Neural network-based eVTOL multi-field coupling stability analysis method and system

Abstract

The application relates to the technical field of aircraft stability analysis, and provides an eVTOL multi-field coupling stability analysis method and system based on a neural network, which comprises the following steps: collecting eVTOL distributed electric propulsion system operation data under a plateau low-temperature environment, adopting a three-field coupling modeling algorithm to construct an aerodynamic-thermal-force coupling model, performing stability prediction modeling based on a physical information constraint neural network, performing multi-field coupling stability analysis by using Lyapunov stability theory, generating a control strategy according to an analysis result, and sending a control instruction through a flight control system. The application realizes coupling prediction and stability control of aerodynamic-thermal-force multi-field effects of the eVTOL distributed electric propulsion system under the plateau low-temperature environment, and improves the accuracy of stability judgment and flight safety under a complex environment.

Description

Technical Field

[0001] This invention relates to the field of aircraft stability analysis technology, and in particular to a method and system for eVTOL multi-field coupling stability analysis based on neural networks. Background Technology

[0002] The neural network-based eVTOL multi-field coupling stability analysis technology is a technology that integrates the coupling effects of multiple physical fields such as aerodynamic field, thermal field, and force field to achieve stability prediction and control of electric vertical take-off and landing aircraft in complex environments. It has advantages such as strong adaptability, high prediction accuracy, and fast response speed. It can improve the flight safety of eVTOL in extreme environments such as high altitude and low temperature through multi-field collaborative analysis and intelligent prediction. It is a key technology direction for ensuring the stability of new aircraft.

[0003] In existing technologies, eVTOL stability analysis mainly employs single-field analysis methods. Aerodynamic stability analysis is primarily based on aerodynamic models under standard atmospheric conditions, thermal field analysis relies mainly on the thermal characteristic parameters of the battery and thruster, and force field analysis is calculated based on the thruster's thrust distribution. When eVTOL distributed electric propulsion systems operate in high-altitude, low-temperature environments, strong coupling effects exist between the various physical fields, but a collaborative analysis mechanism is lacking. Specifically, when eVTOLs fly in high-altitude, low-temperature environments, low-density air conditions significantly affect the aerodynamic effects of the distributed thrusters. Simultaneously, the low-temperature environment alters the battery's thermal characteristics and the thruster's thermal load. These field effects interact, but independent analysis cannot accurately predict system stability, leading to inaccurate stability assessments in complex environments and impacting flight safety and control accuracy. Summary of the Invention

[0004] In view of this, the present invention proposes a neural network-based multi-field coupling stability analysis method and system for eVTOL, which solves the problem that when eVTOL flies in a high-altitude and low-temperature environment, the low-density air conditions will significantly affect the aerodynamic effect of the distributed thruster. At the same time, the low-temperature environment will change the thermal characteristics of the battery and the thermal load of the thruster. The various field effects interact with each other, but independent analysis cannot accurately predict the system stability, resulting in inaccurate stability judgment of the stability analysis method in complex environments, which affects flight safety and control accuracy.

[0005] The technical solution of this invention is implemented as follows: On one hand, this invention provides a method for eVTOL multi-field coupling stability analysis based on neural networks, including the following steps: The system collects operational data of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions. The operational data includes distributed thruster operating status data and battery thermal characteristic data. Based on the operational data, an aerodynamic-thermal-mechanical coupling model is constructed using a plateau environment three-field coupling modeling algorithm, and a three-field coupling parameter matrix is ​​output. Based on the operational data and the three-field coupling parameter matrix, a physical information constrained neural network algorithm is used for stability prediction modeling, and an adaptive stability prediction model for the plateau environment is output. Based on the adaptive stability prediction model for plateau environment, a multi-field coupled stability analysis of the eVTOL distributed electric propulsion system was conducted. The Lyapunov stability theory was used to construct the system stability criterion, and the stability analysis results were obtained. A control strategy is generated based on the stability analysis results, and control commands are sent to the thruster and battery management system through the flight control system based on the control strategy.

[0006] In some embodiments, based on the operational data, an aerodynamic-thermal-mechanical coupled model is constructed using a three-field coupling modeling algorithm for plateau environments, outputting a three-field coupling parameter matrix. Based on the operational data and the three-field coupling parameter matrix, a physical information constrained neural network algorithm is used for stability prediction modeling, outputting an adaptive stability prediction model for plateau environments, including: The operational data is subjected to plateau environment feature extraction to obtain plateau low temperature environment parameters and distributed electric propulsion system state parameters; Based on the plateau low-temperature environment parameters and the state parameters of the distributed electric propulsion system, an aerodynamic field model, a thermal field model and a force field model are constructed using a plateau environment three-field coupling modeling algorithm, and the three-field coupling parameter matrix is ​​output. Based on the aforementioned operational data, high-altitude low-temperature environmental parameters, distributed electric propulsion system state parameters, and three-field coupling parameter matrix, a physical information constrained neural network algorithm is used for stability prediction modeling, outputting an adaptive stability prediction model for the high-altitude environment.

[0007] In some embodiments, the plateau environment three-field coupling modeling algorithm includes: An aerodynamic field model is constructed based on a slipstream effect correction model, which dynamically corrects the slipstream field generated by the distributed thruster according to the low-density air conditions at high altitudes. A thermal field model is constructed based on the battery thermal decay model and the thruster thermal load model. A force field model is constructed based on a multi-thruster collaborative constraint model, and a three-field coupling parameter matrix is ​​generated through real-time data interaction between the three-field models.

