Unmanned aerial vehicle life prediction method and device, terminal and storage medium

Abstract

The invention provides an unmanned aerial vehicle life prediction method and device, a terminal and a storage medium, and relates to the technical field of unmanned aerial vehicle life prediction. The method comprises the following steps: acquiring operation data of the unmanned aerial vehicle, wherein the operation data comprises flight time, flight attitude, flight position, motion trail, environmental parameters, component state and mechanical stress; based on the operation data, an unmanned aerial vehicle digital twin model is constructed, the unmanned aerial vehicle digital twin model is used for simulating the operation states of the unmanned aerial vehicle under different conditions, virtual data are generated, and the unmanned aerial vehicle digital twin model comprises a geometric sub-model, a kinematics sub-model, a dynamology sub-model, an environment sub-model and a control sub-model; and constructing an enhanced data set based on the operation data and the virtual data, and training an unmanned aerial vehicle life prediction model by using the enhanced data set. The method can improve the accuracy of unmanned aerial vehicle life prediction under the condition of data shortage.

Description

Technical Field

[0001] The present application relates to the technical field of drone life prediction, and in particular to a drone life prediction method, device, terminal, and storage medium. Background Art

[0002] With the widespread application of drones in various fields, their reliability and safety are becoming increasingly important. Accurately predicting the remaining useful life of drones is crucial for ensuring the smooth progress of flight missions, reducing maintenance costs, and improving overall drone performance. However, in practical applications, due to the difficulty of collecting drone flight data and the complexity of flight conditions, insufficient data is often encountered. Traditional life prediction methods struggle to accurately estimate drone life in this data-poor environment. Summary of the Invention

[0003] The present application provides a drone life prediction method, device, terminal and storage medium to solve the problem of low accuracy of drone life prediction in the existing technology under insufficient data conditions.

[0004] In a first aspect, the present application provides a method for predicting the lifespan of a drone, comprising:

[0005] Collecting the UAV's operational data, including flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status, and mechanical stress;

[0006] Based on the operating data, a digital twin model of the drone is constructed, and the operating state of the drone under different conditions is simulated using the digital twin model of the drone to generate virtual data. The digital twin model of the drone includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model, and a control sub-model.

[0007] An enhanced data set is constructed based on the operating data and the virtual data, and a UAV life prediction model is trained using the enhanced data set.

[0008] In a second aspect, the present application provides a drone life prediction device, comprising:

[0009] A data acquisition module is used to collect the operating data of the UAV, including flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status and mechanical stress;

[0010] A simulation data generation module is used to build a UAV digital twin model based on the operating data, and use the UAV digital twin model to simulate the operating status of the UAV under different conditions to generate virtual data. The UAV digital twin model includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model, and a control sub-model;

[0011] The model training module is used to construct an enhanced data set based on the operating data and the virtual data, and use the enhanced data set to train a UAV life prediction model.

[0012] In a third aspect, the present application provides a terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method described in the first aspect or any possible implementation of the first aspect are implemented.

[0013] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the method described in the first aspect or any possible implementation of the first aspect.

[0014] The present application provides a method, device, terminal, and storage medium for predicting the life of a drone. The method collects the drone's operating data, including flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status, and mechanical stress. Based on the operating data, a drone digital twin model is constructed, and the drone digital twin model is used to simulate the drone's operating status under different conditions, generating virtual data. The drone digital twin model includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model, and a control sub-model. An enhanced data set is constructed based on the operating data and virtual data, and the enhanced data set is used to train the drone's life prediction model. By constructing a drone digital twin model, the present application can obtain the drone's operating status under different conditions in real time, thereby expanding the test data. The acquired operating data and virtual data are then used to construct an enhanced data set and train the drone's life prediction model, thereby improving prediction accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0016] Figure 1 This is a schematic diagram of the drone digital twin and deep learning model architecture provided by the embodiment of the present application;

[0017] Figure 2 This is a flowchart of the implementation of the drone life prediction method provided in the embodiment of the present application;

[0018] Figure 3 This is a schematic diagram of the structure of the UAV life prediction device provided in an embodiment of the present application;

[0019] Figure 4 It is a schematic diagram of a terminal provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0021] In order to make the purpose, technical solutions and advantages of this application clearer, specific embodiments will be described below with reference to the accompanying drawings.

[0022] To address the difficulty of traditional lifespan prediction methods in accurately estimating drone lifespan in data-poor environments, this application aims to provide a drone lifespan prediction method using digital twins and deep learning in this data-poor environment. By building a digital twin model of the drone and combining it with a deep learning algorithm, the method achieves accurate and reliable prediction of drone lifespan.

[0023] Reference Figure 1 The drone digital twin and deep learning model architecture enables real-time monitoring, simulation, and optimized control of drones. This architecture not only improves the intelligence level of drones but also provides strong support for the execution of complex tasks. It mainly includes the physical entity layer, data interaction layer, data management and storage, digital twin model layer, deep learning layer, and application layer. Specifically:

[0024] Physical Layer: The physical layer is the physical foundation of the drone digital twin system and primarily includes the drone's hardware and sensors. These sensors collect real-time drone flight status data, such as position, speed, attitude, acceleration, and temperature, and transmit this data to the digital twin system.

[0025] Drone hardware: including fuselage, power system, flight control system, sensors (such as IMU, GPS, camera), etc.

[0026] Sensor network: used to monitor the status and environmental information of drones in real time.

[0027] Data Interaction Layer: The data interaction layer bridges the gap between physical entities and data management and storage, responsible for data transmission and processing. It receives sensor data from physical entities and transmits it to data management and storage, while also feeding back control instructions generated by the digital twin model to the physical entity.

[0028] Data transmission module: Realize real-time data transmission through wireless communication technology (such as Wi-Fi, 5G) or wired connection.

