise balancing, thereby improving the adaptability and matching accuracy to complex and variable flight states, and ensuring that the simulation model can quickly and reliably reproduce the state of the real flight target under various operating conditions.

[0082] To address the challenge of accurately predicting the dynamic changes of an unstable axis system and determining the precise stress state accordingly, in some embodiments, step 304—predicting the dynamic change trend of the unstable axis system using a long short-term memory network based on the flight parameter state, and determining the final stress state of the corresponding axis system based on the dynamic change trend—includes:

[0083] Step 401: Extract the aerodynamic state and engine thrust state from the flight parameter state.

[0084] In step 401, the aerodynamic state refers to a comprehensive description of the various forces generated when a flight target moves through the air, mainly including lift, drag, and lateral forces. The engine thrust state refers to the magnitude and direction of the thrust generated by the flight target's engines. The extraction process separates these two specific data points from the overall flight parameter state.

[0085] In this embodiment, the system first receives real-time acquired flight parameter states, which is a dataset containing various parameters. Then, according to predetermined data classification rules, the system identifies and separates all aerodynamically related parameter data from this dataset, grouping them into aerodynamic states. Simultaneously, the system also identifies and separates all engine output-related parameter data, grouping them into engine thrust states. These two extracted state data sets will be used for subsequent sequence construction and analysis.

[0086] Step 402: Arrange the force vectors contained in the aerodynamic state and the engine thrust state in chronological order to form a parameter sequence.

[0087] In step 402, a force vector is an information representation that includes both the magnitude and direction of a force. A parameter sequence is an ordered list of force vector data from multiple consecutive time points arranged chronologically to reflect the change of force over time. A force vector refers to the force vector contained in each of these two states. The aerodynamic state includes aerodynamic vectors such as lift, drag, and side force, while the engine thrust state includes the thrust vector and its directional components. These force vectors collectively constitute the input for the dynamic analysis.

[0088] In this embodiment, the system acquires the aerodynamic state and engine thrust state obtained in step 401, both of which contain force vector information. The system extracts the aerodynamic vector data and engine thrust vector data for each sampling moment within a recent period, according to the data's timestamp. Then, strictly following the chronological order, these force vector data are arranged sequentially to form a continuous time parameter sequence. This sequence characterizes the recent changes in the resultant force acting on the flight target.

[0089] Step 403: Input the parameter sequence into the Long Short-Term Memory (LSTM) network, extract the time dependency from the parameter sequence through the LSM network, and output the dynamic change trend of the unstable axis system within a future preset time period based on the time dependency.

[0090] In step 403, the Long Short-Term Memory (LSTM) network is a special computational model capable of learning and remembering dependencies in long-term data series. In this application, temporal dependencies specifically refer to the changing patterns and mutual influences of aerodynamic forces and engine thrust force vectors at consecutive time points. This relationship reflects the temporal evolution characteristics of the aircraft's dynamic state and forms the basis for LTM network prediction.

[0091] In this embodiment, the system inputs the parameter sequence constructed in step 402 into a long short-term memory network model that has been pre-trained with a large amount of data. The analysis unit inside the network model processes each data point in the sequence step by step and learns the historical patterns and laws of force vector changes, i.e., extracts the time dependency. Then, the network uses this learned relationship to extrapolate the force situation over a preset period of time in the future and finally outputs a prediction result. This result describes how the unstable axis system will change in the future, such as the trend of its angular velocity or acceleration.

[0092] Step 404: According to the preset attitude decomposition method, the force vectors contained in the aerodynamic state and the engine thrust state are decomposed to the corresponding longitudinal axis system, transverse axis system and normal axis system to obtain the preliminary force state under each axis system.

[0093] In step 404, the preset attitude decomposition method is a set of predefined mathematical rules used to decompose a force vector in space into three mutually perpendicular reference directions. The longitudinal axis can be the direction pointing from the head to the tail of the flight target. The lateral axis can be the direction pointing from the left to the right of the flight target. The normal axis can be the direction pointing from the top to the bottom of the flight target. The initial force state refers to the magnitude of the force component in each axis direction obtained after decomposition.

