Unmanned aerial vehicle landing gear fatigue life prediction method, device, equipment and storage medium

Abstract

The application discloses an unmanned aerial vehicle landing gear fatigue life prediction method and device, equipment and a storage medium, relates to the unmanned aerial vehicle field, and comprises the following steps: obtaining original flight parameter data collected by a target flight parameter recorder carried by a target unmanned aerial vehicle, performing feature dimension reduction processing on the original flight parameter data to obtain target flight parameter data, and determining an input feature set corresponding to each prediction target based on the target flight parameter data; determining target load time series data corresponding to the target flight parameter data based on the input feature set by using a target load prediction model; applying the target load time series data to a target three-dimensional model, performing stress analysis on a target stress point corresponding to a target landing gear of the target unmanned aerial vehicle to determine single-flight total damage corresponding to the target flight parameter data, and determining fatigue life corresponding to the target landing gear based on the single-flight total damage; and the target three-dimensional model is a full-size three-dimensional model of the target landing gear. The application realizes accurate prediction of the fatigue life of the unmanned aerial vehicle landing gear.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicles (UAVs), and particularly to a method, apparatus, equipment, and storage medium for predicting the fatigue life of UAV landing gear. Background Technology

[0002] During high-intensity, complex flight missions and landings, modern unmanned aerial vehicles (UAVs) must withstand severe multi-directional impact loads and long-term dynamic cyclic loads on their landing gear structures. This complex stress environment easily leads to metal fatigue, loosening of connectors, or microcracks in the landing gear. Failure to monitor and assess these structural damages in a timely and accurate manner can potentially cause the landing gear to break or fail completely in subsequent missions, seriously jeopardizing flight safety and causing mission failure. In recent years, as aerospace technology has evolved towards Prognostics and Health Management (PHM) systems, traditional scheduled maintenance or reactive maintenance models can no longer meet the high-frequency, high-reliability maintenance requirements of UAVs. Currently, the field of UAV landing gear maintenance and health monitoring faces three major technical challenges: difficulty in near-real-time assessment of structural damage, lack of precise quantification of fatigue life, and poor engineering feasibility of technical processes. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a method, apparatus, device, and storage medium for predicting the fatigue life of UAV landing gear, capable of accurately predicting the fatigue life of UAV landing gear. The specific solution is as follows: In a first aspect, this application discloses a method for predicting the fatigue life of a UAV landing gear, including: The raw flight parameter data collected by the target flight parameter recorder on the target UAV is acquired, and the raw flight parameter data is subjected to feature dimensionality reduction processing to obtain target flight parameter data. Based on the target flight parameter data, the input feature set corresponding to each predicted target is determined. The predicted target includes the landing gear lateral force at the landing gear touchdown point, the landing gear yaw force at the landing gear touchdown point, the wheel center yaw load, and the vertical load of the target UAV. The target load time series data corresponding to the target flight parameter data is determined using the target load prediction model based on the input feature set; the target load prediction model is a load prediction model based on an ensemble learning framework architecture of stacked ensemble networks; The target load time series data is applied to the target three-dimensional model, and stress analysis is performed on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage per sortie corresponding to the target flight parameter data, and the fatigue life corresponding to the target landing gear is determined based on the total damage per sortie; the target three-dimensional model is a full-size three-dimensional model of the target landing gear.

[0004] Optionally, before acquiring the raw flight parameter data collected by the target flight parameter recorder mounted on the target UAV, the method further includes: The target running working directory is received through the interface event-driven mechanism, and it is verified whether a model file with the target suffix already exists in the target running working directory; If the target working directory does not contain a model file with the target suffix, then the target resource path parsing function is used to search the target underlying resource library to see if there is a target source model file corresponding to the target suffix. If the target source model file exists in the target underlying resource library, the target source model file is saved to the target running working directory using the target copy command.

[0005] Optionally, the step of performing feature dimensionality reduction processing on the original flight parameter data to obtain the target flight parameter data includes: Determine the degree of nonlinear dependence between each flight parameter variable and the target payload variable in the original flight parameter data, and perform feature dimensionality reduction on the original flight parameter data based on the degree of nonlinear dependence to obtain the first flight parameter data; Determine the target variance corresponding to each flight parameter variable in the first flight parameter data, and perform feature dimensionality reduction on the first flight parameter data based on the preset variance threshold and the target variance to obtain the second flight parameter data; The target flight parameter data is obtained by filtering the flight parameter variables in the second flight parameter data using the target correlation matrix.

[0006] Optionally, the step of using the target payload prediction model to determine the target payload time-series data corresponding to the target flight parameter data based on the input feature set includes: The input feature set is input into the target load prediction model, so as to generate the load prediction sequence of each intermediate layer corresponding to the input feature set by using the target base learners of the first network layer of the target load prediction model. Using the second network layer of the target load prediction model, the load prediction sequences of each intermediate layer are matrix-concatenated to obtain the target load time series data corresponding to the target flight parameter data.