[0029] Data preprocessing unit: performs preprocessing operations such as cleaning, denoising, and format conversion on the collected data to improve data quality.

[0030] Data management and storage: Data management and storage are responsible for storing and managing data from physical entities and digital twin models. The digital twin system needs to reflect the status and behavior of physical entities in real time, so the data management module must support low-latency, high-frequency data updates and synchronization.

[0031] Real-time database: stores sensor data and real-time flight status information.

[0032] Historical database: Saves historical flight data for analysis and model training.

[0033] Cloud storage: Digital twin systems need to process large amounts of real-time data, historical data, and predictive data, so the data storage module must have high capacity, high performance, and scalability.

[0034] Digital twin model layer: The digital twin model layer is the virtual part of the drone digital twin system, used to simulate and reflect the state and behavior of physical entities. It usually includes the following submodules:

[0035] Geometric model: describes the geometry and appearance of the drone for visualization in a virtual environment.

[0036] Kinematic model: simulates the motion state of the drone, including position, speed, and attitude.

[0037] Dynamic model: describes the dynamic behavior of the drone, such as acceleration, angular velocity, etc.

[0038] Environmental model: Build a virtual flight environment to simulate the flight scenario of the UAV, including terrain, weather conditions, etc.

[0039] Control Model: An algorithm based on machine learning or traditional control theory used to optimize the UAV’s flight path and mission planning.

[0040] Deep Learning Layer: The lightweight CNN-Transformer-BiLSTM architecture effectively combines the local feature extraction capabilities of convolutional neural networks (CNNs), the global modeling capabilities of transformers, and the time series modeling capabilities of bidirectional long short-term memory networks (BiLSTMs). This architecture enables accurate modeling of complex tasks while maintaining efficient computation.

[0041] This architectural design can not only effectively process multi-source heterogeneous data from drones, but also run efficiently on resource-constrained edge devices, providing support for real-time decision-making of drones.

[0042] (1) Lightweight CNN module: The CNN module is mainly used to extract local features of input data and is suitable for processing temporal and spatial correlations in drone sensor data.

[0043] (2) Transformer module: used to capture global dependencies and enhance the model’s ability to model long-distance dependencies.

[0044] Multi-Head Self-Attention (MHSA): Multiple attention heads are used to process input features simultaneously and capture dependencies in different subspaces.

[0045] Locally Enhanced Feed-Forward (LeFF): Combined with convolution operations to enhance local feature extraction capabilities.

[0046] (3) BiLSTM module: used to process time series data and capture the previous and next dependencies in the time dimension.

[0047] Bidirectional LSTM (BiLSTM): Combines forward and backward LSTM to model time series from two directions simultaneously, enhancing the model's ability to capture time dependencies.

[0048] Application layer: mainly oriented towards user needs, providing intuitive interactive interface and diversified functional support.

[0049] (1) 3D model visualization: The application layer uses 3D model visualization to provide users with an intuitive display of the drone's operating status. Users can view the drone's flight attitude, position, and surrounding environment in real time. A 3D visualization interface is constructed using 3D modeling technology, such as Unity3D. Real-time data is transmitted to the front end via a communication protocol, dynamically updating the 3D model. The communication protocol can be WebSocket or other low-latency communication protocols.

[0050] Real-time 3D modeling: Dynamically update the position, attitude, and motion trajectory of the 3D model based on the real-time sensor data of the drone (such as GPS, IMU).

[0051] Environmental rendering: Combined with environmental sensor data (such as meteorological data and terrain information), the real flight environment is rendered in the 3D model, including terrain and weather changes.

[0052] Interactive operation: Users can operate the drone's viewing angle, zoom in / out, etc. through the interface to obtain more detailed flight information.

[0053] (2) Lifespan prediction: The application layer provides drone lifespan prediction functionality. Based on the output of a deep learning model (a lightweight CNN-Transformer-BiLSTM), it provides users with drone health status assessment and remaining useful life (RUL) prediction. Data visualization tools (such as ECharts and D3.js) are used to draw lifespan prediction curves and health status dashboards.

[0054] Health status monitoring: Displays real-time health indicators of key drone components (such as batteries, motors, and sensors).

[0055] Life Prediction: Based on the prediction results of the deep learning model, it shows the remaining service life of the drone and provides the prediction confidence interval.

[0056] Early warning and suggestions: When the predicted life is lower than the safety threshold, the system automatically issues an early warning and provides maintenance or replacement suggestions.

[0057] (3) Human-computer interaction interface: The application layer provides a user-friendly human-computer interaction interface, allowing users to easily interact with the system, obtain information, and issue instructions. For example, HTML5, CSS3, and JavaScript can be used to build the web interface. React or Vue.js frameworks can also be used to improve development efficiency. The application programming interface communicates with the back-end system to obtain data and update the interface in real time.

[0058] User interface design: Provides a simple and intuitive graphical interface, including dashboards, map views, 3D model views, etc.

[0059] Flight mission planning: Users can set the drone's flight path, mission objectives, etc. through the interface.

[0060] Real-time control: Users can manually adjust the drone's flight parameters (such as speed, altitude) or switch flight modes.

[0061] Data query and analysis: Users can query historical flight data, health records, and perform data analysis.

[0062] Figure 2 The implementation flow chart of the drone life prediction method provided in the embodiment of the present application is detailed as follows:

[0063] In step 101, the operation data of the UAV is collected, and the operation data includes flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status and mechanical stress.

[0064] In an embodiment of the present application, the operating data of the drone is collected through sensors of various hardware devices on the drone, including but not limited to flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status and mechanical stress.

[0065] Since the collected operating data is the basic data for the subsequent construction of the drone digital twin model and training of the drone life prediction model in this embodiment, the collected operating data needs to be preprocessed, including data cleaning, normalization and feature extraction. The purpose is to avoid omissions, errors or duplications in the collected data, which may cause the finally constructed drone digital twin model and drone life prediction model to be inconsistent with the actual situation.