[0008] In some embodiments, the physical information constrained neural network algorithm includes: A physical constraint loss function is constructed, which includes constraint terms of the aerodynamic-thermal-mechanical conservation equations and stability boundary constraint terms under plateau conditions. A neural network model is trained based on a few-shot learning architecture and the physical constraint loss function. An online adaptive optimization mechanism is used to adjust the parameters of the trained neural network model in real time, and an adaptive stability prediction model for plateau environment is output.

[0009] In some embodiments, the plateau environment three-field coupling modeling algorithm and the physical information constrained neural network algorithm employ an iterative optimization mechanism, including: Based on the prediction accuracy evaluation results of the adaptive stability prediction model for the plateau environment, the coupling weight parameters in the three-field coupling modeling algorithm for the plateau environment are dynamically adjusted; the three-field coupling parameter matrix is ​​recalculated based on the adjusted coupling weight parameters; the recalculated three-field coupling parameter matrix is ​​input into the physical information constrained neural network algorithm for model optimization until the prediction accuracy meets the set threshold.

[0010] In some embodiments, the multi-field coupled stability analysis of the eVTOL distributed electric propulsion system based on the plateau environment adaptive stability prediction model, the system stability criterion constructed using Lyapunov stability theory, and the resulting stability analysis results include: Based on the adaptive stability prediction model for the plateau environment, multi-field coupled state prediction is performed to obtain the aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector of the eVTOL distributed electric propulsion system. Based on the aerodynamic field stability state vector, thermal field stability state vector and force field stability state vector, a multi-field coupled stability criterion for plateau environment is constructed using Lyapunov stability theory, and a stability criterion matrix is ​​output. The stability of the eVTOL distributed electric propulsion system is evaluated based on the stability criterion matrix, and the stability analysis results are output.

[0011] In some embodiments, the multi-field coupling state prediction includes: The operational data is input into the plateau environment adaptive stability prediction model for forward inference, and an initial stability prediction vector is output. Based on the aerodynamic-thermal-mechanical coupling constraint conditions under the plateau low temperature environment, the initial stability prediction vector is corrected by inter-field coupling, and the aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector are output.

[0012] In some embodiments, the construction of a multi-field coupled stability criterion for a plateau environment using Lyapunov stability theory includes: A single-field stability criterion is constructed based on the Lyapunov function, and the aerodynamic field criterion vector, thermal field criterion vector, and force field criterion vector are output. A multi-field coupled stability criterion is constructed based on the inter-field coupling constraint condition, and the coupled stability criterion vector is output. The single-field stability criterion and the multi-field coupled stability criterion are combined to obtain the stability criterion matrix.

[0013] In some embodiments, the stability assessment includes: The aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector are compared and analyzed with the stability criterion matrix to output a stability deviation vector. Based on the plateau environment stability threshold, the stability deviation vector is used to determine the stability level and output the stability level and stability margin. Based on the stability level and stability margin, the stability analysis results are generated.

[0014] On the other hand, the present invention also provides a neural network-based eVTOL multi-field coupling stability analysis system, the system comprising: The data acquisition module is used to collect the operating data of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions. The operating data includes the operating status data of the distributed propulsion unit and the thermal characteristics data of the battery. The modeling and prediction module is used to construct an aerodynamic-thermal-mechanical coupling model based on the operational data using a three-field coupling modeling algorithm for plateau environment, and output a three-field coupling parameter matrix. Based on the operational data and the three-field coupling parameter matrix, a physical information constrained neural network algorithm is used to perform stability prediction modeling and output an adaptive stability prediction model for plateau environment. The stability analysis module is used to perform multi-field coupling stability analysis on the eVTOL distributed electric propulsion system based on the plateau environment adaptive stability prediction model. The system stability criterion is constructed using Lyapunov stability theory to obtain the stability analysis results. The control execution module is used to generate a control strategy based on the stability analysis results, and send control commands to the thruster and battery management system through the flight control system based on the control strategy.

[0015] The present invention provides a neural network-based method and system for eVTOL multi-field coupling stability analysis, which has the following advantages over existing technologies: (1) This application collects the operating data of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions, constructs an aerodynamic-thermal-mechanical coupling model using a three-field coupling modeling algorithm for high-altitude environments, and performs stability prediction modeling based on a physical information constrained neural network algorithm. It utilizes the nonlinear mapping capability of the neural network to capture the complex coupling relationship between the aerodynamic field, thermal field, and force field. Subsequently, it uses Lyapunov stability theory to construct a multi-field coupling stability criterion for collaborative analysis, generates accurate stability analysis results, and generates targeted control strategies accordingly. Thus, by replacing the traditional single-field independent analysis method with a multi-field coupling collaborative analysis mechanism, it realizes the coupling prediction and stability control of the aerodynamic-thermal-mechanical multi-field effects of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions, and improves the accuracy of stability judgment and flight safety under complex conditions. (2) This application obtains environmental parameters and system state parameters by extracting plateau environmental features from the operational data. It constructs aerodynamic field, thermal field, and force field models by using slip flow effect correction model, battery thermal decay model, thruster thermal load model, and multi-thruster cooperative constraint model, respectively, and realizes the interaction of three field data. Then, it constructs a physical constraint loss function containing aerodynamic-thermal-mechanical conservation equation constraint terms and stability boundary constraint terms. It trains a neural network model based on a small sample learning architecture and uses an online adaptive optimization mechanism to adjust parameters in real time. It dynamically adjusts coupling weight parameters and model parameters through an iterative optimization mechanism. Thus, by deeply integrating physical constraints and neural networks, it replaces the traditional empirical modeling method, realizes accurate modeling and adaptive stability prediction of the aerodynamic-thermal-mechanical three-field coupling effect of the eVTOL distributed electric propulsion system under plateau low temperature environment, and improves the accuracy of multi-field coupling modeling and the adaptability of prediction model under complex environmental conditions. (3) This application generates an initial stability prediction vector by inputting the operating data into the plateau environment adaptive stability prediction model for forward reasoning, and obtains the stability state vector of each field by performing inter-field coupling correction based on the aerodynamic-thermal-mechanical coupling constraint conditions; then, based on the Lyapunov function, it constructs single-field stability criteria and multi-field coupling stability criteria respectively and combines them to form a stability criteria matrix. By comparing and analyzing the state vector with the criteria matrix, it outputs the stability deviation vector, and generates the stability level and stability margin based on the plateau environment stability threshold. Thus, by replacing the traditional single-criteria evaluation method with a multi-level stability evaluation mechanism, it realizes the accurate prediction and quantitative evaluation of the multi-field coupling stability state of the eVTOL distributed electric propulsion system under plateau low temperature environment, and improves the accuracy and reliability of stability analysis under complex environmental conditions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a neural network-based eVTOL multi-field coupling stability analysis method according to the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Currently, stability analysis of eVTOL in high-altitude and low-temperature environments often relies on extensive flight tests and adjustments based on engineers' experience. Stability parameters vary significantly under different environmental conditions and system states. Stability analysis for each operating condition requires multiple experimental verifications, which is time-consuming and costly, resulting in long analysis cycles and poor safety assurance capabilities. Existing single-field analysis models suffer from a lack of rapid adaptability, insufficient accuracy, and difficulty in accurately modeling the mapping relationship between complex physical field coupling effects and stability states in complex multi-field coupled environments.

