Method and system for landing gear actuator fatigue estimation

Abstract

The application provides a kind of flight landing gear actuator fatigue estimation method and system, comprising: obtaining the service history sample set of flight landing gear actuator;The service history sample set is pretreated, and the service state is marked;Respectively using GRNN, RBF and DNN are trained, obtain multiple actuator fatigue estimation models, and the model output is actuator failure flag;The prediction accuracy value of each model in the preset service time cycle is counted;The prediction accuracy value of each model in the preset service time cycle is weighted and summed to determine the weighted prediction accuracy;The model with the highest weighted prediction accuracy is obtained as the final fatigue estimation prediction model;The feature vector of the aircraft is used as input and input into the fatigue estimation prediction model, and the predicted actuator failure flag is output.The application can be used for the failure warning of the actuator of the aircraft landing gear in the service state, to reduce the defects of artificial judgment and subjective judgment.

Description

Technical Field

[0001] This invention relates to the field of flight simulator technology, and in particular to the simulation of flight landing gear actuators, specifically to a method and system for fatigue estimation of flight landing gear actuators. Background Technology

[0002] Actuators are the main actuators in hydraulic transmission systems, used in many important locations on aircraft such as elevators, rudders, doors, and landing gear. They play a crucial role in servo control, steering, and power control systems, and their performance and safety are directly related to the aircraft's performance. The quality of actuator performance and its safe operation directly affect the aircraft's performance and safe flight.

[0003] The landing gear of an aircraft is crucial for safe takeoff and landing. Whether using a tricycle, tailwheel, or multi-row design, the landing gear bears the majority of the aircraft's weight. Its high-speed takeoff and landing, as well as its smooth deployment and retraction, directly impact flight safety. In both civil and military aircraft, there have been numerous crashes / emergency landings caused by landing gear malfunctions. At the moment of impact, the tens or hundreds of tons of pressure borne by the wheels are directly applied to the landing gear, relying on the main shock absorbers and the retraction / retraction actuators. If the hydraulic struts have been in service for a long time and have developed hidden risks or are damaged, it can ultimately cause the fuselage or wings to strike the ground, resulting in a flight safety accident.

[0004] Actuators used in the landing gear of aircraft, also known as actuator cylinders, are often tested for their performance using load testing machines during the design phase. For example, Chinese patent CN203350054U discloses a method that uses springs to constrain the flight control mechanical control system behind the actuator, thereby achieving accurate loading of the test piece during movement and simulating the movement process of the flight control mechanical control system behind the actuator in flight. Especially in the performance testing of actuators using electromechanical systems, stiffness testing is achieved through motors and lead screw mechanisms. For example, the electromechanical actuator stiffness testing device and method disclosed in patent CN107727341 uses a symmetrical layout of lead screws II and I on both sides of the test piece, and employs a symmetrical manual loading method to avoid damage to the electromechanical actuator itself due to excessive active loading control force. High displacement measurement accuracy laser sensors I and II are used, and they are placed on a parallel line to improve measurement accuracy. The lead screws II and I with self-locking T-shaped large-pitch threads, in conjunction with lead screw connecting gears I and II, a force-applying gear, and a force-applying gear lever arm, are used to apply the loading force. The loading force test is made more accurate by directly connecting the tension and compression sensors in series with the electromechanical actuator itself.

[0005] Existing performance testing or experimental equipment is primarily designed for actuator design, aiming to test the performance of the installed actuators. However, after actual service, such as in the actuator cylinders used for aircraft landing gear lifting, their strength and stiffness will change due to the combined effects of various factors such as aircraft load, ground friction, acceleration, and service time. This can even lead to sudden and severe changes. The results of performance tests such as stiffness testing, durability testing, and instability load testing during the design phase will no longer be applicable at this time, and it will be difficult to actually disassemble the aircraft for testing and early warning. Summary of the Invention

[0006] A first aspect of the present invention provides a method for estimating fatigue in a flight landing gear actuator, comprising:

[0007] Obtain a service history sample set of the flight landing gear actuators, including actuator service data and landing gear main shock absorber service data; the actuator service data includes external force load, ground friction, ambient temperature, hydraulic medium temperature, instantaneous landing velocity, acceleration, piston displacement, design stiffness, service time, and service status; the landing gear main shock absorber service data includes main shock absorber design load, coefficient of restitution, service time, and service status.

