A method for identifying aircraft flight courses based on time-series flight parameter data

Abstract

This invention belongs to the field of aircraft structural fatigue life design, specifically involving a method for identifying aircraft flight courses based on time-series flight parameter data. The invention extracts time, external load, total aircraft weight, center-of-gravity normal overload, and center-of-gravity lateral overload history from multivariate time-series flight parameter data, performing error correction, filtering, and compression. Then, based on correlation analysis between physical parameters and flight parameters, it mines flight parameter data strongly correlated with center-of-gravity overload. Features are extracted from the above flight parameter data to construct a key feature database for flight course identification. Finally, machine learning models such as support vector machines and random forests are used to construct a flight course prediction model, realizing aircraft flight course identification technology based on time-series flight parameter data. The identification method proposed in this invention fills the gap in intelligent flight course discrimination, saves labor costs, and has high identification efficiency and accuracy, effectively avoiding the problems of mislabeling and omission of flight courses in manual marking.

Description

Technical Field

[0001] This invention belongs to the field of aircraft structural fatigue life design, specifically relating to a method for identifying aircraft flight courses based on time-series flight parameter data. Background Technology

[0002] During the aircraft design phase, a design load spectrum reflecting the expected usage is compiled based on the design mission profile. Based on this, durability and damage tolerance analyses and tests are conducted. Since the design load spectrum cannot accurately reflect the aircraft's actual service life, a measured load spectrum needs to be compiled to correct the analysis and test results from the design phase, ensuring the flight safety of the aircraft in service.

[0003] GJB67.6A, "Repetitive Loads, Durability and Damage Tolerance", requires that when full-size damage tolerance tests are conducted after strength flight tests or batch production, the test load spectrum should, as far as possible, take into account the correction of the design measured load spectrum by the measured load data.

[0004] The measured load spectrum needs to represent the overall usage of the aircraft fleet. Compiling the measured load spectrum requires considering the proportion of each flight type's takeoffs and landings within the total number of takeoffs and landings during fleet operation, combined with typical takeoff and landing load data for each flight type. Therefore, obtaining the proportion of each flight type's takeoffs and landings within the total number of takeoffs and landings is crucial for compiling the measured load spectrum.

[0005] Currently, the main method for obtaining the flight course ratio for an aircraft fleet relies on ground crew manually marking flight parameters after takeoff and landing. This process is required after each flight for every aircraft, which not only increases the workload significantly but also increases the risk of mismarking and omissions. To address this issue, it is necessary to develop a flight course identification technology based on flight parameters to achieve automatic identification of flight courses. This would enable the compilation of more accurate measured load spectra and provide support for more accurate lifespan determination of aircraft in the field. Summary of the Invention

[0006] The purpose of this invention is to propose a flight course identification method. By extracting feature values ​​from time-series flight parameter data, a model is constructed to predict flight courses from these features, thus realizing a flight course discrimination method based on flight parameter feature data. This invention fills the gap in intelligent flight course discrimination, supplementing and improving flight missions without recorded courses, and ensuring the integrity of the data used. It is highly efficient and accurate, avoiding problems such as mislabeling and omissions that easily occur when manually marking flight courses. This patent can support the compilation of aircraft measured load spectra, providing support for field aircraft fatigue life determination and ensuring aircraft operational safety.

[0007] Technical solution of the present invention

[0008] A method for identifying aircraft flight maneuvers based on time-series flight parameter data, the method comprising the following steps:

[0009] Step 1: Flight Parameter Data Screening and Cleaning: Initially, time, external load weight, total aircraft weight, center of gravity normal overload, and center of gravity lateral overload are retained by default. Correlation analysis is conducted between the physical basis and flight parameters. For the remaining flight parameters, multicollinearity analysis and stepwise regression are used to screen out flight parameters with high correlation to center of gravity normal overload and center of gravity lateral overload. The final 14 selected flight parameters include: time, external load weight, total aircraft weight, center of gravity normal overload, center of gravity lateral overload, altitude, Mach number, rate of climb, roll rate, pitch rate, yaw rate, bank angle, pitch angle, and heading angle. Error data, out-of-bounds data, and ground segment data are then removed to reduce the overall data size.