[0066] The embodiments of the present application comprehensively cover the physical state of the drone and the interaction characteristics with the environment by collecting various operational data such as flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status and mechanical stress, thereby reducing prediction deviations caused by missing or one-sided data and significantly improving the reliability of the remaining life assessment of the drone.

[0067] In step 102, a UAV digital twin model is constructed based on the operating data, and the UAV digital twin model is used to simulate the operating status of the UAV under different conditions to generate virtual data. The UAV digital twin model includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model and a control sub-model.

[0068] In this embodiment of the present application, the drone's operational data collected in step 101 is used to construct a digital twin model of the drone. This digital twin model includes a geometry sub-model, a kinematic sub-model, a dynamics sub-model, an environmental sub-model, and a control sub-model, enabling real-time mapping between the physical drone and the virtual model. The drone's digital twin model is then used to simulate the drone's operational status under different circumstances, generating a large amount of virtual data. This virtual data includes virtual fault data and virtual operational data.

[0069] In the embodiments of this application, the drone digital twin model can reflect the drone's status in real time and support virtual simulation. The drone digital twin model simulates the drone's physical characteristics, operating environment, and failure modes, generating virtual data to supplement the actual data.

[0070] Furthermore, the digital twin technology in the embodiments of this application enables real-time feedback of the drone's operating status into the virtual model. Combined with the real-time predictions from the deep learning model, managers can keep tabs on the drone's health and remaining lifespan. This real-time monitoring capability provides timely decision-making support for managers, enabling them to quickly make informed decisions based on the drone's actual conditions.

[0071] In one possible implementation, the process of constructing the geometric sub-model is as follows:

[0072] The geometric parameters, actual size and shape of the drone's propeller, as well as the geometric positions of each component, are input into the 3D modeling software to construct a geometric sub-model of the drone;

[0073] Import the UAV's geometric sub-model into the Unreal Engine and optimize the UAV's geometric sub-model by setting the skeleton hierarchy and physical properties.

[0074] Optionally, the geometric sub-model in this embodiment is the basis of the UAV digital twin model, which is used to accurately represent the appearance and structure of the UAV. Use professional 3D modeling software to build the geometric sub-model of the UAV, accurately model the propeller according to its geometric parameters (such as diameter, pitch, and blade shape), build the geometric model of the fuselage and the arm according to the actual size and shape of the UAV, assemble the various components according to the actual geometric position relationship, and ensure that the geometric sub-model is consistent with the physical entity in appearance, size, and shape. Save the constructed geometric sub-model and import it into the virtual engine. By setting a new skeleton hierarchy or adding physical properties, the dynamic performance of the geometric sub-model is further optimized. Among them, the 3D modeling software can be CATIA, the virtual engine can be Unity 3D, or it can be selected according to actual conditions.

[0075] In one possible implementation, the kinematic submodel is constructed as follows:

[0076] Establish a body coordinate system and an inertial coordinate system. The body coordinate system uses the center of mass of the drone as its origin, while the inertial coordinate system uses the earth as its reference coordinate. These are used to describe the absolute position and attitude of the drone.

[0077] Using the flight attitude and flight position of the UAV, the kinematic equation of the UAV is established. The kinematic equation includes the translational motion equation and the rotational motion equation.

[0078] The Euler method is used to solve the kinematic equations to obtain the position and attitude of the UAV at each time step.

[0079] Optionally, the kinematic sub-model in this embodiment describes the position and attitude changes of the drone without considering the effects of forces. The specific construction process is as follows:

[0080] The first step is to build a coordinate system.

[0081] Establish the body coordinate system (B system), take the center of mass of the UAV as the origin, and establish the coordinate system along the three main axes of the body.

[0082] Establish an inertial coordinate system (I system), usually with the earth as the reference, to describe the absolute position and attitude of the drone.

[0083] The second step is to establish the kinematic equations.

[0084] Based on the motion characteristics of the UAV, the kinematic equations including the translational motion equation and the rotational motion equation are derived. The translational motion equation describes the position change of the UAV in the inertial coordinate system, and the rotational motion equation describes the attitude change of the UAV. Among them, the translational motion equation is:

[0085]

[0086] in, is the position of the UAV in the inertial coordinate system, v x 、v y 、v z is the speed in the corresponding direction.

[0087] The equation of rotational motion is:

[0088]

[0089] in, is the Euler angle (roll angle Pitch angle θ, yaw angle ψ), ω x 、ω y 、ω z is the angular velocity in the body coordinate system.

[0090] The third step is to parameterize the angle and angular velocity.

[0091] Euler angles are used to describe the attitude of the drone, and the derivative of the Euler angles (angular velocity) is linked to the angular velocity in the body coordinate system through the coordinate transformation matrix.

[0092] The fourth step is to construct the coordinate transformation matrix.

[0093] Construct a coordinate transformation matrix from the body coordinate system to the inertial coordinate system to convert the body angular velocity into Euler angular velocity. The coordinate transformation matrix A is in the form of:

[0094]

[0095] Among them, φ, θ, is the Euler angle (roll angle Pitch angle θ, yaw angle ψ)

[0096] The relationship between the Euler angle and the body angular velocity is:

[0097]

[0098] Among them, A -1 is the inverse matrix of the coordinate transformation matrix A.

[0099] Step 5: Numerical solution of the kinematic sub-model.

[0100] The kinematic equations are solved using the Euler method of numerical integration to determine the drone's position and attitude at each time step. Sensor data is then used to update the state of the kinematic submodel in real time to ensure consistency between the model and the physical entity.

[0101] In one possible implementation, the dynamics sub-model includes a motor dynamics unit, a rigid body attitude dynamics unit, and a rigid body position dynamics unit. The construction process of the dynamics sub-model is as follows:

[0102] Use the motion trajectory of each motor in the drone to build a motor dynamics unit;

[0103] Use the UAV's flight attitude to construct a rigid body attitude dynamics unit;

[0104] The flight position of the UAV is used to construct a rigid body position dynamics unit.