[0007] Optionally, the target base learner includes a first base learner based on an extreme gradient boosting tree, a second base learner based on a random forest, a third base learner based on an extremely random tree, a fourth base learner based on a bagging model, and a fifth base learner based on a categorical gradient boosting algorithm.

[0008] Optionally, before applying the target load time series data to the target three-dimensional model and performing stress analysis on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage per sortie corresponding to the target flight parameter data, the method further includes: A target three-dimensional model corresponding to the target landing gear is constructed using target finite element software, and unit loads in four directions are applied to the target stress points of the target three-dimensional model to determine the stress tensor component response coefficients corresponding to the target stress points. Based on the stress tensor component response coefficients, a unit load stress response coefficient matrix corresponding to the target landing gear is constructed.

[0009] Optionally, the step of applying the target load time series data to the target three-dimensional model, performing stress analysis on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage per sortie corresponding to the target flight parameter data, and determining the fatigue life corresponding to the target landing gear based on the total damage per sortie, includes: Using the target 3D model, the target load time series data is converted into the actual stress components of the target stress point corresponding to the target landing gear of the target UAV at the target time based on the unit load stress response coefficient matrix. Based on the signed fourth strength theory, the real stress components are converted into a target stress time series, and the fatigue load cycle corresponding to the target flight parameter data is determined based on the target stress time series using a preset rainflow counting algorithm. The fatigue load cycle is corrected using the target boundary criterion equation, and the total damage per sortie corresponding to the target flight parameter data is determined based on the corrected fatigue load cycle and Miner's linear cumulative damage theory; the target boundary criterion equation is the fatigue failure boundary criterion equation determined based on the Goodman curve. The global cumulative damage corresponding to the target landing gear is determined based on the historical cumulative damage data corresponding to the target landing gear and the total damage of a single sortie. The fatigue life corresponding to the target landing gear is determined based on the current equivalent damage evolution rate of the target UAV and the global cumulative damage.

[0010] Secondly, this application discloses a device for predicting the fatigue life of a drone landing gear, comprising: The feature set acquisition module is used to acquire the raw flight parameter data collected by the target flight parameter recorder carried by the target UAV, perform feature dimensionality reduction processing on the raw flight parameter data to obtain target flight parameter data, and determine the input feature set corresponding to each predicted target based on the target flight parameter data; the predicted targets include the landing gear lateral force at the landing gear touchdown point, the landing gear yaw force at the landing gear touchdown point, the wheel center yaw load, and the vertical load of the target UAV. The load data acquisition module is used to determine the target load time series data corresponding to the target flight parameter data based on the input feature set using the target load prediction model; the target load prediction model is a load prediction model based on an ensemble learning framework architecture of stacked ensemble networks. The fatigue life prediction module is used to apply the target load time series data to the target three-dimensional model, perform stress analysis on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage of a single sortie corresponding to the target flight parameter data, and determine the fatigue life of the target landing gear based on the total damage of a single sortie; the target three-dimensional model is a full-size three-dimensional model of the target landing gear.

[0011] Thirdly, this application discloses an electronic device, including: Memory, used to store computer programs; A processor is used to execute the computer program to implement the aforementioned method for predicting the fatigue life of UAV landing gear.

[0012] Fourthly, this application discloses a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned method for predicting the fatigue life of UAV landing gear.

[0013] In this application, when predicting the fatigue life of a UAV landing gear, raw flight parameter data collected by a target flight parameter recorder mounted on the target UAV is acquired. Feature dimensionality reduction processing is performed on the raw flight parameter data to obtain target flight parameter data. Based on the target flight parameter data, an input feature set corresponding to each predicted target is determined. The predicted targets include the lateral force at the landing point, the yaw force at the landing point, the wheel center yaw load, and the vertical load of the target UAV landing gear. A target load prediction model is used to determine the target load time series data corresponding to the target flight parameter data based on the input feature set. The target load prediction model is a load prediction model based on an ensemble learning framework architecture using a stacked ensemble network. The target load time series data is applied to the target 3D model, and stress analysis is performed on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage per sortie corresponding to the target flight parameter data. Based on the total damage per sortie, the fatigue life corresponding to the target landing gear is determined. The target 3D model is a full-size 3D model of the target landing gear. As can be seen, this application cleverly utilizes the inherent conventional airborne flight parameters of the UAV flight parameter recorder and constructs a virtual sensor through software and hardware algorithms. Under the condition of not damaging the original physical structure of the aircraft and not increasing any effective payload, it achieves high-precision load spectrum perception. By performing feature dimensionality reduction on the original flight parameter data, the target load prediction model based on the ensemble learning architecture of stacked ensemble network is used to process the target flight parameter data to identify the influence of transient spatial pose on the load. It has extremely strong regularization ability and significantly reduces computational complexity. It not only reduces the inference computing power requirement from the GPU level to the ordinary CPU level of a portable terminal, but also greatly improves the prediction robustness and generalization ability, and achieves accurate prediction of the fatigue life of UAV landing gear. Attached Figure Description

