Full-size strength test reaction force prediction method based on data fusion

Abstract

The invention relates to the technical field of airplane structure strength tests, in particular to a full-size strength test reaction force prediction method based on data fusion, which comprises the following steps: acquiring historical test load data, and constructing a target prediction model; constructing a bearing reaction calculation model considering deformation; acquiring a loading force line angle and an actual load corresponding to each loading force point of the formal / high-load test; and predicting the bearing reaction in the formal / high-load test process. According to the method, the actual load uncertainty (the actual load mean value and the actual load standard deviation) and the influence of aircraft structure deformation in the test can be effectively considered, and the bearing reaction of the pre-test / low-load test can be predicted more accurately.

Description

Technical Field

[0001] The invention relates to the technical field of aircraft structure strength test, and in particular to a full-scale strength test support reaction force prediction method based on data fusion. Background Art

[0002] Full-scale aircraft structural testing is a complex and large-scale test. It simulates the various complex loads that an aircraft may encounter in actual flight by actively applying external loads indoors. It aims to verify the structural performance of the aircraft under different working conditions and ensure its integrity, durability and safety under various working conditions. In aircraft structural strength tests, it is crucial to accurately evaluate the accuracy of test loading. By comparing the theoretical support reaction force and the actual support reaction force of the constraint point, the accuracy of the test loading can be effectively judged. During the test, factors such as loading equipment errors and structural deformation may cause changes in the load direction and the position of the resultant force point, thereby affecting the reliability of the test results. By comparing the theoretical support reaction force with the actual support reaction force, errors can be discovered and corrected in a timely manner, and the accuracy and reliability of load application can be evaluated. Accurate support reaction force calculation not only helps to determine whether the loading is accurate, but also provides a basis for optimizing the test system and scheme design, ensuring the validity of the test results and the safety of the test system.

[0003] However, due to the high complexity of the full-scale structural strength test system, the calculation of support reaction forces faces the following challenges:

[0004] (1) Influence of structural deformation: Since the aircraft structure will deform under loading during the structural strength test, the deformation of the structure will cause the direction of the load force line to change. Therefore, accurate support reaction force calculation needs to consider the influence of structural deformation. How to consider the support reaction force calculation under deformation conditions is a key issue.

[0005] (2) Influence of load uncertainty: Structural strength test is a complex system engineering. During the test, it may be affected by factors such as loading equipment error, loading method deviation, and constraint support system deviation. There is an error between the actual loading load and the theoretical applied load. How to consider the influence of load uncertainty and accurately predict the actual loading load has become another key issue in the calculation of support reaction force.

[0006] Especially under high-load test conditions, these problems will be further exacerbated. First, as the loading load increases, the deformation of the aircraft structure continues to increase, making the influence of structural deformation on the calculation of support reaction more significant. Second, the loading error of the test system will also be amplified as the load increases, and the load uncertainty will increase accordingly, further increasing the deviation between the actual load and the theoretical load.

[0007] Traditional high-load reaction force calculation methods usually calculate based on the theoretical loads under various loading conditions and assume that the aircraft is a rigid body. Due to the many uncertainties in the test system, the actually applied loads do not exactly match the theoretical loads, and the aircraft will undergo structural deformation during the loading process. The traditional methods do not consider the influence of load uncertainties and structural deformation, resulting in low accuracy of the predicted reaction forces.

[0008] Therefore, it is necessary to provide a reaction force prediction method for full-scale strength tests based on data fusion to solve the above problems. Summary of the Invention

[0009] The present invention provides a reaction force prediction method for full-scale strength tests based on data fusion, which establishes a target prediction model based on historical test load data, combines pre-test / low-load test data, and considers the influence of actual load mean, actual load standard deviation, and structural deformation to comprehensively and accurately predict the reaction forces of formal / high-load tests to solve the existing problems.