[0008] The service data of the actuators and the main shock absorbers of the landing gear are preprocessed and the service status is marked.

[0009] Using the preprocessed service history sample set as input samples, and external force load, ground friction force, ambient temperature, hydraulic medium temperature, instantaneous landing speed, acceleration, piston displacement, design stiffness, main shock absorber design load, recovery coefficient, service time, and service status as feature vectors, the model is trained using generalized regression neural network (GRNN), radial basis function neural network (RBF), and deep learning neural network (DNN) respectively to obtain multiple actuator fatigue estimation models. The model output is the actuator fault indicator.

[0010] Statistical analysis of the prediction accuracy of each model within a preset service life period;

[0011] The weighted prediction accuracy is determined by summing the prediction accuracy values ​​of each model within a preset service time period.

[0012] The model with the highest weighted prediction accuracy is used as the final fatigue estimation prediction model;

[0013] The aircraft's feature vector is obtained as input and fed into the fatigue estimation and prediction model, which outputs the predicted actuator fault flag.

[0014] A second aspect of the present invention provides a fatigue estimation system for a flight landing gear actuator, comprising:

[0015] One or more processors;

[0016] The memory stores operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the aforementioned flight landing gear actuator fatigue estimation method.

[0017] A third aspect of the invention provides a computer-readable medium for storing software, the software including instructions executable by one or more computers, the instructions causing the one or more computers to perform operations including the flow of the aforementioned flight landing gear actuator fatigue estimation method.

[0018] The fatigue estimation method and system for flight landing gear actuators proposed in this invention aim to select the optimal model for different service life conditions by using service data of actuators and service data of landing gear main shock absorbers through multi-model fusion training. Due to the heterogeneous diversity and nonlinear variations of actuator service data and landing gear main shock absorber service data, this invention aims to achieve optimized regression under multiple samples and multiple feature parameters by leveraging the nonlinear mapping capability and learning speed of the GRNN network. It also boasts fast convergence speed and, based on radial basis function networks, exhibits excellent nonlinear approximation performance. Furthermore, the RBF neural network itself is a typical radial propagation neural network, using RBF (radial basis functions) as the "basis" of hidden units to construct the hidden layer space. The hidden layers transform the input vector, converting low-dimensional pattern input data into a high-dimensional space, making linearly inseparable problems in the low-dimensional space linearly separable in the high-dimensional space. This shortens training time and significantly controls training progress when using multi-dimensional actuator and main shock absorber data as feature inputs. In the DNN network structure, the chain-like forward propagation of the DNN deep neural network, with full connections between multiple layers, enables training and prediction output based on massive features.

[0019] Based on this, and considering the different actuator types and design performance of various aircraft, the impact of main shock absorbers and service life is further explored. Through a comprehensive judgment of weighted prediction accuracy, the optimal prediction model is selected to analyze and predict the fatigue state, i.e., the failure characteristics, of the actuators of the same type of aircraft after service, so as to achieve early warning and avoid the shortcomings of subjective judgment based solely on cabin ground inspection and visual observation.

[0020] It should be understood that all combinations of the foregoing concepts and the additional concepts described in more detail below may be considered part of the inventive subject matter of this disclosure, provided that such concepts do not contradict each other. Furthermore, all combinations of the claimed subject matter are considered part of the inventive subject matter of this disclosure.

[0021] The foregoing and other aspects, embodiments, and features of the teachings of the present invention will be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the invention, such as features and / or beneficial effects of exemplary embodiments, will become apparent from the following description or may be learned through practice of specific embodiments according to the teachings of the present invention. Attached Figure Description

[0022] The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in the various figures may be denoted by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the invention will now be described by way of example and with reference to the accompanying drawings, wherein:

[0023] Figure 1 This is a schematic flowchart of a fatigue estimation method for flight landing gear actuators according to an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of a flight landing gear actuator fatigue estimation device according to an embodiment of the present invention. Detailed Implementation

[0025] To better understand the technical content of the present invention, specific embodiments are described below in conjunction with the accompanying drawings.