[0014] 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0015] Figure 1 This application discloses a flowchart of a method for predicting the fatigue life of a drone landing gear. Figure 2 This is a schematic diagram of a specific UAV landing gear fatigue life prediction system architecture disclosed in this application; Figure 3 This is a schematic diagram illustrating the Pearson correlation coefficient between a specific load and flight parameters disclosed in this application. Figure 4 This is a schematic diagram of a specific stacked ensemble learning architecture disclosed in this application; Figure 5 This is a schematic diagram of the structure of a drone landing gear fatigue life prediction device disclosed in this application. Figure 6 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Modern unmanned aerial vehicles (UAVs) must withstand severe multi-directional impact loads and long-term dynamic cyclic loads during high-intensity, complex flight missions and landings. This complex stress environment easily leads to metal fatigue, loosening of connectors, or microcracks in the landing gear. Failure to monitor and assess these structural damages in a timely and accurate manner can potentially cause the landing gear to break or fail completely in subsequent missions, seriously endangering flight safety and leading to mission failure. In recent years, with the evolution of aerospace technology towards predictive and health management systems, traditional periodic or reactive maintenance models can no longer meet the high-frequency, high-reliability maintenance requirements of UAVs. Currently, the field of UAV landing gear maintenance and health monitoring faces three major technical challenges: difficulty in near-real-time assessment of structural damage, lack of precise quantification of fatigue life, and poor engineering feasibility of technical processes. To address these technical problems, this application discloses a method for predicting the fatigue life of UAV landing gear, which can accurately predict the fatigue life of UAV landing gear.

[0018] See Figure 1 As shown, this embodiment of the invention discloses a method for predicting the fatigue life of a UAV landing gear, including: Step S11: Obtain the raw flight parameter data collected by the target flight parameter recorder carried by the target UAV, perform feature dimensionality reduction processing on the raw flight parameter data to obtain target flight parameter data, and determine the input feature set corresponding to each predicted target based on the target flight parameter data; the predicted target includes the landing gear lateral force at the landing gear touchdown point, the landing gear heading force at the landing gear touchdown point, the wheel center heading load and the vertical load of the target UAV.

[0019] In this embodiment, the architecture diagram of the UAV landing gear fatigue life prediction system is as follows: Figure 2As shown, the system includes at least one processor, at least one memory, a data communication interface, and a human-machine interface display peripheral. The memory pre-stores an application instruction set built with PyQt5 (a cross-platform graphical user interface application development framework based on Python) and pre-trained offline machine learning serialization model files (such as .pkl files continuously updated and saved via joblib or pickle). The processor calls the instructions in the memory to execute the entire fatigue life prediction workflow. This system is deployed on a ground station maintenance computing device or a portable maintenance terminal, receiving historical flight status time-series data (which needs to be converted to plain text .csv or table .xls / .xlsx format) downloaded or transmitted by the UAV flight parameter recorder via the data communication interface. Before acquiring the raw flight parameter data collected by the target flight parameter recorder on the target UAV, the process includes: receiving the target running working directory via interface event-driven processing and verifying whether a model file with the target suffix already exists in the target running working directory; if no model file with the target suffix exists in the target running working directory, then using the target resource path parsing function to search for the target source model file corresponding to the target suffix in the target underlying resource library; if the target source model file exists in the target underlying resource library, then using the target copy command to save the target source model file to the target running working directory.

[0020] In one specific implementation, the UAV landing gear fatigue life prediction system first receives the user-selected system operating directory via an interface event-driven mechanism (signal-slot mechanism). Subsequently, the system triggers automated deployment verification logic to load four pre-trained regression prediction models. The system implements a three-layer progressive verification: the first layer verifies whether a model file with the target suffix already exists in the working directory to prevent duplicate I / O overwriting; the second network layer uses its internal resource path resolution function to adapt to the absolute path of the packaging environment and search the underlying resource library for the source model file; the third layer, after meeting the conditions, calls a copy instruction with metadata retention attributes to load the model into the current working directory. This mechanism ensures that even if the user starts the evaluation task from any disk path, the core prediction model engine can be correctly scheduled and mounted into memory, thus providing end-to-end data traceability and foundational support.

[0021] In this embodiment, each takeoff and landing of the UAV generates raw flight parameter data with dimensions of hundreds or even thousands of dimensions. In one specific implementation, 52 core flight parameters that may affect the payload can be predefined. To prevent the machine learning model from overfitting due to multicollinearity and noise interference, after obtaining the raw flight parameter data, this embodiment performs feature dimensionality reduction processing on the raw flight parameter data to obtain target flight parameter data. Specifically, this may include: determining the degree of nonlinear dependence between each flight parameter variable in the raw flight parameter data and the target payload variable, and performing feature dimensionality reduction on the raw flight parameter data based on the degree of nonlinear dependence to obtain first flight parameter data; determining the target variance corresponding to each flight parameter variable in the first flight parameter data, and performing feature dimensionality reduction on the first flight parameter data based on a preset variance threshold and the target variance to obtain second flight parameter data; and using a target correlation matrix to filter each flight parameter variable in the second flight parameter data to obtain the target flight parameter data.