[0094] In this embodiment, the system again utilizes the force vectors contained in the aerodynamic state and engine thrust state extracted in step 401. According to pre-defined decomposition rules, the system mathematically decomposes each force vector into three axial directions: longitudinal, lateral, and normal, calculating the component force in each axial direction. After the calculation, the system obtains the current force situation of each of the three axial directions, which is referred to as the preliminary force state under each axial direction. The specific decomposition process is as follows: The current aerodynamic state (including lift, drag, and lateral force components) and engine thrust state (including thrust magnitude and direction) are mapped to the longitudinal axis (calculating forces in the forward and backward directions, such as drag and thrust components), the lateral axis (calculating forces in the left and right directions, such as lateral force and thrust lateral components), and the normal axis (calculating forces in the up and down directions, such as lift and thrust vertical components) respectively, based on the aircraft's body coordinate system through vector projection and decomposition calculation. For example, if the aircraft is in a climb and turn state, and the engine thrust is forward along the longitudinal axis of the body, it is decomposed into the longitudinal axis as forward thrust. At the same time, the lift in the aerodynamic forces is decomposed into the normal axis, and the lateral force and thrust lateral components generated during the turn are decomposed into the lateral axis.

[0095] Step 405: Based on the dynamic change trend, correct the initial stress state under each shaft system to obtain the final stress state under each shaft system.

[0096] In step 405, correction refers to the process of adjusting and improving the original data based on new information.

[0097] In this embodiment, the system receives the predicted dynamic change trend information of the unstable shaft system from step 403, and the preliminary stress state of each shaft system obtained from step 404. The system first identifies which shaft system is predicted to be unstable. Then, for this specific shaft system, the system adjusts its preliminary stress state value according to the predicted change trend. For example, if the prediction shows that the force on the shaft system will increase, its stress value is increased accordingly. For other stable shaft systems, their preliminary stress state remains basically unchanged. After this adjustment, the system outputs a corrected set of final stress states for each shaft system that more closely reflects the actual future situation.

[0098] Here is a specific example: Following the aforementioned embodiment where flight target A is in an unstable climbing and turning state, and the system has predicted that the roll axis angular velocity will increase from 1.5 degrees per second to 2.1 degrees per second within the next 2 seconds, the system executes the following procedure to accurately determine the final force state. The system first extracts the aerodynamic state from the current flight parameter state, including lift of 12000 N, drag of 2000 N, and side force of 300 N, and the engine thrust state, including thrust of 15000 N in the positive direction along the longitudinal axis of the aircraft. The system then arranges the force vectors contained in the aerodynamic and engine thrust states in chronological order to form a parameter sequence. This sequence contains the resultant force vector data for the three most recent moments: [thrust 14800 N, lift 11900 N, drag 1950 N, side force 280 N], [thrust 14900 N, lift 11950 N, drag 1980 N, side force 290 N], and [thrust 15000 N, lift 12000 N, drag 2000 N, side force 300 N]. The system inputs this parameter sequence into a long short-term memory network. The network analyzes the time dependencies in the sequence and outputs confirmation of the dynamic change trend of the unstable shaft system, i.e., the roll shaft system, predicting that its angular velocity change trend is consistent with the aforementioned trend. Next, the system decomposes the force vectors contained in the current aerodynamic and engine thrust states to the corresponding longitudinal, lateral, and normal shaft systems according to a preset attitude decomposition method. The initial force state of the longitudinal shaft system is 15000 N of thrust minus 2000 N of drag, equaling 13000 N. The initial force state of the normal shaft system is 12000 N of lift, and the initial force state of the lateral shaft system is 300 N of side force. Finally, the system corrects the initial force states of each shaft system based on the dynamic trend of the increasing angular velocity of the roll shaft system. The normal shaft system, with its strong coupling to the roll motion, requires correction. The correction formula is as follows: ,in Represents the final force on the normal axis system, measured in Newtons. The initial force on the representative normal axis system is 12000 Newtons. This represents a coupling coefficient of 0.02 units per degree per second. This represents the predicted change in angular velocity, which is 2.1 - 1.5 = 0.6 degrees per second. Substituting into the calculation, we get: The forces on the longitudinal and transverse shafts remain unchanged, resulting in the final force state of each shaft system as follows: 13000 N for the longitudinal shaft, 12144 N for the normal shaft, and 300 N for the transverse shaft. This result provides accurate force input for accurately locating instability effects and subsequent balancing.

[0099] In the embodiments of this application, the above complete steps form a process that can predict the force conditions of each shaft system in a forward-looking and accurate manner by extracting force information from the original data, constructing time series, using intelligent networks to predict changes, performing scientific force decomposition, and finally completing data correction. This greatly improves the accuracy and reliability of matching and correcting unstable states.