[0026] Various aspects of the invention are described in this disclosure with reference to the accompanying drawings, which illustrate numerous illustrative embodiments. The embodiments of this disclosure are not necessarily intended to encompass all aspects of the invention. It should be understood that the various concepts and embodiments described above, as well as those described in more detail below, can be implemented in any of many ways, because the concepts and embodiments disclosed herein are not limited to any particular implementation. Furthermore, some aspects of the invention disclosed may be used alone or in any suitable combination with other aspects of the invention disclosed.

[0027] Combination Figure 1 The fatigue estimation method for flight landing gear actuators in the exemplary embodiment shown adopts a comprehensive judgment based on multi-model fusion training and weighted prediction accuracy to select the optimal prediction model, which is used for fault warning of aircraft landing gear actuators in service, reducing the defects of manual judgment and subjective judgment.

[0028] Combination Figure 1The flowchart (S100) of the example flight landing gear actuator fatigue estimation method shown includes the following steps:

[0029] S101: Obtain a service history sample set of the flight landing gear actuators, including actuator service data and landing gear main shock absorber service data; the actuator service data includes external load, ground friction, ambient temperature, hydraulic medium temperature, instantaneous landing speed, acceleration, piston displacement, design stiffness, service time, and service status; the landing gear main shock absorber service data includes main shock absorber design load, coefficient of restitution, service time, and service status;

[0030] S102: Preprocess the service data of the actuators and the main shock absorbers of the landing gear, and mark the service status; among them, the fault flag of normal service status is 0, and the fault flag of abnormal service status is 1.

[0031] S103: Using the pre-processed service history sample set as input samples, and external force load, ground friction force, ambient temperature, hydraulic medium temperature, instantaneous landing speed, acceleration, piston displacement, design stiffness, main shock absorber design load, recovery coefficient, service time, and service status as feature vectors, the model is trained using generalized regression neural network (GRNN), radial basis function neural network (RBF), and deep learning neural network (DNN) to obtain multiple actuator fatigue estimation models. The model output is the actuator fault indicator.

[0032] S104: Statistically calculate the prediction accuracy of each model within the preset service time period;

[0033] S105: The prediction accuracy values ​​of each model within the preset service time period are weighted and summed to determine the weighted prediction accuracy;

[0034] S106: The model with the highest weighted prediction accuracy is used as the final fatigue estimation prediction model;

[0035] S107: Obtain the aircraft's feature vector as input, input it into the fatigue estimation and prediction model, and output the predicted actuator fault flag.

[0036] In embodiments of the present invention, input samples are divided into a training set and a test set, wherein the training set is used to train the model, and the test set is used to test the accuracy of the model's predictions. The aforementioned prediction accuracy value is obtained based on the model's predictions using the test set.

[0037] For a specific type of aircraft, the aforementioned service history sample set is obtained from the historical flight data of that aircraft type; and the input sample includes at least 5,000 sample data points, of which at least 80% are used to train the model. In an embodiment of the present invention, the training set consists of 4,500 sample data points, and the test set consists of 500 sample data points.

[0038] During the training process in step S103, taking the GRNN network as an example, the training process can adopt existing training procedures. Optional examples of training procedures include:

[0039] First, define the size of the input layer, hidden layer, summation layer, and output layer, construct the GRNN network model, and use cross-validation to determine the optimal smoothing parameters and train the GRNN network.

[0040] Then, based on the trained GRNN network, the test set data is used as input to predict the output. The module's suitability is determined by comparing the output with the actual results. If the prediction accuracy is unsatisfactory, the GRNN model is retrained or optimized until a GRNN neural network prediction model with prediction error meeting the expected requirements is obtained.

[0041] The GRNN network follows the formula below.