[0022] In one specific implementation, the feature reduction process sequentially includes: nonlinear filtering based on mutual information, constant feature removal based on variance thresholding, and collinear feature elimination based on Pearson correlation coefficient. The system first measures each flight parameter variable. With target load variables The degree of nonlinear interdependence between them. Mutual information is calculated based on information entropy theory, and the formula is as follows: ; in, Let be the joint probability distribution of the two variables. If the two variables are completely independent, The system iterates through and calculates 52 input features, setting a threshold constant (preferably 0.1). All mutual information values... Flight parameter variables were determined by the system to contribute very little to the information of the target payload and were directly removed. Through this level of filtering, the feature dimension was reduced from 52 dimensions to about 40 dimensions (the features removed may vary depending on the target payload).

[0023] For the features retained after the first-level filtering, the system further evaluates their activity over time. If a sensor parameter is in a constant dead zone throughout the landing cycle, it is meaningless for the regression mapping of the dynamic payload. The system calculates each retained feature vector. variance : ; System set variance threshold Flight parameter vectors with variance values ​​less than 1 are removed. This step further reduces the features to approximately 30 dimensions.

[0024] To address the multicollinearity problem caused by data from different sensors representing the same physical phenomenon (such as airspeed from different redundant channels), the system establishes a global pairwise correlation coefficient matrix that preserves key features. (Correlation coefficients) The calculation formula is: ; The system retrieves the matrix, and once the absolute correlation coefficient between two flight parameters is found... This indicates the presence of severe redundancy. The system retains the parameter with the higher correlation to the target load and forcibly removes the other. Finally, after three levels of dimensionality reduction, for the four predicted targets—landing gear touchdown lateral force (FX), touchdown directional force (FY), wheel center directional load (PY), and vertical load (PZ)—14-dimensional, 7-dimensional, 8-dimensional, and 18-dimensional core flight parameters without redundancy are selected as the optimal input feature sets, respectively. This method preserves the physical interpretability of the features to the greatest extent possible. Figure 3 The diagram shows a specific Pearson correlation coefficient between load and flight parameters.

[0025] Step S12: Determine the target load time series data corresponding to the target flight parameter data based on the input feature set using the target load prediction model; the target load prediction model is a load prediction model based on an ensemble learning framework architecture of stacked ensemble networks.

[0026] In this embodiment, the target load prediction model is used to determine the target load time series data corresponding to the target flight parameter data based on the input feature set. This includes: inputting the input feature set into the target load prediction model, and using the target base learners of the first network layer of the target load prediction model to generate the intermediate layer load prediction sequences corresponding to the input feature set; and using the second network layer of the target load prediction model to perform matrix concatenation on the intermediate layer load prediction sequences to obtain the target load time series data corresponding to the target flight parameter data. The target base learners include a first base learner based on an extreme gradient boosting tree, a second base learner based on a random forest, a third base learner based on an extremely random tree, a fourth base learner based on a bagging algorithm, and a fifth base learner based on a categorical gradient boosting algorithm. By deeply understanding the mechanical nature of the load being mainly constrained by transient spatial pose, this embodiment decisively abandons the seemingly advanced but actually redundant Long Short-Term Memory network. After supplementing the front end with a triple hard screening of mutual information, variance, and Pearson to filter out collinear noise, the back end is supplemented with XGBoost and multi-layer stacked tree models for multi-path parallel regression. It can be deduced that the method provided by this embodiment has significantly reduced computational complexity and has extremely strong regularization ability. It not only reduces the inference computing power requirement from the GPU level to the ordinary CPU level of a portable terminal, but also greatly improves the prediction robustness and generalization ability. By combining stepwise dimensionality reduction and ensemble learning, it solves the problems of low computational efficiency and easy overfitting of traditional deep neural networks.

[0027] In one specific implementation, such as Figure 2 As shown, the flight parameter matrix after dimensionality reduction and cleaning is fed into the machine learning inference engine. This embodiment uses an ensemble learning framework with extremely nonlinear fitting capabilities and high training efficiency to construct the target load prediction model, specifically covering Extreme Gradient Boosting Tree (XGBoost) and Stacking ensemble networks. Figure 4As shown, the first network layer (base learners) consists of five heterogeneous and powerful models operating in parallel: Extreme Gradient Boosting Tree (XGBoost), Random Forest, ExtraTrees, Bagging, and CatBoost. The system synchronously inputs the dimensionality-reduced flight parameter matrix into these five base learners, with each base learner generating a set of intermediate layer load prediction sequences. The second network layer (meta learner) uses a linear regression model. The system concatenates the prediction sequences output by the five base learners in the first network layer into a matrix, feeding this as new input features into the linear regression model in the second network layer. Leveraging its fast convergence, the linear regression model redistributes weights and performs secondary learning on the prediction results of multiple base models, ultimately outputting smooth and accurate time-series data of the target loads (FX, FY, PY, PZ). This module not only overcomes the poor generalization ability of a single model, but also guides the continuous updating of training parameters by jointly monitoring the loss function based on curve fitting degree (R2), mean squared error (MSE), and mean absolute error (MAE), thus ensuring the robustness of the field prediction.