[0020] In related technologies, experience-based stability assessment systems cannot effectively handle multi-field nonlinear coupling relationships, causing analysis results to deviate from actual operating conditions. Furthermore, traditional analysis methods have limited responsiveness to dynamic environmental changes, making it difficult to meet the high-precision requirements of stability analysis in complex environments such as high altitudes and low temperatures.

[0021] Based on this, in some embodiments, please refer to Figure 1 This invention provides a neural network-based method for multi-field coupling stability analysis of eVTOL, comprising the following steps: S1, collect operational data of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions. The operational data includes distributed thruster operating status data and battery thermal characteristic data.

[0022] It is understandable that step S1 uses a multi-sensor data acquisition system to obtain the real-time operating status of the eVTOL in a high-altitude and low-temperature environment, providing basic data support for multi-field coupling analysis and ensuring the integrity and timeliness of the data.

[0023] S2, Based on the operational data, an aerodynamic-thermal-mechanical coupling model is constructed using a plateau environment three-field coupling modeling algorithm, and a three-field coupling parameter matrix is ​​output. Based on the operational data and the three-field coupling parameter matrix, a physical information constrained neural network algorithm is used for stability prediction modeling, and an adaptive stability prediction model for the plateau environment is output.

[0024] It is understandable that step S2, through the deep integration of physical mechanisms and neural networks, constructs a multi-field coupling model suitable for high-altitude and low-temperature environments, thereby achieving accurate modeling of the coupling effects between physical fields under complex environments and adaptive improvement of stability prediction capabilities.

[0025] S3. Based on the adaptive stability prediction model for the plateau environment, a multi-field coupling stability analysis was performed on the eVTOL distributed electric propulsion system. The Lyapunov stability theory was used to construct the system stability criterion, and the stability analysis results were obtained.

[0026] It is understandable that step S3 combines classical stability theory with modern prediction techniques to achieve quantitative evaluation and accurate determination of the multi-field coupling stability state of the eVTOL system, providing a reliable basis for control decisions.

[0027] S4. A control strategy is generated based on the stability analysis results, and control commands are sent to the thruster and battery management system through the flight control system based on the control strategy.

[0028] It is understandable that step S4 transforms the stability analysis results into specific control actions, and achieves stable operation of the eVTOL system in complex environments through closed-loop control, thus completing the complete technology chain from sensing, analysis to control.

[0029] Through steps S1-S4 described above, this embodiment achieves intelligent analysis and adaptive control of the multi-field coupling stability of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions, significantly improving flight safety and system reliability under complex environmental conditions. This embodiment provides an effective technical solution for ensuring the stability of new aircraft in extreme environments by organically combining multi-field collaborative modeling, neural network prediction, and closed-loop control.

[0030] In some embodiments, based on the above embodiments, step S1 includes: Collect the operating status data of the distributed thrusters of the eVTOL distributed electric propulsion system. The operating status data of the distributed thrusters includes thruster speed data, thruster power data, thruster vibration data and thruster temperature data, and output the distributed thruster operating status dataset.

[0031] It is understandable that this data acquisition process employs a multi-sensor collaborative acquisition architecture. Thruster rotation speed data is monitored in real-time via an optical encoder at a sampling frequency of one thousand times per second. Thruster power data is obtained through a power metering module, acquiring instantaneous and average power values. Thruster vibration data is acquired using a triaxial accelerometer. xyz Vibration acceleration signals in three directions and thruster temperature data are monitored by a thermistor array to detect the temperature distribution of key components such as thruster blades, motor windings and bearings. All data are marked with a unified timestamp to ensure timing consistency.

[0032] Collect battery thermal characteristic data of the eVTOL distributed electric propulsion system, including battery temperature data, battery voltage data, battery current data, and battery capacity data, and output a battery thermal characteristic dataset.

[0033] It is understandable that battery thermal characteristic data acquisition is achieved through the built-in sensor network of the battery management system. Battery temperature data is monitored by a distributed temperature sensor array to monitor the surface temperature and internal temperature gradient of each battery cell inside the battery pack. Battery voltage data is monitored in real time by a high-precision voltage sampling circuit to monitor the individual cell voltage and module voltage. Battery current data is monitored by a Hall current sensor to monitor the instantaneous and cumulative values ​​of charging and discharging current. Battery capacity data is estimated in real time based on the coulomb method combined with the open-circuit voltage method. This estimation process takes into account temperature compensation and aging coefficient correction.