[0100] To address the challenge of accurately invoking control channels and effectively correcting simulation models based on flight status to achieve high-precision state tracking, in some embodiments, step 104: invoking the controller of the corresponding channel based on the composite flight status or the current flight phase, and correcting the state of the simulation model corresponding to the flight target using existing control modes, includes:

[0101] Step 501: Based on the composite flight state or the current flight phase, select at least one controller to be activated from the pitch channel controller, roll channel controller, and yaw channel controller, and designate the selected controller as the target controller.

[0102] In step 501, the pitch channel controller is a control module specifically responsible for regulating the pitch attitude of the simulation model, i.e., the up-and-down swaying motion of the nose. The roll channel controller is a control module specifically responsible for regulating the roll attitude of the simulation model, i.e., the left-and-right tilting motion of the fuselage. The yaw channel controller is a control module specifically responsible for regulating the yaw attitude of the simulation model, i.e., the left-and-right turning motion of the nose. The target controller refers to one or more of the aforementioned specific controllers that need to be activated to perform correction tasks based on the current state.

[0103] In this embodiment, the system first receives information about the currently identified composite flight state or flight phase. Then, the system queries a preset rule mapping table to find the control channels that need to participate in the operation under that state or phase. For example, a climb requires a pitch channel controller, and a turn requires roll and yaw channel controllers. Based on the query results, the system activates the corresponding pitch channel controller, roll channel controller, or yaw channel controller from the available controller library and marks it as the target controller to be used.

[0104] Step 502: Configure a control mode for the target controller, the control mode including steady-state control for maintaining a stable state or tracking control for tracking changes in corresponding parameters in the reverse drive data.

[0105] In step 502, the control mode refers to the specific adjustment rules and strategies followed by the controller during operation. Steady-state control is a control mode whose goal is to eliminate errors and maintain the controlled parameter at a fixed value. Tracking control is another control mode whose goal is to enable the controlled parameter to follow changes in an external input signal in real time.

[0106] In this embodiment, the system assigns a suitable control mode to each activated target controller. The system then determines which mode to use based on the current composite flight state or flight phase. If the state requires maintaining a stable flight attitude, a steady-state control mode is configured for the controller. If the state requires closely tracking changes in the anti-flight data, a tracking control mode is configured for the controller. The configuration process involves setting the controller's internal control laws and parameters to operate according to the selected mode.

[0107] Step 503: Based on the reverse drive data and the current state of the simulation model, drive the target controller with the configured control mode to generate corresponding manipulation commands.

[0108] In step 503, "drive" refers to providing input data to the controller and initiating its calculation process. "Manipulation command" refers to the control signal output by the controller after calculation, used to directly manipulate the control surfaces of the simulation model, such as the rudder deflection angle or throttle position.

[0109] In this embodiment, the system extracts parameters related to the target controller from the anti-drive data, such as pitch angle data for the pitch channel. Simultaneously, the system obtains real-time status feedback from the simulation model. Then, the system inputs both sets of data into the target controller, which has been configured with the control mode. The controller, based on its internal algorithm, compares the target value (from the anti-drive data) with the current value (from the model feedback), performs calculations, and ultimately generates specific control commands, such as the elevator deflection degree.

[0110] Step 504: Based on the control command, correct the attitude, trajectory, or velocity state of the simulation model.

[0111] In step 504, correction refers to applying control commands to the simulation model to change its internal state, thereby reducing the difference between the model and the target state.

[0112] In this embodiment, the system sends all the control commands generated in step 503 to the simulation model. Upon receiving these commands, the simulation model translates them into actual actions on the corresponding control surfaces, such as adjusting the control surface angle or changing the engine thrust. These actions alter the forces and torques acting on the model, leading to corresponding changes in the model's attitude, flight path, or speed. Through this continuous application of commands and adjustment of state, the output of the simulation model is constantly corrected, ultimately achieving a high degree of consistency with the real flight target state described by the anti-drone data.