[0042]

[0043] In the formula, X = [x1, x2, x3, ..., x... m ] T X is the network input, i.e., the feature vector, y is the corresponding prediction output, representing the prediction output of y given the input X, and f(X, y) is the joint probability density function of X and y.

[0044] In an optional embodiment, the training process based on DNN and RBF networks can be implemented in conjunction with existing commercial training processes.

[0045] Therefore, based on the same training set, models with three different network structures are trained to obtain the corresponding prediction models.

[0046] It should be understood that by using generalized regression neural networks (GRNN), radial basis function neural networks (RBF), and deep learning neural networks (DNN) for training, the prediction model is obtained by optimizing the prediction results through the judgment of prediction errors during the model training process.

[0047] Furthermore, based on the three models obtained through training, each model can be input into a test set and output a corresponding predicted output, i.e., an actuator fault flag, which is 1 or 0. This output is then compared with the actual values ​​in the test set to determine the prediction accuracy of each model.

[0048] In an embodiment of the present invention, step S104 involves calculating the prediction accuracy value of each model within a preset service time period, specifically including the following process:

[0049] Using the service year as a period, the prediction accuracy of each actuator fatigue estimation model in the corresponding service year is statistically analyzed, and expressed as: Prediction accuracy of the service year = Number of fault signs accurately predicted in the corresponding service year / Number of samples in the corresponding service year;

[0050] The service year cycle is set to 1 year, such as 1 year, 2 years, 3 years, 4 years, 5 years, etc.

[0051] In an embodiment of the present invention, step S105 involves weighted summation of the prediction accuracy values ​​of each model within a preset service time period to determine the weighted prediction accuracy. This specifically includes the following process:

[0052] The weighted prediction accuracy of each model is calculated as follows:

[0053] f(1)=w1*q11+w2*q12+w3*q13+...+wn*q1n

[0054] f(2)=w1*q21+w2*q22+w3*q23+...+wn*q2n

[0055] f(3)=w1*q31+w2*q32+w3*q33+...+wn*q3n

[0056] Here, the actuator fatigue estimation models trained based on generalized regression neural network (GRNN), radial basis function neural network (RBF) and deep learning neural network (DNN) are defined as the first estimation model, the second estimation model and the third estimation model, respectively; f(1), f(2) and f(3) represent the weighted prediction accuracy of the first estimation model, the second estimation model and the third estimation model, respectively.

[0057] q1n represents the prediction accuracy of the first estimation model in the nth year of service;

[0058] q2n represents the prediction accuracy of the second estimation model in the nth year of service;

[0059] q3n represents the prediction accuracy of the third estimation model in the nth year of service;

[0060] w1, w2, w3, ..., wn represent the weighted values ​​of the prediction accuracy for different service years.

[0061] The weighted values ​​for the prediction accuracy values ​​of different service years are set as follows:

[0062] w1+w2+w3+…+wn=1;

[0063] Furthermore, w1>w2>w3>…>wn.

[0064] The shorter the service life, the closer the performance of the actuator and main shock absorber is to the design standard, and the higher the required prediction accuracy. Therefore, when screening the prediction accuracy of different estimation models, the required prediction accuracy is higher and the weight is higher for actuators with shorter service life during the weighted calculation process.

[0065] For example, in an embodiment of the present invention, the data definitions for the training set and the test set are as follows:

[0066] training set

[0067] m1(x11,x12,x13,x14,x15,x16,x17,x18,x19,x110,x111,x112,mark1)

[0068] m2(x21,x22,x23,x24,x25,x26,x27,x28,x29,x210,x211,x212,mark2)

[0069] m3(x31,x32,x33,x34,x35,x36,x37,x38,x39,x310,x311,x312,mark3) . . .

[0073] mn(xn1,xn2,xn3,xn4,xn5,xn6,xn7,xn8,xn9,xn10,xn11,xn12,markn)

[0074] Where xn1,xn2,xn3,xn4,xn5,xn6,xn7,xn8,xn9,xn10,xn11,xn12 represent the feature vectors of the nth sample data, respectively representing external force load, ground friction force, ambient temperature, hydraulic medium temperature, instantaneous landing velocity, acceleration, piston displacement, design stiffness, main shock absorber design load, recovery coefficient, service time, and service status.