[0034] The distributed thruster operating status dataset and battery thermal characteristic dataset are synchronized in time and standardized in data format to output operating data.

[0035] It is understood that the time synchronization processing in this embodiment adopts a reference clock synchronization mechanism. The data with different sampling frequencies is unified to the same time reference through the timestamp interpolation algorithm. The data format standardization processing includes three sub-steps: data type conversion, dimensional normalization, and outlier detection and filtering. Dimensional normalization uses the maximum-minimum standardization method to map all data to the interval between zero and one. Outlier detection uses the three-standard-deviation criterion based on statistical distribution to identify and remove outliers. The final running data is stored in the form of a time series matrix, where the rows of the matrix represent time steps and the columns represent different types of sensor data.

[0036] In some embodiments, based on the above embodiments, step S2 includes: The operational data is subjected to plateau environment feature extraction to obtain plateau low temperature environment parameters and distributed electric propulsion system state parameters.

[0037] It is understandable that this feature extraction process employs a multidimensional feature separation algorithm, using principal component analysis to decompose the operational data into two subspaces: environmentally relevant features and system-related features. The plateau low-temperature environmental parameters include atmospheric density. Ambient temperature Atmospheric pressure and humidity The state parameters of a distributed electric propulsion system include the thruster speed vector. Power distribution vector Battery state vector ,in, This indicates the number of thrusters, and each parameter is mapped to a normalized value. Within the interval, the influence of dimensions is eliminated.

[0038] Based on the high-altitude low-temperature environment parameters and the state parameters of the distributed electric propulsion system, an aerodynamic field model, a thermal field model, and a force field model are constructed using a three-field coupling modeling algorithm for the high-altitude environment, and the three-field coupling parameter matrix is ​​output.

[0039] For example, this three-field coupling modeling algorithm adopts a hierarchical coupling architecture. First, it constructs individual field models, and then uses a coupling matrix to characterize the interaction between fields. The aerodynamic field model is modified by considering the low-density effect of the plateau. A mapping relationship between the aerodynamic characteristics of the thruster and environmental parameters is established, and the thermal field model is based on the heat conduction equation. The force field model describes the evolution of the temperature field distribution through the torque balance equations of multiple thrusters. and Establish the relationship between thrust distribution and flight status, in which, For flight speed, For characteristic length, The aerodynamic viscosity at high altitudes, Where is the thermal diffusivity, For heat source items, For total mass, For acceleration, Let be the moment of inertia.

[0040] Based on the aforementioned operational data, high-altitude low-temperature environmental parameters, distributed electric propulsion system state parameters, and three-field coupling parameter matrix, a physical information constrained neural network algorithm is used for stability prediction modeling, outputting an adaptive stability prediction model for the high-altitude environment.

[0041] For example, the physical information constrained neural network adopts a deep feedforward network architecture, and the network structure is an input layer. - Fully connected layer - Batch Normalization Layer - ReLU Activation Layer - Fully Connected Layer -Residual connection-ReLU activation layer-Fully connected layer -Dropout layer - Fully connected layer - Output layer The numbers in parentheses represent the number of neurons in each layer, and the Dropout layer is an example. The probability of random inactivation during training is expressed as follows: That is, randomly The neuron outputs are set to zero to prevent overfitting. The physical information constrained neural network adopts an encoder-decoder architecture. The encoder part contains four fully connected layers with the number of neurons [256, 128, 64, 32], followed by a batch normalization layer and a ReLU activation function. Residual connections are added to the second and fourth layers to alleviate the gradient vanishing problem. The decoder part adopts a symmetric structure [32, 64, 128, 16]. Finally, a stable prediction result is output through a Sigmoid activation function. The total number of network parameters is approximately 150,000. The weight parameters are initialized using the Xavier initialization method, and the bias term is initialized to zero. The training process uses mini-batch gradient descent with a batch size of 64, with an initial learning rate of 0.001, which decays by a factor of 0.1 every 50 rounds.

[0042] The activation function uses the modified linear unit. The network output is processed by the Sigmoid function. Mapped to The interval, the loss function includes the data fitting term. and physical constraints The total loss function is ,in, To constrain the weight coefficients, the network uses the Adam optimizer for parameter updates, and the learning rate is scheduled using a cosine annealing strategy. .in, To minimize the learning rate, To maximize the learning rate, This is the current training round number. This represents the total number of training rounds.

[0043] Specifically, constraint weight coefficients The range of values ​​is Initial setup for training To ensure data fit, the value is gradually increased during the training phase. Strengthen physical constraints to stabilize in the later stages of training Achieve physical consistency.

[0044] In one specific embodiment, during the neural network training process, the first 100 training rounds are set as follows: Rounds 100-500 settings Set after the 500th round This ensures that the model can both fit the data and meet physical constraints.

[0045] Specifically, the minimum learning rate in the cosine annealing strategy Maximum learning rate Total number of training rounds Initial learning rate in the AdaGrad strategy Numerical stability term .

[0046] In one specific embodiment, the following is set , , Over 1000 training rounds, the learning rate gradually decreased from 0.01 to 0.00001; this was set in AdaGrad. , .

[0047] This embodiment establishes a complete theoretical framework for multi-field coupling stability analysis, and achieves accurate prediction of the stability of the eVTOL system under high-altitude and low-temperature conditions through the deep integration of physical constraints and neural networks. Through rigorous mathematical modeling and iterative optimization mechanisms, this embodiment significantly improves the accuracy of multi-field coupling effect modeling and the reliability of stability prediction under complex environmental conditions.