[0010] Step 2: Data Processing: After processing the flight parameter data in Step 1, except for the time term, new parameter terms are generated based on the physical meaning, mathematical and / or mechanical models of the flight parameters, such as nonlinear terms like flight altitude × Mach number, Mach number × rudder position signal, rudder position signal × left outer elevator position signal, etc.; then, multicollinearity diagnosis and filtering methods are used to reduce the size of the sample space (i.e., dimensionality reduction) and the data is standardized.

[0011] Step 3: Time-series data feature extraction: For the time-series flight parameter data processed in Step 2, feature extraction is performed to construct a flight parameter feature database; then, combined with the information of each takeoff and landing flight subject, a typical flight parameter feature & flight subject database is constructed.

[0012] Step 4: Using typical flight parameter features and flight subject database as input, construct a flight subject prediction model based on time-series flight parameter data to achieve flight subject identification.

[0013] Furthermore, in step three, during feature extraction, the features for "time" are takeoff time, landing time, and total flight time; the features for "total aircraft weight" are takeoff weight and landing weight; and the features for "external load weight" are takeoff external load weight and landing external load weight. For all other flight parameters besides time, total aircraft weight, and external load weight, the features are their maximum value, minimum value, standard deviation, and mean. Using center-of-gravity normal overload and total aircraft weight as inputs, the equivalent damage and nominal damage for the takeoff and landing are calculated and also used as feature data for that takeoff and landing.

[0014] Furthermore, in step one, each takeoff and landing flight parameter is processed for bit errors, out-of-bounds data removal, short-range filtering, and peak filtering; and the flight parameter data is judged by the flight altitude to remove the ground segment data and retain only the flight parameter data in the air segment, thereby reducing the scale of input data.

[0015] Furthermore, by using time-series flight parameter data, different methods are employed to extract feature values ​​from flight parameter history data for different flight parameters. Based on the fatigue damage algorithm, the equivalent damage and nominal damage of the takeoff and landing are calculated to realize the overall technical process of flight subject prediction.

[0016] Furthermore, in step one, during the flight parameter screening for the flight subject identification model, prior knowledge is used to initially screen the flight parameter data and determine the initial selection of flight parameter data. This is then combined with regression analysis to ultimately determine the input flight parameters for the flight subject identification model.

[0017] Furthermore, to enable the machine learning model to effectively learn the product features between input independent variables, the flight parameter data product term described in step two is used to augment the data. Considering the possibility of strong linear correlations between data after augmentation, which could unnecessarily increase model training time, multicollinearity diagnosis is performed to retain one set of data with strong linear correlations. This reduces the dimensionality of the model's input independent variables, improving model stability while reducing the overall input data size. Furthermore, peak-valley filtering and small-range filtering algorithms can be used to filter the model input data based on center-of-gravity overload, further reducing the size of the model training input data.

[0018] Furthermore, the data augmentation method described in step two multiplies the input flight parameter data pairwise to obtain the product terms between the flight parameters, so that the machine learning model can learn the product features between the input value variables;

[0019] Furthermore, step two involves multicollinearity diagnosis. This process removes data with strong linear correlations among the input independent variables, thereby avoiding model instability caused by multicollinearity in the input data; it also achieves data dimensionality reduction and avoids unnecessary resource waste during model training.

[0020] Furthermore, the filtering method described in step two uses the center of gravity normal overload as the filtering object, employs small-range filtering to filter out small overload cycles and intermediate data points in each flight maneuver, and uses a peak-valley filtering algorithm to filter out flight parameter data cycles with small peaks and large valleys, further reducing the scale of input data and accelerating the model training speed.

[0021] Furthermore, the "nominal damage" and "equivalent damage" features mentioned in step three are based on the fact that damage varies across different flight maneuvers, while the damage differences within the same flight maneuver are relatively small, but the damage differences between different flight maneuvers are significant. Therefore, a fatigue damage algorithm is used to calculate nominal damage using the center-of-gravity normal overload history, and to calculate equivalent damage using the equivalent overload history obtained from the center-of-gravity normal overload history and the total aircraft weight history during takeoff and landing. Both "equivalent damage" and "nominal damage" are also considered key features of this takeoff and landing.

[0022] Furthermore, the method described in step four, subject identification is essentially a classification between different categories of data. Combining prior knowledge, it employs typical machine learning models suitable for classification algorithms, such as support vector machine classification algorithm and random forest model, to construct a model that uses time-series flight parameter feature values ​​as input to predict flight subjects.