[0113] Here is a specific example: Following the aforementioned implementation where flight target A is in a climb-turn state and the simulation model, after matching, shows an altitude change rate approaching 10 meters per second and a roll angle approaching 15 degrees, the system executes a controller call procedure for continuous and accurate correction. First, based on the current complex flight state of climb-turn, the system selects the pitch and roll channel controllers to be activated as target controllers from the pitch, roll, and yaw channel controllers, as this state requires simultaneous control of pitch and roll attitudes. Next, the system configures control modes for these two target controllers. Since it needs to continuously track changes in altitude change rate and roll angle in the anti-flight data, a tracking control mode is configured for both. Subsequently, based on the target altitude change rate of 10 meters per second and target roll angle of 15 degrees extracted from the anti-drone data, and combined with the current output altitude change rate of 9.8 meters per second and roll angle of 14.7 meters from the simulation model, the system drives the pitch channel controller configured with tracking control mode. This controller calculates the altitude change rate error to be 0.2 meters per second and generates an elevator deflection adjustment command of 0.1 degrees based on its internal scaling factor of 0.5. Simultaneously, it drives the roll channel controller to calculate the roll angle error to be 0.3 degrees and generates an aileron deflection adjustment command of 0.12 degrees based on its scaling factor of 0.4. Finally, the system corrects the simulation model based on these control commands, increasing the elevator deflection by 0.1 degrees and the aileron deflection by 0.12 degrees. These adjustments further stabilize the altitude change rate at 10 meters per second and the roll angle at 15 degrees, thereby achieving continuous high-precision correction of the simulation model's attitude and ensuring complete consistency with the real target state.

[0114] In the embodiments of this application, the above complete steps form an efficient and reliable state correction closed loop by intelligently selecting control channels, flexibly configuring control modes, accurately generating control commands and effectively applying them to the simulation model, thereby improving the tracking accuracy and response speed of the simulation model for real targets under various complex flight conditions.

[0115] To address the challenge of accurately identifying the current operational status of a flight target from multi-source data, in some embodiments, step 101: identifying the current composite flight status of the flight target based on its flight parameters includes:

[0116] Step 601: Based on the EICAS information and anti-drive data in the flight parameters, and according to the existing flight status system, identify the longitudinal flight status corresponding to the current flight phase. Based on the yaw rate and roll angle in the flight parameters, and combined with the control commands in the anti-drive data, identify the lateral flight status.

[0117] In step 601, the Engine Indicating and Crew Alerting System (EICAS) information refers to a series of data about the aircraft system and power system status obtained from the EICAS. The existing flight status system refers to a predefined standard flight phase classification framework. Longitudinal flight status refers to the type describing the aircraft's motion in the vertical plane. Yaw rate refers to the angular velocity at which the aircraft's nose turns left or right. Roll angle refers to the angle at which the aircraft's fuselage tilts around its longitudinal axis. Control commands refer to control signals issued by the pilot or autopilot system. Lateral flight status refers to the type describing the aircraft's lateral motion in the horizontal plane.

[0118] In this embodiment, the system first extracts EICAS information, such as engine thrust parameters and flap positions, from flight parameters, and simultaneously acquires control inputs from the anti-drone data. The system compares this data with standards in the existing flight state system to determine the current flight stage and the corresponding longitudinal flight state, such as climb or descent. Simultaneously, the system reads yaw rate and roll angle data from the flight parameters and, combined with rudder and aileron control commands from the anti-drone data, analyzes lateral motion characteristics to determine the lateral flight state, such as straight flight or turning. Through these two parallel analysis processes, the system obtains longitudinal and lateral state identification results, respectively.

[0119] Step 602: Combine the longitudinal flight state and the lateral flight state into a composite flight state.

[0120] In this embodiment, the system takes the longitudinal and lateral flight states identified in step 601 as input and fuses them according to preset state combination rules. The system queries the state combination mapping table to find the corresponding composite flight state type, for example, combining the longitudinal climb state and the lateral turn state into a composite flight state of climb-turn. Finally, the system outputs the combined composite flight state result, providing complete state information for subsequent processing.

[0121] Here is a specific example: Suppose a flight target A is performing a mission. The system acquires its flight parameters in real time, including engine indications and crew warning system information showing engine thrust at 85% and flaps retracted. The anti-drive data contains clear throttle push and stick control commands. Simultaneously, the flight parameters show altitude increasing continuously at 5 m / s, airspeed maintained at 100 m / s, and heading angle changing steadily at 0.5 degrees / s. Based on the existing flight state system, the system identifies the current flight phase as a climb phase, corresponding to a climb in the longitudinal flight state. Simultaneously, based on the yaw rate of 0.5 degrees / s and roll angle of 15 degrees in the flight parameters, combined with the rudder push and aileron control commands in the anti-drive data, the system identifies the lateral flight state as a stable turn. Finally, the system combines the longitudinal climb and lateral stable turn into a composite climb-turn flight state. This identification result provides a clear state label for subsequent steps.

[0122] In the embodiments of this application, the above-described complete step scheme achieves a comprehensive and accurate identification of the current operating state of the flight target by analyzing the longitudinal and lateral motion characteristics separately and intelligently integrating the two, providing a reliable state basis for subsequent deviation judgment and correction, and improving the pertinence and effectiveness of the entire correction process.