[0048] In some embodiments, based on the above embodiments, the plateau environment three-field coupling modeling algorithm includes: An aerodynamic field model is constructed based on a slipstream effect correction model, which dynamically corrects the slipstream field generated by the distributed thruster according to the low-density air conditions at high altitudes.

[0049] For example, the slipstream effect correction model introduces a plateau density correction factor. Establish the correction relationship for the thruster slipstream velocity. ,in, Atmospheric density at sea level Given the slipstream velocity under standard conditions, the corrected aerodynamic efficiency of the thruster is... The relationship between the modified Reynolds number and polynomial fitting was established. ,coefficient , , The aerodynamic coefficient correction formula is obtained through least squares regression of historical flight data. ,in The standard thrust coefficient, and These are density and temperature deviations, respectively.

[0050] A thermal field model was constructed based on the battery thermal decay model and the thruster thermal load model.

[0051] It is understandable that the battery thermal degradation model uses the Arrhenius equation to describe the effect of temperature on battery performance. ,in, For standard capacity, To activate energy, The gas constant is... For absolute temperature, the thruster thermal load model is derived from the thermal balance equation. Establish the relationship between heat generation and heat dissipation, among which For heat generation power, For convective heat dissipation power, For radiative heat dissipation power, the convective heat dissipation coefficient is obtained through the Nusselt number relationship. Determine the heat transfer coefficient The battery temperature prediction model is ,in, For heat capacity, For surface area, The Stefan-Boltzmann constant is... Emission rate.

[0052] A force field model is constructed based on a multi-thruster collaborative constraint model, and a three-field coupling parameter matrix is ​​generated through real-time data interaction between the three-field models.

[0053] For example, a multi-thruster cooperative constraint model is used to establish a thrust allocation optimization problem. Constraints include thrust balance And torque balance ,in, For the thruster position vector, the three-field coupling parameter matrix is... The weighted combination is expressed as Weighting coefficient , , Satisfy normalization conditions coupling matrices of each field , , The effects of aerodynamic-thermal coupling, thermodynamic-mechanical coupling, and aerodynamic-mechanical coupling are characterized respectively, and the coupling strength is quantified by cross-correlation coefficient.

[0054] Specifically, weighting coefficients , , The range of values ​​is ,in, Indicates the weight of the aerodynamic field. Indicates the thermal field weight. This represents the force field weights, satisfying the normalization constraint conditions. .

[0055] In one specific embodiment, in a high-altitude, low-temperature environment, when the ambient temperature is -20°C and the altitude is 4000 meters, the following settings are implemented: , , At this point, the aerodynamic and force fields dominate, while the thermal field has a relatively small impact. When the battery temperature is too low, the weights are adjusted to... , , Increase the importance of thermal field management.

[0056] In some embodiments, based on the above embodiments, the physical information constrained neural network algorithm includes: A physical constraint loss function is constructed, which includes constraint terms of the aerodynamic-thermal-mechanical conservation equations and stability boundary constraint terms under plateau conditions. A neural network model is trained based on a few-shot learning architecture and the physical constraint loss function. An online adaptive optimization mechanism is used to adjust the parameters of the trained neural network model in real time, and an adaptive stability prediction model for plateau environment is output.

[0057] For example, the physical constraint loss function includes a mass conservation term. Momentum conservation term and energy conservation terms Stability boundary constraints Through Lyapunov functions Derivative constraints To ensure system stability, the few-shot learning architecture employs the Model Independent Meta-Learning (MAML) method for rapid adaptation, with the parameter update rule being: ,in, The inner learning rate is used for outer optimization. Update meta-parameters; online adaptive optimization adjusts network weights based on prediction error within a sliding window. , It is an adaptive learning rate.

[0058] Specifically, inner learning rate Outer learning rate Adaptive learning rate .

[0059] In one specific embodiment, settings are configured in MAML meta-learning. Used for rapid adaptation within a task Used for updating meta parameters. Used for online adaptive adjustment to ensure that the model can quickly adapt to new high-altitude environmental conditions.

[0060] In some embodiments, based on the above embodiments, the plateau environment three-field coupling modeling algorithm and the physical information constrained neural network algorithm adopt an iterative optimization mechanism, including: Based on the prediction accuracy evaluation results of the adaptive stability prediction model for the plateau environment, the coupling weight parameters in the three-field coupling modeling algorithm for the plateau environment are dynamically adjusted; the three-field coupling parameter matrix is ​​recalculated based on the adjusted coupling weight parameters; the recalculated three-field coupling parameter matrix is ​​input into the physical information constrained neural network algorithm for model optimization until the prediction accuracy meets the set threshold.

[0061] For example, the root mean square error and mean absolute error are used as evaluation indicators for prediction accuracy assessment, and the weight parameters are adjusted using the adaptive gradient descent method. The learning rate is adjusted using the AdaGrad strategy. , For the number of iterations, This represents the number of historical iterations. For numerically stable terms, the iteration termination condition is: Or reach the maximum number of iterations , To achieve a preset accuracy threshold, the model optimization process uses alternating updates of the three model parameters and neural network weights to achieve collaborative optimization. Convergence is monitored by assessing the gradient norm of the loss function. Guarantee, among which, For convergence criteria.

[0062] In some embodiments, based on the above embodiments, step S3 includes: Based on the adaptive stability prediction model for the plateau environment, multi-field coupled state prediction is performed to obtain the aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector of the eVTOL distributed electric propulsion system.