[0023] Beneficial effects of the present invention

[0024] This invention establishes a flight course identification method based on time-series flight parameter data features. This method fills the gap in intelligent flight course identification and saves labor costs. Compared with manual marking of flight courses, it avoids the problem of missed marking of flight courses. This method realizes automatic identification of flight courses, providing strong support for the compilation of aircraft measured load spectra, field aircraft fatigue life determination, and ensuring aircraft operational safety. Attached Figure Description

[0025] Figure 1 A flowchart of an aircraft flight course identification method based on time-series flight parameter data;

[0026] Figure 2 Flowchart for inputting flight parameter screening into the flight subject identification model;

[0027] Figure 3 This is for time-series flight parameter feature extraction. Specific Implementation

[0028] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0029] Inspired by the above-mentioned technical concepts, this invention proposes a method for identifying aircraft flight maneuvers based on time-series flight parameter data, comprising the following implementation steps:

[0030] Step 1: Filtering multivariate time-series flight parameter data.

[0031] S1.1: Initial screening of flight parameter data, using prior knowledge to make preliminary judgments, retaining data closely related to flight subjects, such as time, external load weight, total aircraft weight, center of gravity normal overload, and center of gravity lateral overload;

[0032] S1.2: Conduct correlation analysis between physical parameters and flight parameters. Use multicollinearity analysis and stepwise regression to select flight parameters that are highly correlated with center of gravity normal overload and center of gravity lateral overload. The selected flight parameters include: time, altitude, Mach, rate of climb, normal overload, lateral overload, bank angle, pitch angle, heading angle, roll rate, pitch rate, and yaw rate data.

[0033] Step 2: Time series flight parameter data processing.

[0034] S2.1: Data Cleaning. Error and out-of-bounds data removal, short-range filtering, and peak filtering are performed on the flight parameters for each takeoff and landing. The flight parameter data is then classified into air segment data based on flight altitude, and ground segment data is removed, retaining only the air segment flight parameter data to reduce the size of the input data.

[0035] S2.2: Data Augmentation. After processing the flight parameter data in step two, multiply each parameter pairwise (excluding the time term) to obtain the product feature term between the flight parameter data.

[0036] S2.3: Data Compression. Methods such as multicollinearity diagnosis, amplitude filtering, and peak filtering are used to reduce the size of the sample space;

[0037] S2.4: Data Standardization. Due to the inconsistent magnitude of input data, such as the total weight of the machine being on the order of tens of thousands, it naturally carries a certain weight. To minimize the magnitude of the input parameters, data standardization is performed. Simultaneously, data standardization can eliminate the influence of noise. The data standardization method is shown in equation (1).

[0038] (1)

[0039] Where x represents the standardized data; μ represents the mean of the data; and σ represents the standard deviation of the data.

[0040] Step 3: Perform feature extraction on the time-series flight parameter data processed in Step 2.

[0041] S3.1: The characteristics of "time" are take-off time, landing time, and total flight time; the characteristics of "total aircraft weight" are take-off total aircraft weight and landing total aircraft weight; and the characteristics of "external load weight" are take-off external load weight and landing external load weight as time series data characteristics.

[0042] S3.2: Except for the three parameters mentioned in S3.1, the maximum value, minimum value, average value, and variance of the other parameters are extracted as the feature values ​​of their time series data;

[0043] S3.3: Using the center-of-gravity normal overload and the total aircraft weight as inputs, calculate the equivalent and nominal damage for this takeoff and landing, which also serve as characteristic data for this takeoff and landing. The nominal damage calculation uses the center-of-gravity normal overload from the flight parameters. (Nominal overload), equivalent damage is calculated using equivalent overload. Calculation. The equivalent overload is calculated as shown in equation (2), where, For nominal overload, For equivalent overload, W is The corresponding total weight of the machine, Weight designed for strength:

[0044] (2)

[0045] The method for calculating equivalent damage is as follows:

[0046] 1) For the selected load spectrum, extract the complete overload cycle using the rainflow counting method.

[0047] After processing with rainflow counting, the original overload spectrum exhibits overload cycling. The combination, in which The cyclic characteristic (overload ratio) of the i-th overload cycle.

[0048] 2) Convert the overload cycles at each level in the overload spectrum into pulsating cycles (R=0) using the Odin transformation.