[0105] Optionally, the dynamics submodel in this embodiment describes the relationship between the drone's motion and the forces acting on it. The dynamics submodel typically includes a motor dynamics unit, a rigid body attitude dynamics unit, and a rigid body position dynamics unit. The motor dynamics unit describes the relationship between the motor's speed and output torque, the rigid body attitude dynamics unit describes the drone's angular velocity and angular acceleration, and the rigid body position dynamics unit describes the drone's position and velocity.

[0106] The relationship between the motor's power output and speed can be obtained through experiments or theoretical calculations. The lift and torque of the motor can be expressed by the lift coefficient and the square of the speed. The resultant lift f and torque τ x , τ y , τ z It can be expressed as:

[0107]

[0108] Among them, c t is the lift coefficient, c m is the torque coefficient, ω i is the speed of the i-th motor, and d is the distance from the motor to the center of gravity.

[0109] The rigid body attitude dynamics unit describes the angular velocity and angular acceleration of the drone, taking into account the drone's moment of inertia and external torque. Euler angles are used to represent the drone's attitude, and the relationship between angular velocity and Euler angles is used for conversion. The derivative of angular velocity is:

[0110]

[0111] in, is the angular acceleration vector, which represents the rate of change of the drone's angular velocity in radians per second squared (rad / s 2 J is the moment of inertia matrix, which is a 3×3 symmetric matrix that describes the inertia of the drone against rotation. The unit is kilograms per square meter (kg·m 2 M is the external torque vector, representing the external torque acting on the drone, measured in Newton meters (N·m). ω is the angular velocity vector, representing the rotational speed of the drone, measured in radians per second (rad / s). × is the cross product of vectors.

[0112] The rigid body position dynamics unit describes the position changes of the drone, taking into account the effects of gravity, lift, and air resistance. Newton's second law is used to derive the acceleration and velocity of the drone in all directions.

[0113]

[0114] in, Respectively represent the speed of the drone in the x, y, and z directions, in meters per second (m / s), v x 、v y 、v z Respectively represent the speed of the drone in the x, y, and z directions, in meters per second (m / s). Represents the acceleration of the drone in the x, y, and z directions, respectively, in meters per second squared (m / s 2 ), f is the total lift of the drone in Newtons (N), m is the mass of the drone in kilograms (kg), θ is the pitch angle, which is the angle between the drone and the horizontal plane, in radians (rad), φ is the roll angle, which is the rotation angle of the drone around its forward direction, in radians (rad), and g is the acceleration due to gravity, which is approximately 9.81 meters per second squared (m / s 2 ).

[0115] In one possible implementation, the environmental sub-model is used to simulate the external conditions of the drone flight. In the virtual environment, elevation data and geographic information system (GIS) technology can be used to build a three-dimensional terrain model, add static and dynamic obstacle models (such as buildings, trees, vehicles, etc.), and add collision detection functions to the obstacles to ensure that the drone can detect collisions and respond in the simulated environment. According to the type of sensor carried by the drone (such as lidar, camera, IMU, etc.), the corresponding sensor model is configured in the virtual environment. Environmental information such as distance measurement and visual data is obtained through the sensor model. The distance between the drone and the obstacle is detected to determine whether a collision occurs. The distance formula is as follows:

[0116]

[0117] If d collision ≤r dron -r obstacle , a collision occurs, where r dron is the radius of the drone, r obstacle is the radius of the obstacle.

[0118] Real-time data is used to drive dynamic changes in the environmental sub-model, such as adjusting wind speed and direction based on real-time meteorological data. Through digital twin technology, the virtual environment and the physical environment are synchronized in real time. The effect of wind speed on the drone's speed can be expressed as:

[0119] v wind =v drone +v wind_vector (13)

[0120] Among them, v wind is the speed of the UAV in the wind field, v drone is the original speed of the drone, v wind_vector is the wind speed vector.

[0121] The dynamic changes of meteorological parameters can be simulated by probabilistic models or physical models. For example, temperature changes can be expressed as:

[0122]

[0123] Where T(t) is the temperature at time t, T0 is the initial temperature, ΔT is the temperature change amplitude, and P is the change period.

[0124] In one possible implementation, the objective function of the control sub-model is:

[0125] The objective function of the control sub-model is to minimize the total path cost of the UAV. The total path cost of the UAV is the sum of all products of the radar threat cost, strike threat cost, terrain threat cost and energy consumption cost multiplied by the corresponding weight coefficients.

[0126] Optionally, the control sub-model in this embodiment is used to achieve stable flight and mission execution for the UAV. Control objectives, such as path planning, obstacle avoidance, and attitude control, are defined based on the UAV's mission type (e.g., inspection, delivery, search, etc.). Mission-related threat models (e.g., radar threat, strike threat, terrain threat) and cost functions are pre-set, serving as the objective function of the control sub-model.

[0127] The objective function for path planning and optimization can be expressed as:

[0128]

[0129] Among them, v(x) is the total path cost of the UAV, cost r is the radar threat cost, cost m Cost is the cost of attacking the threat. e is the terrain threat cost, cost l is the energy consumption cost, ω r 、ω m 、ω e 、ω l are the weights of each cost item.

[0130] The energy cost is proportional to the flight path length and can be expressed as:

[0131]

[0132] Among them, l i is the length of segment i, and c is the energy cost per unit length.

[0133] In one possible implementation method, after the construction of the geometric sub-model, kinematic sub-model, dynamic sub-model, environmental sub-model and control sub-model is completed, virtual model calibration, real-time simulation and verification and data-driven optimization are required to ensure that the constructed drone digital twin model better meets actual needs.