[0123] To address the challenge of accurately quantifying the state differences between simulation models and real flight targets, in some embodiments, step 102, determining the degree of deviation based on the key and secondary key parameters corresponding to the composite flight state, includes:

[0124] Step 701: Obtain the key parameters and secondary key parameters corresponding to the composite flight state from the preset mapping table of flight states and parameters.

[0125] In step 701, the preset mapping table between flight states and parameters is a predefined database or list that specifies which parameters are the most important key parameters and which are the secondary key parameters for each composite flight state.

[0126] In this embodiment, the system first receives the composite flight state result identified in step 101. Then, the system queries a preset mapping table to find the entry corresponding to the composite flight state. From the found entry, the system reads and obtains the specific names or identifiers of all key and secondary key parameters defined for this state. This parameter information will be used for subsequent deviation calculations and judgments.

[0127] Step 702: Set thresholds for key parameters and secondary key parameters respectively, and determine the degree of deviation according to the thresholds.

[0128] In step 702, the threshold is a threshold value set for each parameter to allow a range of differences.

[0129] In this embodiment, the system sets a corresponding threshold for each key parameter and secondary key parameter obtained in step 701. These thresholds are typically set based on the importance and sensitivity of the parameters, with key parameters usually having more stringent thresholds. Then, the system calculates the difference between the simulation model output value and the corresponding parameter value of the actual flight target. The system compares these differences with their respective set thresholds. Based on the comparison results, the system comprehensively judges the overall deviation of the current state, which can be defined as different levels such as no deviation, slight deviation, or deviation, providing a basis for decision-making regarding whether and how to initiate corrections.

[0130] Here is a specific example: Continuing from the aforementioned embodiment where flight target A is in a climb-turn state and the system has completed state identification, the system executes the following process to determine the degree of deviation. The system first queries the preset mapping table between flight states and parameters to obtain the parameters corresponding to the climb-turn state. Key parameters include altitude change rate, airspeed, and roll angle; secondary key parameters include pitch rate and yaw rate. The system then sets corresponding thresholds for these parameters. The thresholds for the key parameters are set to altitude change rate ±2 m / s, airspeed ±4 m / s, and roll angle ±3 degrees; the thresholds for the secondary key parameters are set to pitch rate ±1 degree / s and yaw rate ±1.5 degrees / s. The system calculates the difference between the simulation model output value and the actual target value, obtaining a difference of 2 m / s in altitude change rate, 5 m / s in airspeed, 3 degrees in roll angle, 0.8 degrees / s in pitch rate, and 1 degree / s in yaw rate. These discrepancies are compared to set thresholds, including altitude change rate difference equal to the threshold, airspeed difference exceeding the threshold, roll angle difference equal to the threshold, pitch rate difference within the threshold, and yaw rate difference within the threshold. Based on the comparison results, the system determines that two of the key parameters have reached or exceeded the thresholds, thus indicating a high degree of deviation and requiring immediate initiation of a correction procedure. This determination provides a clear decision-making basis for subsequently selecting a dynamic tracking and matching method.

[0131] In this embodiment, the above complete steps achieve a rapid, objective, and quantitative assessment of the degree of state deviation by intelligently querying parameter mapping relationships and comparing differences based on thresholds. This provides accurate and reliable input for selecting the correct correction strategy, effectively improving the efficiency and accuracy of the state matching process.

[0132] This application also provides a flight data-based anti-flight simulator reproduction system based on real-time correction of a simulation model, such as... Figure 2As shown, the system includes the following modules: a flight phase / state identification module, a state deviation discrimination module, a target state matching module, and a rapid tracking calibration module. The flight phase / state identification module identifies and classifies the current flight state based on real-time acquired aircraft parameters. The state deviation discrimination module defines key and secondary key parameters for different flight states and sets thresholds to judge the degree of deviation. The target state matching module selects different target state matching methods based on the degree of deviation and the flight state, and sets overdrive activation / suppression settings. The rapid tracking calibration module quickly calibrates the simulation model state after trimming to ensure consistency with the real aircraft state.

[0133] First, the flight phase / state recognition module divides the flight state into two dimensions: longitudinal and lateral. These two dimensions can be combined into multiple composite flight states. The current aircraft state is identified and classified based on real-time acquired aircraft parameters. For different flight states, the state deviation judgment module defines different key and secondary key parameters and sets different thresholds for each parameter to determine the degree of deviation. Based on the degree of deviation of the key and secondary key parameters, it determines whether to perform target state matching, and then selects different target state matching methods based on the deviation parameters and the flight state. After target state matching, long-term data back-drive is performed to check for state deviation, and then a decision is made regarding whether to perform rapid tracking control. Simultaneously, the matching method can be selected through three channels of overdrive.