[0063] It is understandable that this multi-field coupled state prediction obtains the stability states of each physical field through forward propagation calculation of a neural network. The formula for calculating the aerodynamic field stability state vector is as follows: The thermal field stability state vector is The force field stability state vector is ,in, For environmental parameter vectors, For the thruster parameter vector, For the battery parameter vector, A vector of temperature parameters. For the thrust parameter vector, For torque parameter vectors, , , These are the neural network mapping functions for the corresponding physical fields, ensuring the accurate extraction and quantization of the stability features of each physical field.

[0064] Based on the aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector, a multi-field coupled stability criterion for plateau environment is constructed using Lyapunov stability theory, and a stability criterion matrix is ​​output.

[0065] It is understandable that this stability criterion is constructed based on composite Lyapunov functions. ,in, Let Lyapunov be the aerodynamic field function. For the thermal field Lyapunov function, Let Lyapunov be the force field function. For multi-field coupled Lyapunov functions, where, , , It is a positive definite symmetric matrix. These are the coupling weight coefficients. For the coupling matrix, the subscripts are... , Representing different physical field indices, stability criterion matrix This provides a mathematical basis for the stability assessment of multi-field coupled systems.

[0066] Specifically, coupling weight coefficients The range of values ​​is diagonal elements Represents the stability weight of a single game, off-diagonal elements. Represents the inter-field coupling strength, and the total weight satisfies .

[0067] In one specific embodiment, for aerodynamic field self-coupling, a configuration is set... For thermal field self-coupling, set For force field self-coupling, set For aero-thermal coupling, set For aerodynamic-force coupling, set For thermo-mechanical coupling, set .

[0068] The stability of the eVTOL distributed electric propulsion system is evaluated based on the stability criterion matrix, and the stability analysis results are output.

[0069] For example, this stability assessment is performed by comparing functions. Calculate the system stability deviation, where, For the actual system state vector, As a reference stable state vector, the stability index is obtained through The calculated stability analysis results include stability index values ​​and the contribution weights of each physical field. And the overall system stability rating, among which, For the first The stability deviation components of each physical field provide quantitative stability assessment information for flight control decisions.

[0070] In some embodiments, based on the above embodiments, the multi-field coupling state prediction includes: The operational data is input into the plateau environment adaptive stability prediction model for forward inference, and an initial stability prediction vector is output. Based on the aerodynamic-thermal-mechanical coupling constraint conditions under the plateau low temperature environment, the initial stability prediction vector is corrected by inter-field coupling, and the aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector are output.

[0071] For example, the forward inference process obtains the initial prediction result through the forward propagation calculation of the neural network. ,in, To standardize the input feature vector, Represents the neural network mapping function. The initial stability prediction vector is used, and the inter-field coupling correction is achieved through the coupling constraint matrix. The initial prediction is adjusted, and the corrected aerodynamic field stability state vector is: ,in This represents the aerodynamic field component in the initial prediction vector. This is the aerodynamic field coupling correction matrix. To couple the correction increment, the thermal field stability state vector and the force field stability state vector The calculations were performed using the same correction strategy to ensure that the interactions between the various physical fields are accurately reflected.

[0072] In some embodiments, based on the above embodiments, the construction of a multi-field coupled stability criterion for plateau environments using Lyapunov stability theory includes: A single-field stability criterion is constructed based on the Lyapunov function, and the aerodynamic field criterion vector, thermal field criterion vector, and force field criterion vector are output. A multi-field coupled stability criterion is constructed based on the inter-field coupling constraint condition, and the coupled stability criterion vector is output. The single-field stability criterion and the multi-field coupled stability criterion are combined to obtain the stability criterion matrix.

[0073] For example, the single-field stability criterion is based on the quadratic Lyapunov function. Construct, where subscript Indicates the type of physical field ( ), These are the state variables of the corresponding physical field. As a positive definite symmetric matrix, the aerodynamic field criterion vector passes through... The calculated thermal field criterion vector is... Force field criterion vector Using the same method, the multi-field coupling stability criterion is calculated by coupling Lyapunov functions. Construct, in which, These are the coupling weight coefficients. For the coupling matrix, the subscripts are... , These represent the first and second physical field indices, respectively, and the coupling stability criterion vector. The stability criterion matrix is ​​obtained by combining the block matrices. This ensures that all stability conditions for single-field and multi-field coupling are covered.

[0074] In some embodiments, based on the above embodiments, the stability assessment includes: The aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector are compared and analyzed with the stability criterion matrix to output a stability deviation vector. Based on the plateau environment stability threshold, the stability deviation vector is used to determine the stability level and output the stability level and stability margin. Based on the stability level and stability margin, the stability analysis results are generated.

[0075] For example, the stability deviation vector is obtained through The calculation yielded that, This represents the overall stability state vector of the system. The stability threshold vector is used, and the stability level is determined using a piecewise function. ,in, , , These are the safety, warning, and critical stability thresholds, respectively. The stability margin calculation formula is as follows: ,in, The stability analysis results include the stability level, which represents the upper limit of the threshold corresponding to the current level. Stability margin Contribution of each physical field and risk warning information, among which For the first The deviation components corresponding to each physical field provide a comprehensive stability assessment basis for flight control decisions.

[0076] Specifically, safety threshold Warning threshold Critical threshold The thresholds must satisfy the following conditions: .

[0077] In one specific embodiment, a design is provided for high-altitude environments. , , When the stability deviation norm is less than 0.05, the system is in a safe state; between 0.05 and 0.2, it is in a normal state; between 0.2 and 0.5, it is in a warning state; and greater than 0.5, it is in a dangerous state.