[0010] A reaction force prediction method for full-scale strength tests based on data fusion of the present invention adopts the following technical solutions, including:

[0011] Obtain historical test load data collected in the full-scale structural strength test system;

[0012] Based on the historical test load data, construct a target prediction model for predicting the actual load mean and actual load standard deviation;

[0013] According to the coordinates of the constraint points of the statically determinate support system of the full-scale structural strength test system, the load directions of the constraint points, the coordinates of the loading force points, the loading force line angles at each loading stage of the loading force points, and the load values, construct a reaction force calculation model considering deformation;

[0014] According to the loading force line angles during the pre-test / low-load test process, establish a loading force line angle prediction model, and predict the loading force line angles corresponding to each loading force point of the formal / high-load test based on the loading force line angle prediction model;

[0015] Based on the load data of the pre-test / low-load test and the target prediction model, predict the actual load mean and actual load standard deviation. Based on the predicted actual load mean and actual load standard deviation, use Monte Carlo simulation to simulate the loads of the formal / high-load test to obtain the actual loads corresponding to each loading point of the formal / high-load test;

[0016] Based on the actual loads and loading force line angles corresponding to each loading point of the formal / high-load test, and use the reaction force calculation model to predict the reaction forces during the formal / high-load test process.

[0017] Preferably, the steps for obtaining the historical test load data collected in the full-scale structural strength test system are as follows:

[0018] Take the historical test load data collected in the full-scale structural strength test system as the initial test load data;

[0019] Perform data processing on the initial test load data to obtain the historical test load data. The data processing includes: missing value processing, outlier detection, and outlier processing.

[0020] Preferably, the steps for constructing a target prediction model for predicting the actual load mean and the actual load standard deviation based on the historical test load data are as follows:

[0021] Based on the historical test load data, construct an actual load mean prediction model using the least squares method;

[0022] Based on the historical test load data, establish an actual load standard deviation prediction model using the random forest algorithm;

[0023] Use the genetic algorithm to optimize the model parameters of the actual load standard deviation prediction model to obtain the optimized actual load standard deviation prediction model;

[0024] Combine the actual load mean prediction model and the optimized actual load standard deviation prediction model to obtain the target prediction model.

[0025] Preferably, the expression of the actual load mean prediction model is:

[0026] y = 0.999708 * x + 1.1778

[0027] In the formula, x represents the historical test load data; y represents the predicted actual load mean.

[0028] Preferably, the steps for constructing a reaction force calculation model considering deformation are as follows:

[0029] According to the coordinates of the constraint points and the load directions of the constraint points of the statically determinate support system in the full-scale structural strength test system, obtain the constraint point coordinate matrix;

[0030] According to the coordinates of the loading force points of the statically determinate support system in the full-scale structural strength test system, the loading force line angles at each loading stage, and the load values, obtain the load resultant matrix during the test;

[0031] According to the constraint point coordinate matrix and the load resultant matrix, obtain the reaction force calculation model considering deformation.

[0032] Preferably, the expression of the constraint point coordinate matrix is:

[0033]

[0034] In the formula, x in is the x - coordinate of the constraint point i at the n - th loading stage; y in is the y - coordinate of the constraint point i at the n - th loading stage; z in is the z - coordinate of the constraint point i at the n - th loading stage; cosα in is the cosine value of the angle between the load direction and the x - axis direction of the constraint point i at the n - th loading stage; cosβ in is the cosine value of the angle between the load direction and the y - axis direction of the constraint point i at the n - th loading stage; cosγ in is the cosine value of the angle between the load direction and the z - axis direction of the constraint point i at the n - th loading stage; α in is the component of the loading force line angle of the constraint point i at the n - th loading stage in the x - axis direction; β in is the component of the loading force line angle of the constraint point i at the n - th loading stage in the y - axis direction; γ in is the component of the loading force line angle of the constraint point i at the n - th loading stage in the z - axis direction; where i = 1, 2, 3…6 is the number of the constraint point.

[0035] Preferably, the expression of the resultant load matrix is:

[0036]

[0037] In the formula, m represents the total number of loading force points, and the loading force points include the active loading points, the fastening points and the total weight of the aircraft; j represents the loading point number, x jn represents the x - coordinate of the loading force point j at the n - th loading stage; y jn represents the y - coordinate of the loading force point j at the n - th loading stage; z jn represents the z - coordinate of the loading force point j at the n - th loading stage; α jn is the component of the loading force line angle of the loading force point j at the n - th loading stage in the x - axis direction; β jn is the component of the loading force line angle of the loading force point j at the n - th loading stage in the y - axis direction; γ jn is the component of the loading force line angle of the loading force point j at the n - th loading stage in the z - axis direction; F jn represents the load value of the loading force point j at the n - th loading stage; F xn represents the component of the load value of the resultant load at the n - th loading stage in the x - axis direction; F yn represents the component of the load value of the resultant load at the n - th loading stage in the y - axis direction; F zn represents the component of the load value of the resultant load at the n - th loading stage in the z - axis direction; M xnrepresents the component of the resultant loading moment in the x-axis direction at the nth loading level; M yn represents the component of the resultant loading moment in the y-axis direction at the nth loading level; M zn represents the component of the resultant loading moment in the z-axis direction at the nth loading level, where i = 1, 2, …, m is the number of the loading force points.