[0134] Regarding flight phase / state identification, the following explanation is provided: Based on engine indication, EICAS information from the crew alarm system, and anti-flight data, and according to the longitudinal flight status, the current flight phase is identified (e.g., power-on, taxiing out, takeoff 1, takeoff 2, takeoff 3-ground, takeoff 3-climb, cruise, approach, landing, taxiing in, engine shutdown, etc., eleven phases in total). The criteria for flight phase division are shown in Table 1:

[0135] Table 1. Examples of EICAS Information Flight Phase Definitions

[0136] Regarding the EICAS definition above, the phase divisions differ for some flight phases within the overall flight mission profile. However, the main purpose of this module is to perform preliminary classification for state deviation judgment and target state matching. Therefore, some phases can be merged and split, dividing the aircraft state into six states: taxiing, takeoff, climb, level flight, descent, and landing. There is no need for matching states during the power-on and power-off phases, so no identification processing is performed. Details are shown in Table 2.

[0137] Table 2 Comparison of EICAS Information Flight Phase Definitions and Flight Status Matching

[0138] Furthermore, the aforementioned stages are primarily divided according to the longitudinal and pitch states of flight. However, special lateral states during flight also need to be identified to select different target matching methods, such as turns. For these states, flight parameters must be used for identification. The identification of turning and non-turning states can be aided by yaw rate and roll angle, assisted by control commands. The identification of these states is independent of the longitudinal flight profile and can be combined with longitudinal flight states to form composite flight states. These can be categorized into three conditions: turning / non-turning, single-engine / twin-engine, and no or small sideslip / large sideslip, forming composite flight states (such as ground taxiing turns, climb turns, level flight turns, descent turns, single-engine large sideslip climb turns, etc.).

[0139] Regarding the determination of the degree of deviation from the actual flight state, the following explanations are provided: General judgment criteria: For different flight states, key parameters (airspeed, altitude, attitude angles, flight phase) and secondary key parameters (pitch rate, roll rate, yaw rate, acceleration) are defined. Non-key parameters: Other aircraft information. Thresholds are set for key and secondary key parameters to determine the degree of deviation between the simulation model state and the actual aircraft state. Specific details are shown in Table 3.

[0140] Table 3 Design of key and secondary key parameters for each flight phase

[0141] The threshold values ​​for key parameters include consistent requirements across all flight phases: airspeed ±3 knots, altitude ±100 feet, and attitude angle ±2°. The threshold values ​​for secondary key parameters include pitch rate, roll rate, yaw rate, and acceleration in SI units of 1 × 10⁻². Details are shown in Table 4.

[0142] Table 4. Setting of Thresholds for Key Parameters

[0143] Regarding target state matching, the following explanations are provided: Target state matching can be configured with over-control activation and suppression settings. Over-control activation is used when deviations occur in the identification and deviation judgment during the flight phase. Users can manually select a specific matching method to activate target state matching. Over-control suppression is used when performing verification matching for specific subjects. If the model's realism differs significantly from the real aircraft, over-tolerance situations may occur during the subject's actions. In this case, to reflect the model's true state, target state matching should not be performed. Therefore, the target state matching function can be suppressed during the subject's actions. When over-control is activated and key parameters exceed thresholds, target state matching is initiated. Based on flight phase / state identification, it can be determined whether the aircraft is in a stable state. If it is stable, stable state matching is used, and the trim method in Table 5 is selected based on the flight phase / state identification; otherwise, unstable state matching is used. Stable state matching: By setting parameters such as the simulation model's position, attitude, and velocity, and performing trim (such as level flight trim, constant climb angle trim), target matching in a stable state is achieved. Typically, target state matching is considered under relatively stable conditions. Steady-state matching is achieved through parameter setting and balancing. The position, attitude, velocity, and other parameters of the simulation model are set sequentially, and then balancing is performed under this state to achieve stable flight under the target state. The specific process is as follows: Figure 3 As shown, the trim phase offers a variety of trim methods to choose from for different flight states. Different trim methods are selected based on the specific flight parameters under different flight states. These include rapid trim methods for six degrees of freedom, such as level flight trim, constant climb angle trim, constant thrust trim, and single-engine failure trim. In addition to the aforementioned rapid six-degree-of-freedom trim, single or multiple degree-of-freedom combined trims can also be performed based on the current aircraft state. Independent trim methods are provided for the six degrees of freedom: longitudinal, lateral, normal, pitch, roll, and yaw. Different combinations of these six degrees of freedom can be used to meet the different flight states of the aircraft.