[0028] Step S13: Apply the target load time series data to the target three-dimensional model, perform stress analysis on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage per sortie corresponding to the target flight parameter data, and determine the fatigue life corresponding to the target landing gear based on the total damage per sortie; the target three-dimensional model is a full-size three-dimensional model of the target landing gear.

[0029] In this embodiment, before applying the target load time series data to the target 3D model and performing stress analysis on the target stress points corresponding to the target UAV's landing gear to determine the total damage per sortie corresponding to the target flight parameter data, the method further includes: constructing a target 3D model corresponding to the target landing gear using target finite element software, and applying unit loads in four directions to the target stress points of the target 3D model to determine the stress tensor component response coefficients corresponding to the target stress points, and constructing a unit load stress response coefficient matrix corresponding to the target landing gear based on the stress tensor component response coefficients. In a specific implementation, during the system offline construction stage (virtual part), a full-size 3D model of the landing gear is pre-constructed using Abaqus finite element software, and unit loads in four directions are applied to the key stress weak points of the landing gear (i.e., target stress points, such as the maximum principal stress node at the connection between the titanium alloy component and the composite material strut). After static solution, the six stress tensor component response coefficients corresponding to the node are extracted, forming a 4×6 unit load stress response coefficient matrix C.

[0030] In this embodiment, target load time-series data is applied to the target 3D model, and stress analysis is performed on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage per sortie corresponding to the target flight parameter data. Based on the total damage per sortie, the fatigue life of the target landing gear is determined. This includes: using the target 3D model, converting the target load time-series data into the true stress components of the target stress points corresponding to the target landing gear of the target UAV at the target time based on the unit load stress response coefficient matrix; converting the true stress components into a target stress time-series sequence based on the signed fourth strength theory, and using a preset rainflow counting algorithm to determine the fatigue load cycle corresponding to the target flight parameter data based on the target stress time-series sequence; correcting the fatigue load cycle using the target boundary criterion equation, and determining the total damage per sortie corresponding to the target flight parameter data based on the corrected fatigue load cycle and Miner's linear cumulative damage theory; the target boundary criterion equation is the fatigue failure boundary criterion equation determined based on the Goodman curve; determining the global cumulative damage corresponding to the target landing gear based on the historical cumulative damage data and the total damage per sortie, and determining the fatigue life of the target landing gear based on the current equivalent damage evolution rate of the target UAV and the global cumulative damage.

[0031] In one specific implementation, during the online prediction phase, the system obtains the current time. The four physical load column vectors of the predicted output The system executes matrix multiplication logic: ; Through linear superposition matrix operations, the system transforms the four-dimensional independent loads into the loads at the weak node. The six true stress components at time: three normal stresses and three shear stresses .

[0032] To establish a scalar benchmark for fatigue calculations while preserving the alternating tensile and compressive characteristics of the load to meet the waveform requirements of subsequent rainflow counting, the system employs a signed fourth strength theory (applicable to tough materials such as titanium alloys). First, the system transforms the six-dimensional stress tensor into one-dimensional equivalent absolute stress values: ; Subsequently, the system extracts the sign of the hydrostatic stress (or the principal stress with the largest absolute value) at the node at the current moment. The underlying calculation formula called by the code is: ; The generated signed equivalent stress time series It can accurately reflect the zero-crossing fluctuations of the load and is used as the sole input for subsequent rainflow counting.

[0033] The system is sensitive to highly fluctuating The rainflow counting algorithm is executed sequentially to extract fatigue load cycles. To improve the efficiency of the algorithm, the system first performs peak and trough filtering to remove all intermediate transition points and repetition points that are continuously monotonically increasing or decreasing, and only retains extreme points (alternating peak and trough arrays).

[0034] Then, the core three-point verification rule is executed: the system takes two adjacent stress intervals formed by three consecutive extreme points, and calculates the absolute amplitude of the first interval. The absolute amplitude of the second interval .like If the first interval is completely contained within the second interval, the system determines that it is a complete closed load cycle. The system extracts this cycle and records its stress amplitude. and mean stress =(peak value + valley value) / 2. The system then removes the two extreme points that form a loop from the array, connects the two ends of the breakpoint, and repeats this logic until the remaining nodes in the array can no longer form a closed loop.

[0035] Because the fatigue life SN curves (stress-life curves) obtained in the laboratory for materials are all in symmetrical cycles (mean stress is zero, stress ratio) The measurements were taken under conditions where the rainflow extracted the vast majority of cycles, which were asymmetric cycles. Therefore, the system calls the Goodman fatigue failure boundary criterion equation to perform amplitude conversion: ; The system will extract each group Converted to equivalent symmetrical cyclic stress amplitude : ; in, This represents the ultimate tensile strength of the structural material (e.g., 905 MPa for internal curing of the TC4 titanium alloy system).

[0036] For each corrected equivalent stress amplitude The system first performs fatigue limit filtering: if Less than the fatigue limit threshold of the material If the small high-frequency oscillation does not cause substantial fatigue damage, it is skipped directly, thereby significantly reducing the computational power consumption of invalid logarithmic operations.