[0075] Markn represents the fault flag for the nth sample data.

[0076] Three prediction models generated through training are input into the test set, and the models predict the corresponding fault indicators. Then, the prediction accuracy of each model within a preset service time period is obtained by calculating the prediction accuracy of the service year as described above. An example is shown below:

[0077] First estimation model accuracy Second estimation model accuracy Third estimation model accuracy 1 year of service q11 q21 q31 2 years of service q12 q22 q32 3 years of service q13 q23 q33 ... ... ... ... n years of service q1n q2n q3n

[0078] Combination Figure 2 As shown in the embodiment of the present invention, a fatigue estimation device 200 for a flight landing gear actuator is also proposed, including a service history sample set acquisition module, a service history sample set preprocessing module, a training module, a prediction accuracy value statistics module, a weighted prediction accuracy acquisition module, a fatigue estimation prediction model screening module, and a prediction output module.

[0079] The service history sample set acquisition module is used to acquire the service history sample set of the flight landing gear actuators, including actuator service data and landing gear main shock absorber service data. The actuator service data includes external load, ground friction, ambient temperature, hydraulic medium temperature, instantaneous landing speed, acceleration, piston displacement, design stiffness, service time, and service status; the landing gear main shock absorber service data includes main shock absorber design load, coefficient of restitution, service time, and service status.

[0080] The service history sample set preprocessing module is used to preprocess the service data of actuators and the service data of landing gear main shock absorbers, and to mark the service status; among them, the fault flag of normal service status is 0, and the fault flag of abnormal service status is 1.

[0081] The training module uses a preprocessed service history sample set as input samples and external force load, ground friction, ambient temperature, hydraulic medium temperature, instantaneous landing velocity, acceleration, piston displacement, design stiffness, main shock absorber design load, recovery coefficient, service time, and service status as feature vectors. It is trained using a generalized regression neural network (GRNN), a radial basis function neural network (RBF), and a deep learning neural network (DNN) to obtain multiple actuator fatigue estimation models. The output of each model is an actuator fault indicator.

[0082] The prediction accuracy statistics module is used to calculate the prediction accuracy of each model within a preset service time period.

[0083] The weighted prediction accuracy acquisition module is used to sum the prediction accuracy values ​​of each model within a preset service time period to determine the weighted prediction accuracy.

[0084] The fatigue estimation prediction model screening module is used to select the model with the highest weighted prediction accuracy as the final fatigue estimation prediction model.

[0085] The prediction output module is used to obtain the aircraft's feature vector as input, input it into the fatigue estimation prediction model, and output the predicted actuator fault flag.

[0086] According to an embodiment of the present invention, a flight landing gear actuator fatigue estimation system is also proposed, comprising: one or more processors; and a memory storing operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the flight landing gear actuator fatigue estimation method of the foregoing embodiments.

[0087] According to embodiments of the present invention, a computer-readable medium for storing software including instructions executable by one or more computers, which, when executed, cause the one or more computers to perform operations including the flow of the flight landing gear actuator fatigue estimation method of the foregoing embodiments.