[0144] Table 5 Balancing Methods

[0145] Unstable state matching: A combined balancing method is adopted. Unstable shaft systems are not balanced. Based on the aerodynamic and engine thrust states, the system is decomposed according to the set attitude to reflect the actual force state of the corresponding shaft system (such as climbing with acceleration).

[0146] For rapid tracking calibration, the following explanation is provided: After trimming, based on the flight phase / state identification results, the corresponding controllers (such as roll, pitch, and yaw channels) are invoked to quickly calibrate the simulation model state. Control modes include steady-state control and tracking control. Steady-state control means controlling the tracking parameters to 0, suitable for maintaining lateral stability. Tracking control means tracking the anti-drag data target, and controlling the tracking parameters to anti-drag data values ​​during control, suitable for performance-related subjects such as takeoff and climb. Controller framework: Pitch channel (elevator / thrust): Control parameters include: pitch angle, pitch rate, pitch acceleration, barometric altitude (radio altitude), rate of climb, normal acceleration, corrected airspeed, and longitudinal acceleration. Control modes: Steady-state control, tracking control. Roll channel (aileron): Control parameters: roll angle, roll rate, roll acceleration, and yaw angle. Control modes: Steady-state control, tracking control. Yaw channel (rudder): Control parameters: yaw angle, yaw rate, yaw acceleration, runway lateral distance. Control modes: steady-state control, tracking control. Details are shown in Tables 6 and 7.

[0147] Table 6 Fast Calibration Tracking Controller Framework

[0148] Table 7. Schematic diagram of the relationship between rapid calibration tracking parameters and flight status.

[0149] In this embodiment, the simulator continuously tracks the flight status of the real aircraft, ensuring consistency between the flight parameters reproduced by the simulator and the anti-flight data, thereby improving the accuracy of accident investigation and simulator reproduction. Aircraft status recognition is employed, enabling rapid target trim matching. Continuous tracking control after trim is also considered, employing rapid process tracking control to ensure the continuity of the anti-flight simulator process.

[0150] Figure 4 A schematic diagram of a data-driven flight simulator model calibration device provided in this application embodiment is shown below. The specific implementation section describes the following:

[0151] The identification module 41 is used to identify the current composite flight state of the flight target based on the flight parameters of the flight target, including anti-drive data.

[0152] The determination module 42 is used to determine the degree of deviation based on the key parameters and secondary key parameters corresponding to the composite flight state.

[0153] Selection module 43 is used to select the corresponding target state matching method according to the degree of deviation and the composite flight state, and to perform state matching according to the target state matching method.

[0154] The calling module 44 is used to call the controller of the corresponding channel according to the composite flight state or the current flight stage after a successful match, and correct the state of the simulation model corresponding to the flight target through the existing control mode.

[0155] The data-driven flight simulator model calibration device of this application embodiment is used to implement the aforementioned data-driven flight simulator model calibration method. Therefore, the specific implementation of the data-driven flight simulator model calibration device can be found in the embodiment section of the data-driven flight simulator model calibration method above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0156] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described data-driven flight simulator simulation model correction methods.

[0157] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described data-driven flight simulator simulation model correction methods.

[0158] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0159] The embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the data-driven flight simulator simulation model correction method.

[0160] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0161] The foregoing has provided a detailed description of a data-driven flight simulator simulation model correction method, apparatus, electronic device, and storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A data-driven method for calibrating a flight simulator model, characterized in that, include: Identify the current composite flight state of the flight target based on its flight parameters, including anti-drive data; The degree of deviation is determined based on the key and secondary key parameters corresponding to the composite flight state; Select the corresponding target state matching method based on the degree of deviation and the composite flight state, and perform state matching according to the target state matching method; After a successful match, based on the composite flight state or the current flight stage, the controller of the corresponding channel is invoked to correct the state of the simulation model corresponding to the flight target through the existing control mode.

2. The data-driven flight simulator simulation model calibration method according to claim 1, characterized in that, The step of selecting the corresponding target state matching method based on the degree of deviation and the composite flight state, and performing state matching according to the target state matching method, includes: Based on the degree of deviation, determine whether the composite flight state is a stable state; When the composite flight state is a stable state, state matching is performed based on the flight parameter state and the first trim method corresponding to the stable state, depending on the different composite flight states. Alternatively, when the composite flight state is an unstable state, state matching is performed according to the second trim method corresponding to the unstable state.