[0037] for The system substitutes the effective cycle into the logarithmic empirical equation of the SN curve of the material solidified inside (e.g., This allows for the reverse calculation of the total number of life cycles required for the structure to withstand the stress cycle until it fractures. .

[0038] According to Miner's linear fatigue cumulative damage theory, the relative damage to the landing gear caused by this single cycle is... .

[0039] The system extracts all data from a single flight. The damage from each effective cycle is algebraically summed to obtain the total damage per sortie for the current sortie. : ; Subsequently, the system reads historical cumulative damage data from the local continuously updated log. Calculate the global cumulative damage up to the end of this sortie: ; The system defines the fatigue failure threshold as follows: The remaining serviceable flight life is derived based on the equivalent damage evolution rate of the current sortie. All calculated data (single-slot damage, expected life, cumulative damage, and remaining life) are ultimately fed back to the read-only text control in the system's GUI interface for rendering and display, and are automatically overwritten and updated in TXT format in the global work audit log. Through the above process, this embodiment establishes a fully integrated automatic execution engine from flight parameter import to fatigue life output, freeing maintenance personnel from the cumbersome Abaqus finite element operations, rainflow counting programming, and damage formula conversion. The system's underlying layer can automatically perform tensor superposition, criterion equivalence, and Miner's theory accumulation within milliseconds, while the outer layer directly provides the operator with intuitive visual charts and remaining serviceable flight life values. This one-stop closed-loop design improves the feasibility of technology implementation and provides a reference solution for the intelligent and dynamic health assessment of UAV formations. By converting the virtual macroscopic four-dimensional physical loads predicted at the front end into microscopic damage at the bottom layer of the physical landing gear structure, a virtual-physical fusion method for predicting the fatigue life of UAV landing gear is achieved. This reduces the high maintenance costs and failure risks of traditional field physical sensors. Instead of relying on large-area physical strain gauge arrays, it cleverly utilizes the inherent onboard conventional flight parameters (real) and finite element prior response matrices (virtual) of the UAV flight parameter recorder to construct a virtual sensor through software and hardware algorithms. This transformation logic achieves load spectrum sensing accuracy comparable to or even exceeding that of physical measurements without damaging the original physical structure of the aircraft or adding any payload, significantly reducing the monitoring budget for the entire life cycle of the aircraft.

[0040] As can be seen, this application cleverly utilizes the inherent conventional airborne flight parameters of the UAV flight parameter recorder and constructs a virtual sensor through software and hardware algorithms. Under the condition of not damaging the original physical structure of the aircraft and not increasing any effective payload, it achieves high-precision load spectrum perception. By performing feature dimensionality reduction on the original flight parameter data, the target load prediction model based on the ensemble learning architecture of stacked ensemble network is used to process the target flight parameter data to identify the influence of transient spatial pose on the load. It has extremely strong regularization ability and significantly reduces computational complexity. It not only reduces the inference computing power requirement from the GPU level to the ordinary CPU level of a portable terminal, but also greatly improves the prediction robustness and generalization ability, and achieves accurate prediction of the fatigue life of UAV landing gear.

[0041] See Figure 5 As shown, this application discloses a device for predicting the fatigue life of a UAV landing gear, comprising: The feature set acquisition module 11 is used to acquire the raw flight parameter data collected by the target flight parameter recorder carried by the target UAV, perform feature dimensionality reduction processing on the raw flight parameter data to obtain target flight parameter data, and determine the input feature set corresponding to each predicted target based on the target flight parameter data; the predicted target includes the landing gear lateral force at the landing gear touchdown point, the landing gear heading force at the landing gear touchdown point, the wheel center heading load and the vertical load of the target UAV. The load data acquisition module 12 is used to determine the target load time series data corresponding to the target flight parameter data based on the input feature set using the target load prediction model; the target load prediction model is a load prediction model based on an ensemble learning framework architecture of stacked ensemble networks. The fatigue life prediction module 13 is used to apply the target load time series data to the target three-dimensional model, perform stress analysis on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage of a single sortie corresponding to the target flight parameter data, and determine the fatigue life of the target landing gear based on the total damage of a single sortie; the target three-dimensional model is a full-size three-dimensional model of the target landing gear.

[0042] As can be seen, this application cleverly utilizes the inherent conventional airborne flight parameters of the UAV flight parameter recorder and constructs a virtual sensor through software and hardware algorithms. Under the condition of not damaging the original physical structure of the aircraft and not increasing any effective payload, it achieves high-precision load spectrum perception. By performing feature dimensionality reduction on the original flight parameter data, the target load prediction model based on the ensemble learning architecture of stacked ensemble network is used to process the target flight parameter data to identify the influence of transient spatial pose on the load. It has extremely strong regularization ability and significantly reduces computational complexity. It not only reduces the inference computing power requirement from the GPU level to the ordinary CPU level of a portable terminal, but also greatly improves the prediction robustness and generalization ability, and achieves accurate prediction of the fatigue life of UAV landing gear.