[0038] Preferably, the expression of the reaction force calculation model considering deformation is:

[0039] A n X n =B n n = 1, 2, …, D

[0040] In the formula, X n is the reaction force matrix of the constraint points corresponding to the nth load level; A n is the constraint point coordinate matrix; B n is the resultant load matrix during the test; n represents the nth load level; D is the total number of loading levels.

[0041] Preferably, according to the loading force line angle during the preliminary test / low-load test, the least squares method is used to establish a loading force line angle prediction model.

[0042] A reaction force prediction system for full-scale strength test based on data fusion adopts the following technical solutions, including:

[0043] A historical parameter acquisition module, which is used to acquire the historical test load data collected in the full-scale structural strength test system;

[0044] A target prediction model construction module, which is used to construct a target prediction model for predicting the actual load mean and the actual load standard deviation based on the historical test load data;

[0045] A reaction force calculation model construction module, which is used to construct a reaction force calculation model considering deformation according to the coordinates of the constraint points of the statically determinate support system in the full-scale structural strength test system, the load directions of the constraint points, the coordinates of the loading force points, the loading force line angles and the load values of the loading force points at each loading level;

[0046] A loading force line angle prediction module, which is used to establish a loading force line angle prediction model according to the loading force line angle during the preliminary test / low-load test, and predict the loading force line angles corresponding to each loading force point in the formal / high-load test based on the loading force line angle prediction model;

[0047] An actual load acquisition module, which is used to predict the actual load mean value and the actual load standard deviation based on the pretest / low-load test load data and the target prediction model, and based on the predicted actual load mean value and the actual load standard deviation, and use Monte Carlo simulation to simulate the load of the formal / high-load test to obtain the actual load corresponding to the formal / high-load test;

[0048] A reaction force prediction module, which is used to predict the reaction force during the formal / high-load test based on the actual load corresponding to the formal / high-load test, the loading force line angles corresponding to each loading force point of the formal / high-load test, and use the reaction force calculation model.

[0049] The beneficial effects of the present invention are:

[0050] Through multi-source data fusion, a target prediction model for predicting the actual load mean value and the actual load standard deviation (uncertainty parameters) is established based on historical test data, and combined with the pretest / low-load test load data, the actual load of the formal / high-load test and its uncertainty parameters are accurately predicted. At the same time, a reaction force calculation model considering structural deformation is established, that is, a method proposed by the present invention that can effectively consider the actual load uncertainty (actual load mean value and actual load standard deviation) in the test and the influence of aircraft structural deformation, and more accurately predict the reaction force of the pretest / low-load test. It provides strong support for optimizing the test loading scheme, reducing equipment loss, shortening the test cycle and controlling economic costs. In addition, the method of the present invention is not only applicable to the reaction force calculation in the full-scale structural strength test system, but also can be extended and applied to the reaction force prediction of other complex systems, promoting the progress and application development of related technologies, and having a wide application prospect. Description of the Drawings

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0052] Figure 1 It is a flowchart of a full-scale strength test reaction force prediction method based on data fusion according to the present invention;

[0053] Figure 2 It is a flowchart of the reaction force prediction process in the formal / high-load test in the embodiment of the present invention;

[0054] Figure 3 It is a Pearson rank correlation coefficient heat map in the embodiment of the present invention;

[0055] Figure 4 This is the fitting result graph of the actual load mean prediction model using the least squares method in the embodiments of the present invention;

[0056] Figure 5 This is the prediction result graph of the actual load standard deviation prediction model based on GA_RF in the embodiments of the present invention;

[0057] Figure 6 This is the prediction result graph of the reaction force in the formal / high load test in the embodiments of the present invention;

[0058] Figure 7 This is the prediction result graph of the reaction force at the 140% loading level in the formal / high load test in the embodiments of the present invention. Detailed implementation manners

[0059] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0060] An embodiment of a method for predicting the reaction force of a full-scale strength test based on data fusion according to the present invention is as Figure 1 and Figure 2 shown, and includes:

[0061] S1. Obtain historical test load data and construct a target prediction model;

[0062] Specifically, S11. Obtain historical test load data collected in the full-scale structural strength test system; S12. Based on the historical test load data, construct a target prediction model for predicting the actual load mean and the actual load standard deviation.