[0134] The virtual simulation calibration is as follows: inputting the actual UAV operation data collected in step 101 into the virtual model, calibrating and optimizing the model so that the virtual model can accurately reflect the actual operation status of the physical UAV.

[0135] Real-time simulation and verification are: real-time simulation of the UAV in a virtual environment, simulating the operation of the UAV under different working conditions, and comparing and verifying with the actual UAV operation data to ensure the accuracy and reliability of the virtual model.

[0136] Data-driven optimization means that as the drone operates and data continues to accumulate, the drone digital twin model is continuously optimized and improved using new data to improve the model's accuracy and predictive capabilities.

[0137] In step 103, an enhanced data set is constructed based on the operating data and the virtual data, and the enhanced data set is used to train a UAV life prediction model.

[0138] In an embodiment of the present application, an enhanced data set is constructed using virtual data and collected drone operation data, and the enhanced data is used to train a drone life prediction model.

[0139] The embodiment of the present application innovatively integrates digital twin technology and deep learning models, and can perform high-precision life prediction for data-deficient drones. It is applicable to a variety of drones and has good versatility and scalability. In addition, through accurate life prediction, managers can plan the maintenance and replacement of drones in advance. Before the drone is about to reach the end of its service life, maintenance personnel are arranged and the necessary parts are prepared in a targeted manner, avoiding emergency repairs caused by sudden failures of the drone and reducing unnecessary waste of manpower, material and financial resources. At the same time, excessive maintenance is avoided, that is, unnecessary maintenance operations are avoided when the drone still has a long service life, thereby reducing maintenance costs and improving the economic benefits of the enterprise.

[0140] Furthermore, based on the lifespan prediction results of this embodiment, companies can promptly identify potential problems during the drone's lifespan and take appropriate measures, such as replacing aging components and optimizing drone operating parameters. This helps maintain stable drone operation, reduces the probability of drone failures, and improves drone reliability. High drone reliability is crucial for the communications industry, ensuring the continuity and stability of communications services, improving user experience, and enhancing companies' market competitiveness.

[0141] In one possible implementation, building an enhanced dataset based on operational data and virtual data includes:

[0142] The running data and virtual data are fused according to the preset ratio to construct an enhanced dataset.

[0143] Optionally, this embodiment utilizes a drone digital twin model to simulate the drone's operating status under different circumstances, generating a large amount of virtual data. Prior to fusion, the virtual data requires processing such as noise addition and data offset to enhance its diversity and authenticity. The collected actual operating data is then fused with the virtual data generated by the drone digital twin model in a preset ratio to form an enhanced dataset. Features reflecting the drone's health status are extracted from the training data, and the enhanced dataset is preprocessed with normalization and denoising to improve its quality.

[0144] In one possible implementation, the enhanced dataset includes the drone's operational data, health status, and remaining service life. Using the enhanced dataset, a drone life prediction model is trained, including:

[0145] A lightweight CNN model, a Transformer model, and a BiLSTM model are sequentially connected to construct a lightweight CNN-Transformer-BiLSTM model.

[0146] Taking the UAV's operating data as input and the corresponding UAV's health status and remaining service life as output, a lightweight CNN-Transformer-BiLSTM model is trained to obtain a UAV life prediction model.

[0147] Optionally, this embodiment uses a deep learning algorithm to train the enhanced dataset and construct a drone lifespan prediction model. This model combines a lightweight CNN model to extract local features, a Transformer model to capture global dependencies, and a BiLSTM model to model long-term dependencies in time series. The fully connected layer outputs the prediction results. Through these steps, the drone lifespan prediction model effectively combines local features, global dependencies, and time series information to accurately predict drone lifespan. The steps are as follows:

[0148] (1) Dataset

[0149] The enhanced data set includes the operating data, health status, and remaining service life of the drone. In this embodiment, the enhanced data is divided into a training set, a validation set, and a test set, for example, 70% of the training set, 15% of the validation set, and 15% of the test set.

[0150] (2) Model construction

[0151] (1) The lightweight CNN model includes input layer, convolution layer, activation function, and pooling layer.

[0152] The time series in the input layer are mostly sensor data with the shape of (B, T, F), where B is the batch size, T is the time steps, and F is the feature dimensions. For example, the sensor data of a drone may contain features such as position, velocity, attitude, acceleration, and temperature.

[0153] The goal of the lightweight CNN model is to extract local features from time series data. It processes time series data using one-dimensional convolution (Conv1D). The convolution kernel is (K, F, C), where K is the kernel size, F is the input feature dimension, and C is the number of output channels.

[0154]

[0155] Among them, x t+k-1,f is the fth feature of the input data at the t+k-1th time step, w k,f,c is the weight of the convolution kernel k-th position, f-th input feature and c-th output channel, b c is the bias term.

[0156] This embodiment uses the ReLU activation function: ReLU(y)=max(0,y).

[0157] This embodiment uses maximum pooling (MaxPooling1D) to reduce the time dimension, that is:

[0158]

[0159] The output is: (B, T″, C), where T″ is the time step after pooling.

[0160] (2) The Transformer model is used for global feature extraction, including input, position encoding, multi-head attention mechanism, feedforward neural network and output.

[0161] Where the input is: (B, T″, C).

[0162] The position code is:

[0163]

[0164] Among them, pos is the time step position, i is the dimension index, and d is the feature dimension.

[0165] Then in the multi-head attention mechanism, the input is: (B, T″, C).

[0166] Compute the query, key, and value:

[0167] Q=XW Q ,K=XW K ,V=XW V (twenty one)

[0168] Among them, W Q 、W K 、W V is a learnable weight matrix.

[0169] The attention score is calculated as:

[0170]

[0171] Among them, d k is the dimension of the key, divided by This is to prevent the dot product result from being too large.

[0172] Multi-Head Attention is:

[0173] MultiHead(Q,K,V)=Concat(head1,…,head h )W o (twenty three)

[0174] in, W o is the output weight matrix and h is the number of heads.