[0049] The Odin transformation formula is as follows:

[0050] (3)

[0051] In formula (3) Indicates the peak value of stress cycles. If the stress ratio is given, then the equivalent load corresponding to the i-th cycle is...

[0052] (4)

[0053] 3) Calculate the equivalent damage for each cycle based on the S-N curve (m=4) under pulsating circulation;

[0054] Under cyclic pulsating loads, the stress fatigue performance S-N curve of metallic materials can be described by a power function, expressed as:

[0055] (5)

[0056] In equation (5), N represents the number of cycles, and S represents the stress. When calculating nominal damage or equivalent damage, the following formula is used: or Replace S and calculate the damage for each cycle. The damage for the i-th cycle can then be expressed as:

[0057] (6)

[0058] The total damage value is denoted as ,but

[0059] (7)

[0060] S3.4: Combine the flight parameter characteristic values ​​of each takeoff and landing sequence obtained from S3.2 and S3.3 with the corresponding flight course information to construct a flight parameter characteristic & flight course database;

[0061] Step 4: Using machine learning methods such as support vector machines and random forests, construct a flight subject prediction model based on time-series flight parameter data values ​​to achieve flight subject identification.

[0062] Example 1: Taking a certain type of aircraft as an example, the specific technical process of the present invention is as follows: Figure 1 As shown, a flight course identification model is constructed based on the design concept of this invention. There are 6344 takeoff and landing data for this type of aircraft, of which 2605 takeoffs and landings were missing or obviously mislabeled, and these were removed, leaving 3739 takeoffs and landings. Since it is impossible to identify takeoffs and landings with correct but mislabeled flight courses, the flight course identification model is constructed using the 3739 takeoff and landing data with flight course information:

[0063] 1) Following step one, using the time-series flight parameter data and flight course information as input, and then... Figure 2 The data processing method shown involves filtering the input flight parameters of multivariate time-series flight data. The 14 flight parameters obtained after filtering are: time, altitude, Mach, rate of climb, normal overload, lateral overload, bank angle, pitch angle, yaw angle, roll rate, pitch rate, and yaw rate.

[0064] For multivariate flight parameters, stepwise regression and principal component analysis were used to determine the flight parameters required for flight subject identification, namely: equivalent overload, roll rate voting signal, pitch rate voting signal, yaw rate voting signal, flight altitude, Mach number, rudder position signal, left outer elevon position signal, left inner elevon position signal, right outer elevon position signal, right inner elevon position signal, left leading edge flap position signal, right leading edge flap position signal, and lateral acceleration voting signal.

[0065] 2) Following step two, first multiply the flight parameter data in pairs, except for "time", to generate new parameter terms, such as flight altitude × Mach number, Mach number × rudder position signal, rudder position signal × left outer elevator position signal, and other nonlinear terms; then, use methods such as multicollinearity diagnosis, amplitude filtering, and peak filtering to reduce the size of the sample space (dimensionality reduction) and standardize the data.

[0066] 3) Following step three, extract the time-series flight parameter features for takeoff and landing of aircraft 3739. The features for "time" are takeoff time, landing time, and total flight time; the features for "total aircraft weight" are takeoff weight and landing weight; the features for "external load weight" are takeoff external load weight and landing external load weight; and the features for other flight parameters are maximum value, minimum value, standard deviation, and mean value. Nominal damage is calculated using nominal overload data, and equivalent overload is calculated using nominal overload combined with total aircraft weight to determine equivalent damage. Furthermore, by combining information from each takeoff and landing flight maneuver, a "flight parameter feature data and flight maneuver database" is constructed, such as... Figure 3 As shown.

[0067] 4) Following step four: From the 3739 takeoff and landing data points in the "Flight Parameter Feature Data and Flight Subject Database," 2991 takeoff and landing data points (80%) were randomly selected to construct a model for predicting flight subjects based on input flight parameter features. The prediction model used a Support Vector Machine (SVM) model, with a second-order polynomial kernel, a penalty coefficient C of 5.0, and independent terms coef0=5.0 for training the flight subject prediction model. After model training, the remaining 748 (20%) takeoff and landing data points that were not used in model training were used to test the model's performance. Due to insufficient sample data and human labeling errors, the accuracy of the constructed model was 93.5%.