3. The data-driven flight simulator simulation model calibration method according to claim 2, characterized in that, When the composite flight state is stable, state matching is performed based on the flight parameter state and the first trim method corresponding to the stable state for different composite flight states; or, when the composite flight state is unstable, state matching is performed according to the second trim method corresponding to the unstable state, including: When the composite flight state is a stable state, the position parameters, attitude parameters, and velocity parameters of the simulation model corresponding to the flight target are set, and the position parameters, attitude parameters, and velocity parameters are used as the initial conditions for state matching. Based on the flight parameter state and the initial conditions, select the first trim method that matches the composite flight state from a variety of preset trim methods; Using the selected first balancing method, the simulation model under the initial conditions is balanced to complete state matching; When the composite flight state is unstable, based on the flight parameter state, the dynamic change trend of the unstable axis system is predicted using a long short-term memory network, and based on the dynamic change trend, the final force state under the corresponding axis system is determined. Based on the final stress state, a second balancing method that does not require balancing of the unstable shaft system is selected. The simulation model is balanced using the selected second balancing method to complete state matching.

4. The data-driven flight simulator simulation model calibration method according to claim 3, characterized in that, The process of predicting the dynamic change trend of the unstable axis system based on the flight parameter state using a long short-term memory network, and determining the final force state under the corresponding axis system based on the dynamic change trend, includes: Extract the aerodynamic state and engine thrust state from the flight parameter states; The force vectors contained in the aerodynamic state and the engine thrust state are arranged in chronological order to form a parameter sequence; The parameter sequence is input into a long short-term memory network. The network extracts time dependencies from the parameter sequence and outputs the dynamic change trend of the unstable axis system within a future preset time period based on the time dependencies. According to the preset attitude decomposition method, the force vectors contained in the aerodynamic state and the engine thrust state are decomposed to the corresponding longitudinal axis system, transverse axis system and normal axis system to obtain the preliminary force state under each axis system. Based on the aforementioned dynamic change trend, the initial stress state under each shaft system is corrected to obtain the final stress state under each shaft system.

5. The data-driven flight simulator simulation model calibration method according to claim 1, characterized in that, The step of invoking the controller of the corresponding channel according to the composite flight state or the current flight stage, and correcting the state of the simulation model corresponding to the flight target through the existing control mode, includes: Based on the composite flight state or the current flight phase, select at least one controller to be activated from the pitch channel controller, roll channel controller, and yaw channel controller, and designate the selected controller as the target controller. Configure a control mode for the target controller, the control mode including steady-state control for maintaining a stable state or tracking control for tracking changes in corresponding parameters in the reverse drive data; Based on the anti-drive data and combined with the current state of the simulation model, the target controller with the configured control mode is driven to generate corresponding manipulation commands. Based on the control commands, the attitude, trajectory, or velocity state of the simulation model is corrected.

6. The data-driven flight simulator simulation model calibration method according to claim 1, characterized in that, The process of identifying the current composite flight state of a flight target based on its flight parameters includes: Based on the EICAS information and anti-flight data in the flight parameters, and according to the existing flight status system, the longitudinal flight status corresponding to the current flight phase is identified. Based on the yaw rate and roll angle in the flight parameters, and combined with the control commands in the anti-flight data, the lateral flight status is identified. The longitudinal flight state and the lateral flight state are combined into a composite flight state.

7. The data-driven flight simulator model calibration method according to claim 1, characterized in that, The determination of the degree of deviation based on the key parameters and secondary key parameters corresponding to the composite flight state includes: Obtain the key parameters and secondary key parameters corresponding to the composite flight state from the preset mapping table of flight states and parameters; Set thresholds for key parameters and secondary key parameters respectively, and determine the degree of deviation according to the thresholds.

8. A data-driven flight simulator model calibration device, characterized in that, include: The identification module is used to identify the current composite flight state of the flight target based on the flight parameters of the flight target, including anti-drive data; The determination module is used to determine the degree of deviation based on the key parameters and secondary key parameters corresponding to the composite flight state; The selection module is used to select the corresponding target state matching method based on the degree of deviation and the composite flight state, and to perform state matching according to the target state matching method. The calling module is used to call the controller of the corresponding channel according to the composite flight state or the current flight stage after a successful match, and correct the state of the simulation model corresponding to the flight target through the existing control mode.

Images