[0175] The feedforward neural network uses a two-layer fully connected network, namely:

[0176] FFN(X)=max(0,xW1+b1)W2+b2 (24)

[0177] Among them, W1 and W2 are the weights of the linear transformation, and b1 and b2 are biases.

[0178] The final output is: (B, T″, C).

[0179] (3) The BiLSTM model is used for time series modeling, with the goal of capturing long-term dependencies in time series data.

[0180] The input is: (B,T″,C).

[0181] Perform LSTM calculations, the calculation formula is:

[0182]

[0183] Among them, σ is the sigmoid function, W f is the weight matrix, b f is the bias term, tanh is the hyperbolic tangent function, ht is the hidden state at the current moment, f t For the forget gate, i t is the input gate, is the candidate memory unit, C t is the memory unit, o t is the output gate.

[0184] Then calculate the forward and backward LSTM respectively, namely:

[0185]

[0186] Concatenate the forward and backward hidden states to get:

[0187] The final output is (B, T″, 2H), where H is the number of hidden units of LSTM.

[0188] (4) The fully connected layer is used for prediction, with the input being (B, T″, 2H).

[0189] Use a fully connected layer to map the features to the output dimension, that is:

[0190] y=W f h+b f (27)

[0191] Among them, W f is the weight matrix, b f is the bias term, and h is the final output of the BiLSTM layer.

[0192] The final output is (B, T″, 1), which is the predicted lifespan value at each time step.

[0193] (3) Model training

[0194] The goal of the drone life prediction model in this embodiment is to minimize the error between the predicted value and the true value. The commonly used loss function is the mean square error (MSE), which is:

[0195]

[0196] Among them, y i is the true value, i.e. the actual life of the drone, is the model prediction value of the UAV life prediction model, and N is the number of samples.

[0197] For batch data, the loss function is calculated as:

[0198]

[0199] Where B is the batch size.

[0200] The Adam optimizer is then used to update the model parameters of the drone life prediction model. Adam is an adaptive learning rate optimization algorithm that combines the advantages of momentum and RMSProp. The main optimization process is as follows:

[0201] Calculate the gradient as follows:

[0202]

[0203] Among them, θ t is the model parameter, g t is the gradient of the loss function with respect to the parameters.

[0204] Calculate the momentum using the first-order moment estimate, namely:

[0205] m t =β1m t-1 +(1-β1)g t (31)

[0206] Among them, m t is the momentum, β1 is the momentum decay rate, usually set to 0.9.

[0207] To calculate RMSProp, we use the second-order moment estimation, namely:

[0208]

[0209] Among them, v t is the second-order moment estimate, and β2 is the decay rate, which is usually set to 0.999.

[0210] Bias correction, i.e.:

[0211]

[0212] in, and is the bias-corrected momentum.

[0213] Parameter update, namely:

[0214]

[0215] Where η is the learning rate, usually set to 0.001, ∈ is a constant, usually set to 10 -8 , for numerical stability, to prevent division by zero.

[0216] Furthermore, backpropagation is used to calculate the gradient of the loss function with respect to the model parameters. i is the true value, i.e. the actual lifespan of the drone, To predict the value for the model, here are the detailed steps:

[0217] The gradient of the output layer is:

[0218]

[0219] The gradient of the fully connected layer is:

[0220]

[0221] Among them, W f is the weight matrix, b f is the bias term, h is the input of the fully connected layer, and comes from the output of the previous layer.

[0222] The gradient of BiLSTM is calculated using the chain rule, that is:

[0223]

[0224] Among them, W f is the weight matrix.

[0225] The gradient of the LSTM unit is calculated through back propagation through time (BPTT), that is:

[0226]

[0227] Among them, t is the activation value of the output gate (i.e., the output of the sigmoid function).

[0228] Calculate the Transformer gradient, where the Transformer gradient includes the gradient of the multi-head attention and feedforward neural network, namely:

[0229]

[0230] Among them, Q is the query matrix, K is the key matrix, and V is the value matrix.

[0231] Calculate the lightweight CNN gradient, that is:

[0232]

[0233] Among them, W is the convolution kernel weight, which is used to map the input feature map to the output feature map, b is the bias term, which adds an offset to the output feature map, and x is the input feature map, which comes from the output of the previous layer.

[0234] Secondly, the Adam optimizer is used to update the model parameters, referring to formula (34).

[0235] Finally, the training process is as follows:

[0236] Forward propagation: Input data passes through CNN, Transformer, BiLSTM and fully connected layers to calculate the predicted value

[0237] Calculate the loss: Use MSE to calculate the loss L.

[0238] Backpropagation: Calculate the gradient of the loss function with respect to the model parameters.

[0239] Gradient Update: Update the model parameters using the Adam optimizer.

[0240] Repeat iteration: Repeat the above steps until the loss function converges or the maximum number of iterations is reached.

[0241] (4) Model Validation and Deployment

[0242] 1. Model validation: After model training is completed, the performance of the model needs to be tested on the validation set and the test set to ensure its generalization ability.

[0243] (1) Validation set evaluation

[0244] Input: Validation set data (B_val, T, F).

[0245] Output: Validation set prediction results (B_val,T,1).

[0246] Evaluation metrics include:

[0247] Mean square error (MSE), that is:

[0248]

[0249] Among them, y i is the true value of the i-th sample, is the predicted value of the i-th sample.

[0250] Mean absolute error (MAE), that is:

[0251]

[0252] Coefficient of determination (R 2 ),Right now:

[0253]

[0254] in, is the mean of the true values of the validation set.

[0255] (2) Test set evaluation

[0256] Input: test set data (B_teat, T, F).

[0257] Output: Test set prediction results (B_teat, T, 1).

[0258] Evaluation metrics: Use the same metrics as the validation set (MSE, MAE, R 2 )Evaluate the performance of the model on the test set.