[0068] The above description is merely one specific embodiment of the present invention, providing a detailed description of the invention. Parts not covered herein are conventional techniques. However, the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for identifying aircraft flight maneuvers based on time-series flight parameter data, characterized in that, The method includes the following steps: Step S1: Flight parameter data screening and cleaning; initially, the default retention time, external load weight, total aircraft weight, center of gravity normal overload, and center of gravity lateral overload are retained; by conducting correlation analysis between physical parameters and flight parameters, stepwise regression is used to screen out flight parameters with high correlation to center of gravity normal overload and center of gravity lateral overload; then, error data, out-of-bounds data, and ground segment data of the flight parameters are removed to reduce the overall data size; Step S2: Data processing; After processing the flight parameter data in step S1, except for the time term, new parameter terms are generated based on the physical meaning, mathematical and / or mechanical model of the flight parameters; then, multicollinearity diagnosis and filtering methods are used to reduce the size of the sample space and standardize the data to improve model stability. Step S3: Time-series data feature extraction; for the time-series flight parameter data processed in step S2, feature extraction is performed to construct a flight parameter feature database; and using the center of gravity normal overload and total aircraft weight as inputs, the equivalent damage and nominal damage of takeoff and landing are calculated, which are also used as feature data of the takeoff and landing; then, combined with the information of each takeoff and landing flight subject, a typical flight parameter feature & flight subject database is constructed. Step S4: Using typical flight parameter features and flight subject database as input, construct a flight subject prediction model based on time-series flight parameter data to achieve flight subject identification.

2. The method for identifying aircraft flight maneuvers based on time-series flight parameter data according to claim 1, characterized in that: In step S1, during the flight parameter data screening, the flight parameter data is initially screened based on prior knowledge, and combined with a regression analysis algorithm to determine the final input flight parameters.

3. The method for identifying aircraft flight maneuvers based on time-series flight parameter data according to claim 1, characterized in that: In step S1, error correction, out-of-bounds data removal, short-range filtering, and peak filtering are performed on the flight parameters for each takeoff and landing. The flight parameter data is then classified into air segment data based on the flight altitude, and ground segment data is removed, retaining only the air segment flight parameter data to reduce the size of the input data.

4. The method for identifying aircraft flight maneuvers based on time-series flight parameter data according to claim 1, characterized in that: In step S2, during data processing, the data is augmented using the product term of flight parameter data. Considering the possibility of strong linear correlation between data after data augmentation, which may cause an unnecessary increase in model training time, multicollinearity diagnosis is carried out to retain one of the multiple sets of data with strong linear correlation, thereby achieving dimensionality reduction of the model input independent variables.

5. The method for identifying aircraft flight maneuvers based on time-series flight parameter data according to claim 4, characterized in that: By employing peak-valley filtering and small-range filtering algorithms, and using center-of-gravity overload as a benchmark, the model input data is filtered to further reduce the scale of the model training input data.

6. The method for identifying aircraft flight maneuvers based on time-series flight parameter data according to claim 1, characterized in that: Feature extraction of time-series flight parameter data: the features of "time" are takeoff time, landing time and total flight time, the features of "total aircraft weight" are takeoff weight and landing weight, and the features of "external load weight" are takeoff external load weight and landing external load weight; for the other flight parameters other than time, total aircraft weight and external load weight, the features are their maximum value, minimum value, standard deviation and mean.

7. The method for identifying aircraft flight maneuvers based on time-series flight parameter data according to claim 1, characterized in that: In step S3, the fatigue damage algorithm is used to calculate the "equivalent damage" and "nominal damage" obtained by the center of gravity normal overload history and the total weight history of the aircraft during takeoff and landing, respectively. These are also used as key features of the takeoff and landing.

8. The method for identifying aircraft flight maneuvers based on time-series flight parameter data according to claim 1, characterized in that: We employ machine learning methods such as support vector machines and random forests to construct a time-series flight parameter feature value-flight subject prediction model based on support vector machine and random forest machine learning models.

9. The method for identifying aircraft flight maneuvers based on time-series flight parameter data according to claim 1, characterized in that: In step S1, the 14 flight parameters finally selected include: time, external load weight, total weight, center of gravity normal overload, center of gravity lateral overload, altitude, Mach number, rate of climb and climb, roll rate, pitch rate, yaw rate, bank angle, pitch angle, and heading angle.

Images