[0088] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Those skilled in the art can make various modifications and refinements without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. A method for estimating fatigue in a flight landing gear actuator, characterized in that, include: Obtain a service history sample set of the flight landing gear actuators, including actuator service data and landing gear main shock absorber service data; the actuator service data includes external force load, ground friction force, ambient temperature, hydraulic medium temperature, instantaneous landing speed, acceleration, piston displacement, design stiffness, service time and service status. The service data of the main landing gear shock absorber includes the design load, recovery coefficient, service time, and service status of the main shock absorber. The service data of the actuators and the main shock absorbers of the landing gear are preprocessed and the service status is marked. Using the preprocessed service history sample set as input samples, and external force load, ground friction force, ambient temperature, hydraulic medium temperature, instantaneous landing speed, acceleration, piston displacement, design stiffness, main shock absorber design load, recovery coefficient, service time, and service status as feature vectors, the model is trained using a generalized regression neural network (GRNN), a radial basis function neural network (RBF), and a deep learning neural network (DNN) to obtain multiple actuator fatigue estimation models. The model output is the actuator fault indicator. The prediction accuracy of each model within a preset service life period is calculated, where the preset service life period is set to 1 year. The weighted prediction accuracy is determined by summing the prediction accuracy values ​​of each model within a preset service time period. The model with the highest weighted prediction accuracy is used as the final fatigue estimation prediction model; The aircraft's feature vector is obtained as input and fed into the fatigue estimation and prediction model, which outputs the predicted actuator fault flag. The step of weighted summing of the prediction accuracy values ​​of each model within a preset service time period to determine the weighted prediction accuracy includes: The weighted prediction accuracy of each model is calculated as follows: f(1) = w1*q11+ w2*q12+ w3*q13+…+ wn*q1n f(2) = w1*q21+ w2*q22+ w3*q23+…+ wn*q2n f(3) = w1*q31+ w2*q32+ w3*q33+…+ wn*q3n Here, the actuator fatigue estimation models trained based on generalized regression neural network (GRNN), radial basis function neural network (RBF) and deep learning neural network (DNN) are defined as the first estimation model, the second estimation model and the third estimation model, respectively; f(1), f(2) and f(3) represent the weighted prediction accuracy of the first estimation model, the second estimation model and the third estimation model, respectively. q1n represents the prediction accuracy of the first estimation model in the nth year of service; q2n represents the prediction accuracy of the second estimation model in the nth year of service; q3n represents the prediction accuracy of the third estimation model in the nth year of service; w1, w2, w3, ..., wn represent the weighted values ​​of the prediction accuracy for different service years.

2. The fatigue estimation method for flight landing gear actuators according to claim 1, characterized in that, The service status is marked, where a fault flag of normal service status is 0, and a fault flag of abnormal service status is 1.

3. The fatigue estimation method for flight landing gear actuators according to claim 1, characterized in that, The method employs a generalized regression neural network (GRNN), a radial basis function neural network (RBF), and a deep learning neural network (DNN) for training. During model training, the prediction error of the prediction results is evaluated and optimized to obtain the corresponding prediction model.

4. The fatigue estimation method for flight landing gear actuators according to claim 1, characterized in that, The statistical analysis of the prediction accuracy of each model within a preset service life period includes: Using the service year as a period, the prediction accuracy of each actuator fatigue estimation model in the corresponding service year is statistically analyzed, and expressed as: Prediction accuracy of the service year = Number of fault signs accurately predicted in the corresponding service year / Number of samples in the corresponding service year.

5. The fatigue estimation method for flight landing gear actuators according to claim 1, characterized in that, The weighted values ​​of the prediction accuracy values ​​for different service years are set as follows: w1+w2+w3+…+wn=1; Furthermore, w1 > w2 > w3 > … > wn.

6. The fatigue estimation method for flight landing gear actuators according to any one of claims 1-5, characterized in that, The input samples are divided into a training set and a test set, wherein the training set is used to train the model and the test set is used to test the accuracy of the model's predictions. The aforementioned prediction accuracy value is obtained by model prediction based on the test set.

7. The fatigue estimation method for flight landing gear actuators according to claim 6, characterized in that, For a specific type of aircraft, the service history sample set is obtained from the historical flight data of that type of aircraft; and the input sample includes at least 5,000 sample data, of which at least 80% is used to train the model.

8. A fatigue estimation system for flight landing gear actuators, characterized in that, include: One or more processors; The memory stores operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the flight landing gear actuator fatigue estimation method as described in any one of claims 1-7.

9. A computer-readable medium for storing software, characterized in that, The software includes instructions executable by one or more computers, which, when executed, cause the one or more computers to perform operations including the flow of the flight landing gear actuator fatigue estimation method as described in any one of claims 1-7.

Images