[0043] In one specific embodiment, the device may further include: The file verification module is used to receive the target running working directory through interface event driving, and to verify whether a model file with the target suffix already exists in the target running working directory; The file retrieval module is used to search the target underlying resource library for the target source model file corresponding to the target suffix if the target running working directory does not contain a model file with the target suffix. The file copy module is used to save the target source model file to the target running working directory using a target copy command if the target source model file exists in the target underlying resource library.

[0044] In one specific embodiment, the feature set acquisition module 11 may include: The first dimensionality reduction unit is used to determine the degree of nonlinear dependence between each flight parameter variable and the target load variable in the original flight parameter data, and to perform feature dimensionality reduction on the original flight parameter data based on the degree of nonlinear dependence to obtain the first flight parameter data. The second dimensionality reduction unit is used to determine the target variance corresponding to each of the flight parameter variables in the first flight parameter data, and to perform feature dimensionality reduction on the first flight parameter data based on a preset variance threshold and the target variance to obtain the second flight parameter data. The third dimensionality reduction unit is used to filter the flight parameter variables in the second flight parameter data using the target correlation matrix to obtain the target flight parameter data.

[0045] In one specific embodiment, the load data acquisition module 12 may include: The load prediction unit is used to input the input feature set into the target load prediction model, so as to generate the load prediction sequence of each intermediate layer corresponding to the input feature set by using the target base learners of the first network layer of the target load prediction model. The sequence splicing unit is used to perform matrix splicing on each of the intermediate layer load prediction sequences using the second network layer of the target load prediction model to obtain the target load time series data corresponding to the target flight parameter data.

[0046] In one specific embodiment, the device may further include: The matrix construction module is used to construct a target three-dimensional model corresponding to the target landing gear using target finite element software, and apply unit loads in four directions to the target stress points of the target three-dimensional model to determine the stress tensor component response coefficients corresponding to the target stress points, and construct the unit load stress response coefficient matrix corresponding to the target landing gear based on the stress tensor component response coefficients.

[0047] In one specific embodiment, the fatigue life prediction module 13 may include: The real stress component determination unit is used to convert the target load time series data into the real stress components of the target stress point corresponding to the target landing gear of the target UAV at the target time using the target three-dimensional model and based on the unit load stress response coefficient matrix. The fatigue load cycle determination unit is used to convert the real stress components into a target stress time series based on the signed fourth strength theory, and to determine the fatigue load cycle corresponding to the target flight parameter data based on the target stress time series using a preset rainflow counting algorithm. The single-flight total damage determination unit is used to correct the fatigue load cycle using the target boundary criterion equation, so as to determine the single-flight total damage corresponding to the target flight parameter data based on the corrected fatigue load cycle and Miner's linear cumulative damage theory; the target boundary criterion equation is the fatigue failure boundary criterion equation determined based on the Goodman curve. The fatigue life determination unit is used to determine the global cumulative damage corresponding to the target landing gear based on the historical cumulative damage data corresponding to the target landing gear and the total damage of a single sortie, and to determine the fatigue life corresponding to the target landing gear based on the current equivalent damage evolution rate of the target UAV and the global cumulative damage.

[0048] Furthermore, embodiments of this application also disclose an electronic device, Figure 6 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0049] Figure 6 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the UAV landing gear fatigue life prediction method disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0050] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0051] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk, or optical disk, etc. The resources stored thereon can include an operating system 221, computer programs 222, etc., and the storage method can be temporary storage or permanent storage.

[0052] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the UAV landing gear fatigue life prediction method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0053] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned method for predicting the fatigue life of UAV landing gear. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0054] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0055] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0056] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0057] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0058] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predicting the fatigue life of a UAV landing gear, characterized in that, include: The raw flight parameter data collected by the target flight parameter recorder on the target UAV is acquired, and the raw flight parameter data is subjected to feature dimensionality reduction processing to obtain target flight parameter data. Based on the target flight parameter data, the input feature set corresponding to each predicted target is determined. The predicted target includes the landing gear lateral force at the landing gear touchdown point, the landing gear yaw force at the landing gear touchdown point, the wheel center yaw load, and the vertical load of the target UAV. The target payload time series data corresponding to the target flight parameter data is determined using the target payload prediction model based on the input feature set; The target load prediction model is a load prediction model based on an ensemble learning framework architecture with stacked ensemble networks. The target load time series data is applied to the target three-dimensional model, and stress analysis is performed on the target stress point corresponding to the target landing gear of the target UAV to determine the total damage of a single sortie corresponding to the target flight parameter data, and the fatigue life of the target landing gear is determined based on the total damage of a single sortie. The target 3D model is a full-size 3D model of the target landing gear.

2. The method for predicting the fatigue life of UAV landing gear according to claim 1, characterized in that, Before acquiring the raw flight parameter data collected by the target flight parameter recorder mounted on the target UAV, the process also includes: The target running working directory is received through the interface event-driven mechanism, and it is verified whether a model file with the target suffix already exists in the target running working directory; If the target working directory does not contain a model file with the target suffix, then the target resource path parsing function is used to search the target underlying resource library to see if there is a target source model file corresponding to the target suffix. If the target source model file exists in the target underlying resource library, the target source model file is saved to the target running working directory using the target copy command.