[0259] (3) Result visualization

[0260] Plot the comparison between true and predicted values to visually analyze model performance. Plot the error distribution to analyze the source of error.

[0261] 2. Model Deployment

[0262] After the model is verified, it is deployed to the actual application environment for prediction of new data.

[0263] (1) Save the model and save the trained model parameters to a file.

[0264] (2) Model loading: loading model parameters in the deployment environment.

[0265] (3) Model reasoning, making predictions on new data:

[0266] Input: New data (B_new, T, F).

[0267] Forward propagation: Among them, x new For new data, For the prediction results.

[0268] Output: prediction result (B_new,T,1).

[0269] (4) Result post-processing: post-process the prediction results.

[0270] 3. Deployment Environment

[0271] Local deployment: Deploy the model to a local server or edge device, and use Python scripts or C++ libraries for inference.

[0272] Cloud deployment: Deploy the model to a cloud platform (such as AWS, Azure, and Google Cloud) and provide an application programming interface for external calls.

[0273] Embedded deployment: Deploy the model to the drone’s embedded device to predict lifespan in real time.

[0274] 4. Continuous monitoring and updates

[0275] Monitor model performance: Regularly evaluate the model's performance on new data to ensure its predictive accuracy.

[0276] Model update: If model performance degrades, retrain the model with new data and update the deployment.

[0277] In one possible implementation, after using the enhanced dataset to train the drone life prediction model, the method may further include:

[0278] Using the trained UAV life prediction model, the remaining useful life (RUL) of the UAV is predicted and the prediction results are output. At the same time, the prediction results are evaluated using actual data, and the model parameters of the UAV life prediction model are adjusted to improve the prediction accuracy.

[0279] The present application provides a method for predicting the life of a drone. The method collects the drone's operating data, including flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status, and mechanical stress. Based on the operating data, a drone digital twin model is constructed, and the drone digital twin model is used to simulate the drone's operating status under different conditions to generate virtual data. The drone digital twin model includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model, and a control sub-model. An enhanced data set is constructed based on the operating data and the virtual data, and the enhanced data set is used to train the drone life prediction model. By constructing a drone digital twin model, the present application can obtain the drone's operating status under different conditions in real time, thereby expanding the test data. The acquired operating data and virtual data are then used to construct an enhanced data set and train the drone life prediction model, which can improve the prediction accuracy and reliability.

[0280] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0281] The following are device embodiments of the present application. For details not fully described therein, please refer to the corresponding method embodiments described above.

[0282] Figure 3 The following is a schematic diagram of the structure of the UAV life prediction device provided by the embodiment of the present application. For the sake of convenience, only the parts related to the embodiment of the present application are shown, which are detailed as follows:

[0283] like Figure 3 As shown, the UAV life prediction device 3 includes:

[0284] The data acquisition module 31 is used to collect the operation data of the UAV, including flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status and mechanical stress;

[0285] A simulation data generation module 32 is used to build a UAV digital twin model based on the operating data, and use the UAV digital twin model to simulate the operating status of the UAV under different conditions to generate virtual data. The UAV digital twin model includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model, and a control sub-model;

[0286] The model training process 33 is used to construct an enhanced data set based on the operating data and the virtual data, and use the enhanced data set to train the UAV life prediction model.

[0287] The present application provides a drone life prediction device. The device collects the drone's operating data, including flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status, and mechanical stress. Based on the operating data, a drone digital twin model is constructed, and the drone digital twin model is used to simulate the drone's operating status under different conditions to generate virtual data. The drone digital twin model includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model, and a control sub-model. An enhanced data set is constructed based on the operating data and virtual data, and the enhanced data set is used to train the drone life prediction model. By constructing a drone digital twin model, the present application can obtain the drone's operating status under different conditions in real time, thereby expanding the test data. The acquired operating data and virtual data are then used to construct an enhanced data set and train the drone life prediction model, which can improve the prediction accuracy and reliability.

[0288] In one possible implementation, the simulation data generation module may be used to:

[0289] The running data and virtual data are fused according to the preset ratio to construct an enhanced dataset.

[0290] In one possible implementation, the process of constructing the geometric sub-model is as follows:

[0291] The geometric parameters, actual size and shape of the drone's propeller, as well as the geometric positions of each component, are input into the 3D modeling software to construct a geometric sub-model of the drone;

[0292] Import the UAV's geometric sub-model into the Unreal Engine and optimize the UAV's geometric sub-model by setting the skeleton hierarchy and physical properties.

[0293] In one possible implementation, the kinematic submodel is constructed as follows:

[0294] Establish a body coordinate system and an inertial coordinate system. The body coordinate system uses the center of mass of the drone as its origin, while the inertial coordinate system uses the earth as its reference coordinate. These are used to describe the absolute position and attitude of the drone.

[0295] Using the flight attitude and flight position of the UAV, the kinematic equation of the UAV is established. The kinematic equation includes the translational motion equation and the rotational motion equation.

[0296] The Euler method is used to solve the kinematic equations to obtain the position and attitude of the UAV at each time step.

[0297] In one possible implementation, the dynamics sub-model includes a motor dynamics unit, a rigid body attitude dynamics unit, and a rigid body position dynamics unit. The construction process of the dynamics sub-model is as follows:

[0298] Use the motion trajectory of each motor in the drone to build a motor dynamics unit;

[0299] Use the UAV's flight attitude to construct a rigid body attitude dynamics unit;

[0300] The flight position of the UAV is used to construct a rigid body position dynamics unit.

[0301] In one possible implementation, the objective function of the control sub-model is:

[0302] The objective function of the control sub-model is to minimize the total path cost of the UAV. The total path cost of the UAV is the sum of all products of the radar threat cost, strike threat cost, terrain threat cost and energy consumption cost multiplied by the corresponding weight coefficients.