[0078] This embodiment establishes a complete multi-field coupling stability assessment system through stability analysis algorithms. By organically combining Lyapunov stability theory with neural network prediction, it achieves accurate assessment and early warning of eVTOL system stability under high-altitude and low-temperature environments. Through rigorous mathematical modeling and quantitative analysis, this embodiment improves the scientific rigor and reliability of stability assessment for multi-physics coupled systems under complex environmental conditions, providing a solid theoretical foundation and technical support for the safe flight control of eVTOL in extreme environments.

[0079] In some embodiments, based on the above embodiments, step S4 includes: Based on the stability analysis results, control parameters are calculated and control logic is constructed. The stability analysis results include thruster stability state data and battery thermal management state data, and the control strategy of the eVTOL distributed electric propulsion system is output.

[0080] For example, the control parameter calculation process uses the stability deviation vector. and stability margin The thruster adjustment parameters are calculated, and the formula for calculating the thruster speed adjustment is as follows: The thruster power adjustment amount is ,in, Indicates the thruster index number, , , These are the proportional, integral, and derivative gain coefficients for speed control. For the first Aerodynamic field stability deviation of each thruster For the first Force field stability deviation of each thruster , These are the proportional and integral gain coefficients for power control, respectively. Battery management parameters, including charge / discharge current limits, are calculated using thermal stability data. and temperature control target value ,in, For the maximum allowable current, This is the threshold for thermal field stability deviation. For optimal operating temperature, For temperature correction, the control logic is constructed to optimize parameter combinations through a fuzzy logic rule base.

[0081] Specifically, the speed control gain coefficient range is: , , The power control gain coefficient range is: , .

[0082] In one specific embodiment, for an eVTOL system with four thrusters, a configuration is provided. , , , , This enables rapid response and avoids overshoot.

[0083] Based on the control strategy of the eVTOL distributed electric propulsion system, control commands are sent from the flight control system to the thrusters and battery management system. These control commands include thruster control commands and battery management control commands.

[0084] For example, this control command sending process converts the control strategy into specific execution instructions, and the thruster control instructions are generated through a formula using PWM signals. The calculation yielded that, As the reference PWM duty cycle, For maximum speed, For PWM adjustment range, battery management control commands are transmitted via charging control signals. and thermal management control signals The formula for generating the thermal management control signal is: ; in, To request charging current, This represents the actual battery temperature. For the maximum allowable temperature, , These are the minimum and maximum fan speeds, respectively. Control commands are transmitted in real time to each actuator via the CAN bus protocol to ensure the timeliness and accuracy of the system response.

[0085] This embodiment establishes a complete eVTOL multi-field coupled stability control execution system by generating control strategies and describing command transmission algorithms. By effectively transforming stability analysis results into specific control actions, adaptive stability control of the eVTOL system under high-altitude and low-temperature environments is achieved. This embodiment improves the accuracy and response speed of multi-physics coupled system control under complex environmental conditions through mathematical modeling and real-time control algorithms.

[0086] In some embodiments, the present invention also provides a neural network-based eVTOL multi-field coupling stability analysis system, the system comprising: The data acquisition module is used to collect the operating data of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions. The operating data includes the operating status data of the distributed propulsion unit and the thermal characteristics data of the battery. The modeling and prediction module is used to construct an aerodynamic-thermal-mechanical coupling model based on the operational data using a three-field coupling modeling algorithm for plateau environment, and output a three-field coupling parameter matrix. Based on the operational data and the three-field coupling parameter matrix, a physical information constrained neural network algorithm is used to perform stability prediction modeling and output an adaptive stability prediction model for plateau environment. The stability analysis module is used to perform multi-field coupling stability analysis on the eVTOL distributed electric propulsion system based on the plateau environment adaptive stability prediction model. The system stability criterion is constructed using Lyapunov stability theory to obtain the stability analysis results. The control execution module is used to generate a control strategy based on the stability analysis results, and send control commands to the thruster and battery management system through the flight control system based on the control strategy.

[0087] This embodiment presents a neural network-based eVTOL multi-field coupling stability analysis system, which realizes intelligent analysis and adaptive control of the multi-field coupling stability of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions, significantly improving flight safety and system reliability under complex environmental conditions. This embodiment provides an effective technical solution for ensuring the stability of new aircraft in extreme environments by organically combining multi-field collaborative modeling, neural network prediction, and closed-loop control.

[0088] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for multi-field coupling stability analysis of eVTOL based on neural networks, characterized in that, Includes the following steps: The system collects operational data of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions. The operational data includes distributed thruster operating status data and battery thermal characteristic data. Based on the operational data, an aerodynamic-thermal-mechanical coupling model is constructed using a plateau environment three-field coupling modeling algorithm, and a three-field coupling parameter matrix is ​​output. Based on the operational data and the three-field coupling parameter matrix, a physical information constrained neural network algorithm is used for stability prediction modeling, and an adaptive stability prediction model for the plateau environment is output. Based on the adaptive stability prediction model for plateau environment, a multi-field coupled stability analysis of the eVTOL distributed electric propulsion system was conducted. The Lyapunov stability theory was used to construct the system stability criterion, and the stability analysis results were obtained. A control strategy is generated based on the stability analysis results, and control commands are sent to the thruster and battery management system through the flight control system based on the control strategy.

2. The method for eVTOL multi-field coupling stability analysis based on neural networks as described in claim 1, characterized in that, Based on the operational data, an aerodynamic-thermal-mechanical coupled model is constructed using a three-field coupling modeling algorithm for plateau environments, outputting a three-field coupling parameter matrix. Based on the operational data and the three-field coupling parameter matrix, a physical information constrained neural network algorithm is used for stability prediction modeling, outputting an adaptive stability prediction model for plateau environments, including: The operational data is subjected to plateau environment feature extraction to obtain plateau low temperature environment parameters and distributed electric propulsion system state parameters; Based on the plateau low-temperature environment parameters and the state parameters of the distributed electric propulsion system, an aerodynamic field model, a thermal field model and a force field model are constructed using a plateau environment three-field coupling modeling algorithm, and the three-field coupling parameter matrix is ​​output. Based on the aforementioned operational data, high-altitude low-temperature environmental parameters, distributed electric propulsion system state parameters, and three-field coupling parameter matrix, a physical information constrained neural network algorithm is used for stability prediction modeling, outputting an adaptive stability prediction model for the high-altitude environment.