3. The method for predicting the fatigue life of UAV landing gear according to claim 1, characterized in that, The step of performing feature dimensionality reduction processing on the original flight parameter data to obtain the target flight parameter data includes: Determine the degree of nonlinear dependence between each flight parameter variable and the target payload variable in the original flight parameter data, and perform feature dimensionality reduction on the original flight parameter data based on the degree of nonlinear dependence to obtain the first flight parameter data; Determine the target variance corresponding to each flight parameter variable in the first flight parameter data, and perform feature dimensionality reduction on the first flight parameter data based on the preset variance threshold and the target variance to obtain the second flight parameter data; The target flight parameter data is obtained by filtering the flight parameter variables in the second flight parameter data using the target correlation matrix.

4. The method for predicting the fatigue life of UAV landing gear according to claim 1, characterized in that, The step of using the target payload prediction model to determine the target payload time series data corresponding to the target flight parameter data based on the input feature set includes: The input feature set is input into the target load prediction model, so as to generate the load prediction sequence of each intermediate layer corresponding to the input feature set by using the target base learners of the first network layer of the target load prediction model. Using the second network layer of the target load prediction model, the load prediction sequences of each intermediate layer are matrix-concatenated to obtain the target load time series data corresponding to the target flight parameter data.

5. The method for predicting the fatigue life of UAV landing gear according to claim 4, characterized in that, The target base learners include a first base learner based on extreme gradient boosting trees, a second base learner based on random forests, a third base learner based on extremely random trees, a fourth base learner based on bagging model, and a fifth base learner based on categorical gradient boosting algorithm.

6. The method for predicting the fatigue life of a UAV landing gear according to any one of claims 1 to 5, characterized in that, Before applying the target load time series data to the target 3D model and performing stress analysis on the target landing gear corresponding to the target UAV to determine the total damage per sortie corresponding to the target flight parameter data, the method further includes: A target three-dimensional model corresponding to the target landing gear is constructed using target finite element software, and unit loads in four directions are applied to the target stress points of the target three-dimensional model to determine the stress tensor component response coefficients corresponding to the target stress points. Based on the stress tensor component response coefficients, a unit load stress response coefficient matrix corresponding to the target landing gear is constructed.

7. The method for predicting the fatigue life of UAV landing gear according to claim 6, characterized in that, The step of applying the target load time-series data to the target 3D model, performing stress analysis on the target stress points corresponding to the target landing gear of the target UAV to determine the total damage per sortie corresponding to the target flight parameter data, and determining the fatigue life corresponding to the target landing gear based on the total damage per sortie, includes: Using the target 3D model, the target load time series data is converted into the actual stress components of the target stress point corresponding to the target landing gear of the target UAV at the target time based on the unit load stress response coefficient matrix. Based on the signed fourth strength theory, the real stress components are converted into a target stress time series, and the fatigue load cycle corresponding to the target flight parameter data is determined based on the target stress time series using a preset rainflow counting algorithm. The fatigue load cycle is corrected using the target boundary criterion equation, and the total damage per sortie corresponding to the target flight parameter data is determined based on the corrected fatigue load cycle and Miner's linear cumulative damage theory; the target boundary criterion equation is the fatigue failure boundary criterion equation determined based on the Goodman curve. The global cumulative damage corresponding to the target landing gear is determined based on the historical cumulative damage data corresponding to the target landing gear and the total damage of a single sortie. The fatigue life corresponding to the target landing gear is determined based on the current equivalent damage evolution rate of the target UAV and the global cumulative damage.

8. A device for predicting the fatigue life of a UAV landing gear, characterized in that, include: The feature set acquisition module is used to acquire the raw flight parameter data collected by the target flight parameter recorder carried by the target UAV, perform feature dimensionality reduction processing on the raw flight parameter data to obtain target flight parameter data, and determine the input feature set corresponding to each predicted target based on the target flight parameter data; the predicted targets include the landing gear lateral force at the landing gear touchdown point, the landing gear yaw force at the landing gear touchdown point, the wheel center yaw load, and the vertical load of the target UAV. The load data acquisition module is used to determine the target load time series data corresponding to the target flight parameter data based on the input feature set using the target load prediction model; the target load prediction model is a load prediction model based on an ensemble learning framework architecture of stacked ensemble networks. The fatigue life prediction module is used to apply the target load time series data to the target three-dimensional model, perform stress analysis on the target stress point corresponding to the target landing gear of the target UAV to determine the total damage of a single sortie corresponding to the target flight parameter data, and determine the fatigue life of the target landing gear based on the total damage of a single sortie. The target 3D model is a full-size 3D model of the target landing gear.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the method for predicting the fatigue life of unmanned aerial vehicle landing gear as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the method for predicting the fatigue life of unmanned aerial vehicle landing gear as described in any one of claims 1 to 7.

Images