[0303] In one possible implementation, the augmented dataset includes the drone’s operational data, health status, and remaining useful life. The model training module can be used to:

[0304] A lightweight CNN model, a Transformer model, and a BiLSTM model are sequentially connected to construct a lightweight CNN-Transformer-BiLSTM model.

[0305] Taking the UAV's operating data as input and the corresponding UAV's health status and remaining service life as output, a lightweight CNN-Transformer-BiLSTM model is trained to obtain a UAV life prediction model.

[0306] Figure 4 Schematic diagram of the terminal provided in the embodiment of the present application. Figure 4 As shown, the terminal 4 of this embodiment includes: a processor 40, a memory 41, and a computer program 42 stored in the memory 41 and executable on the processor 40. When the processor 40 executes the computer program 42, the steps in each of the above-mentioned UAV life prediction method embodiments are implemented, such as Figure 1Alternatively, when the processor 40 executes the computer program 42, the functions of the modules / units in the above-mentioned device embodiments are realized, for example, Figure 3 The functions of each module are shown.

[0307] For example, the computer program 42 may be divided into one or more modules / units, which are stored in the memory 41 and executed by the processor 40 to complete the present application. The one or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program 42 in the terminal 4. For example, the computer program 42 may be divided into Figure 3 The modules shown.

[0308] The terminal 4 can be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal 4 can include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will understand that Figure 4 It is only an example of terminal 4 and does not constitute a limitation on terminal 4. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal may also include input and output devices, network access devices, buses, etc.

[0309] The processor 40 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0310] The memory 41 may be an internal storage unit of the terminal 4, such as a hard disk or memory of the terminal 4. The memory 41 may also be an external storage device of the terminal 4, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash memory card, etc. equipped on the terminal 4. Furthermore, the memory 41 may include both an internal storage unit of the terminal 4 and an external storage device. The memory 41 is used to store the computer program and other programs and data required by the terminal. The memory 41 may also be used to temporarily store data that has been output or is about to be output.

[0311] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0312] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0313] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel 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.

[0314] In the embodiments provided in this application, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0315] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0316] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0317] If the integrated module / unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned drone life prediction method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. It should be noted that the content of the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0318] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A method for predicting the life of an unmanned aerial vehicle, characterized in that: include: Collecting the UAV's operational data, including flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status, and mechanical stress; Based on the operating data, a digital twin model of the drone is constructed, and the operating state of the drone under different conditions is simulated using the digital twin model of the drone to generate virtual data. The digital twin model of the drone includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model, and a control sub-model. An enhanced data set is constructed based on the operating data and the virtual data, and a UAV life prediction model is trained using the enhanced data set.

2. The UAV life prediction method according to claim 1, characterized in that: The constructing of an enhanced data set based on the operating data and the virtual data includes: The operating data and the virtual data are fused according to a preset ratio to construct the enhanced data set.

3. The UAV life prediction method according to claim 1, characterized in that: The construction process of the geometric sub-model is as follows: Inputting the geometric parameters, actual size and shape of the propeller of the UAV and the geometric positions of each component into the 3D modeling software to construct a geometric sub-model of the UAV; The geometric sub-model of the UAV is imported into a virtual engine, and the geometric sub-model of the UAV is optimized by setting a skeleton hierarchy and physical properties.

4. The UAV life prediction method according to claim 1, characterized in that: The construction process of the kinematic sub-model is as follows: Establish a body coordinate system and an inertial coordinate system. The body coordinate system uses the center of mass of the UAV as its origin, and the inertial coordinate system uses the Earth as its reference coordinate to describe the absolute position and attitude of the UAV. Using the flight attitude and flight position of the UAV, a kinematic equation of the UAV is established, wherein the kinematic equation includes a translational motion equation and a rotational motion equation; The kinematic equations are solved using the Euler method to obtain the position and attitude of the UAV at each time step.

5. The UAV life prediction method according to claim 1, characterized in that: The dynamics sub-model includes a motor dynamics unit, a rigid body posture dynamics unit, and a rigid body position dynamics unit. The construction process of the dynamics sub-model is as follows: Use the motion trajectory of each motor in the drone to build a motor dynamics unit; Use the UAV's flight attitude to construct a rigid body attitude dynamics unit; The flight position of the UAV is used to construct a rigid body position dynamics unit.

6. The UAV life prediction method according to claim 1, characterized in that: The objective function of the control sub-model is: The objective function of the control sub-model is to minimize the total path cost of the UAV. The total path cost of the UAV is the sum of all products of the radar threat cost, the strike threat cost, the terrain threat cost and the energy consumption cost multiplied by the corresponding weight coefficients.

7. The UAV life prediction method according to claim 1, characterized in that: The enhanced data set includes the operating data, health status, and remaining service life of the UAV. The use of the enhanced data set to train the UAV life prediction model includes: A lightweight CNN model, a Transformer model, and a BiLSTM model are sequentially connected to construct a lightweight CNN-Transformer-BiLSTM model. The UAV operation data is used as input, and the health status and remaining service life of the corresponding UAV is used as output to train the lightweight CNN-Transformer-BiLSTM model to obtain the UAV life prediction model.

8. A drone life prediction device, characterized in that: include: A data acquisition module is used to collect the operating data of the UAV, including flight time, flight attitude, flight position, motion trajectory, environmental parameters, component status and mechanical stress; A simulation data generation module is used to build a UAV digital twin model based on the operating data, and use the UAV digital twin model to simulate the operating status of the UAV under different conditions to generate virtual data. The UAV digital twin model includes a geometric sub-model, a kinematic sub-model, a dynamic sub-model, an environmental sub-model, and a control sub-model; The model training module is used to construct an enhanced data set based on the operating data and the virtual data, and use the enhanced data set to train a UAV life prediction model.

9. A terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the drone life prediction method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the drone life prediction method as described in any one of claims 1 to 7 are implemented.

Images