3. The method for eVTOL multi-field coupling stability analysis based on neural networks as described in claim 2, characterized in that, The three-field coupling modeling algorithm for the plateau environment includes: An aerodynamic field model is constructed based on a slipstream effect correction model, which dynamically corrects the slipstream field generated by the distributed thruster according to the low-density air conditions at high altitudes. A thermal field model is constructed based on the battery thermal decay model and the thruster thermal load model. A force field model is constructed based on a multi-thruster collaborative constraint model, and a three-field coupling parameter matrix is ​​generated through real-time data interaction of the three-field models.

4. The method for eVTOL multi-field coupling stability analysis based on neural networks as described in claim 2, characterized in that, The physical information constrained neural network algorithm includes: A physical constraint loss function is constructed, which includes constraint terms of the aerodynamic-thermal-mechanical conservation equations and stability boundary constraint terms under plateau conditions. A neural network model is trained based on a few-shot learning architecture and the physical constraint loss function. An online adaptive optimization mechanism is used to adjust the parameters of the trained neural network model in real time, and an adaptive stability prediction model for plateau environment is output.

5. The method for eVTOL multi-field coupling stability analysis based on neural networks as described in claim 2, characterized in that, The plateau environment three-field coupling modeling algorithm and the physical information constrained neural network algorithm adopt an iterative optimization mechanism, including: Based on the prediction accuracy evaluation results of the adaptive stability prediction model for the plateau environment, the coupling weight parameters in the three-field coupling modeling algorithm for the plateau environment are dynamically adjusted; the three-field coupling parameter matrix is ​​recalculated based on the adjusted coupling weight parameters; the recalculated three-field coupling parameter matrix is ​​input into the physical information constrained neural network algorithm for model optimization until the prediction accuracy meets the set threshold.

6. The method for eVTOL multi-field coupling stability analysis based on neural networks as described in claim 1, characterized in that, The above-mentioned adaptive stability prediction model for the plateau environment is used to conduct multi-field coupled stability analysis on the eVTOL distributed electric propulsion system. Lyapunov stability theory is used to construct system stability criteria, and the stability analysis results are obtained, including: Based on the adaptive stability prediction model for the plateau environment, multi-field coupled state prediction is performed to obtain the aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector of the eVTOL distributed electric propulsion system. Based on the aerodynamic field stability state vector, thermal field stability state vector and force field stability state vector, a multi-field coupled stability criterion for plateau environment is constructed using Lyapunov stability theory, and a stability criterion matrix is ​​output. The stability of the eVTOL distributed electric propulsion system is evaluated based on the stability criterion matrix, and the stability analysis results are output.

7. The method for multi-field coupling stability analysis of eVTOL based on neural networks as described in claim 6, characterized in that, The multi-field coupled state prediction includes: The operational data is input into the plateau environment adaptive stability prediction model for forward inference, and an initial stability prediction vector is output. Based on the aerodynamic-thermal-mechanical coupling constraint conditions under the plateau low temperature environment, the initial stability prediction vector is corrected by inter-field coupling, and the aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector are output.

8. The method for multi-field coupling stability analysis of eVTOL based on neural networks as described in claim 6, characterized in that, The multi-field coupled stability criterion for plateau environments constructed using Lyapunov stability theory includes: A single-field stability criterion is constructed based on the Lyapunov function, and the aerodynamic field criterion vector, thermal field criterion vector, and force field criterion vector are output. A multi-field coupled stability criterion is constructed based on the inter-field coupling constraint condition, and the coupled stability criterion vector is output. The single-field stability criterion and the multi-field coupled stability criterion are combined to obtain the stability criterion matrix.

9. The method for multi-field coupling stability analysis of eVTOL based on neural networks as described in claim 6, characterized in that, The stability assessment includes: The aerodynamic field stability state vector, thermal field stability state vector, and force field stability state vector are compared and analyzed with the stability criterion matrix to output a stability deviation vector. Based on the plateau environment stability threshold, the stability deviation vector is used to determine the stability level and output the stability level and stability margin. Based on the stability level and stability margin, the stability analysis results are generated.

10. A neural network-based eVTOL multi-field coupling stability analysis system, used to execute the neural network-based eVTOL multi-field coupling stability analysis method as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition module is used to collect the operating data of the eVTOL distributed electric propulsion system under high-altitude and low-temperature conditions. The operating data includes the operating status data of the distributed propulsion unit and the thermal characteristics data of the battery. The modeling and prediction module is used to construct an aerodynamic-thermal-mechanical coupling model based on the operational data using a three-field coupling modeling algorithm for plateau environment, and output a three-field coupling parameter matrix. Based on the operational data and the three-field coupling parameter matrix, a physical information constrained neural network algorithm is used to perform stability prediction modeling and output an adaptive stability prediction model for plateau environment. The stability analysis module is used to perform multi-field coupling stability analysis on the eVTOL distributed electric propulsion system based on the plateau environment adaptive stability prediction model. The system stability criterion is constructed using Lyapunov stability theory to obtain the stability analysis results. The control execution module is used to generate a control strategy based on the stability analysis results, and send control commands to the thruster and battery management system through the flight control system based on the control strategy.

Images