[0063] The steps of S11 for obtaining historical test load data collected in the full-scale structural strength test system are: taking the historical test load data collected in the full-scale structural strength test system as the initial test load data; performing data processing on the initial test load data to obtain historical test load data, and the data processing includes: missing value processing, outlier detection, and outlier processing,

[0064] Exemplarily, in a specific embodiment, in step S11, historical data of three aircraft structural strength tests are collected as initial test load data, missing values in the load data are identified, and polynomial interpolation method is used to fill the missing values; since the distribution type of the load data is unknown, the interquartile range method (IQR) is used for outlier detection. Since the process of aircraft structural strength test is complex, the cost of re-collecting load data is high and the scale of collected load is large. Therefore, in this embodiment, the direct deletion method is adopted for outlier processing of the load data outliers, and the data after missing value processing, outlier detection and outlier processing are used as historical test load data.

[0065] S12. The steps of constructing a target prediction model for predicting the actual load mean and the actual load standard deviation based on the historical test load data are as follows: Based on the historical test load data, the least squares method is used to construct an actual load mean prediction model; based on the historical test load data, the random forest algorithm is used to establish an actual load standard deviation prediction model; the genetic algorithm is used to optimize the model parameters of the actual load standard deviation prediction model to obtain an optimized actual load standard deviation prediction model; the actual load mean prediction model and the optimized actual load standard deviation prediction model are combined to obtain a target prediction model.

[0066] Exemplarily, in a specific embodiment, first, on the basis of step S11, the distribution type analysis of the historical test load data and the feature analysis of the historical test load data are carried out; among them, for the distribution type analysis of the historical test load data: in this embodiment, the graphical method and the D-K test (D'Agostino's K-squared test) method are used to determine the distribution type of the load data. First, the histogram is used for preliminary analysis of the load distribution type, and it can be intuitively seen that the histogram shapes of each group of load data are "bell-shaped", conforming to the characteristics of normal distribution. Subsequently, the D-K test is used to carry out further quantitative analysis on the load distribution type. The results show that at the 95% confidence level, all groups of data accept the normal distribution hypothesis. In summary, it can be considered that the load data follows a normal distribution; among them, for the feature analysis of the historical test load data: the Pearson rank correlation coefficient method is used to carry out feature analysis on the load uncertainty parameters. The Pearson rank correlation coefficient heat map, as Figure 3 shown, through correlation analysis, it can be obtained that the theoretical load and the actual load mean have a completely positive linear correlation; there is a strong positive correlation between the theoretical load, the load change amount and the actual load standard deviation, and there is a weak equal-strength positive correlation between the number of loading levels and the actual load standard deviation. Therefore, the theoretical load is selected to predict the actual load mean, and the theoretical load, the load change amount and the number of loading levels are selected to predict the actual load standard deviation.

[0067] Combined with the results of the distribution type analysis and the feature analysis of historical test load data, since there is a linear relationship between the actual load mean and the theoretical load, the least squares method is used in this embodiment to establish an actual load mean prediction model. To improve the generalization ability of the prediction model, K-fold cross-validation is used to divide the data set into a training set and a test set, and the fitting results are as follows Figure 4 shown. The mean squared error (MSE) and the coefficient of determination (R 2 ) are used as evaluation indicators for the model prediction ability. The MSE of the actual load mean prediction model based on the least squares method is 29.5826, and R 2 is 0.9999, indicating that the model has good prediction performance, so there is no need to further optimize the model parameters. The expression of the actual load mean prediction model is obtained as:

[0068] y = 0.999708 * x + 1.1778

[0069] where x represents the historical test load data; y represents the predicted actual load mean.

[0070] Exemplarily, in a specific embodiment, the selected theoretical load, load change amount, and loading level in the results of the feature analysis of historical test load data are used to predict the actual load standard deviation. In this embodiment, the random forest algorithm (Random Forest, RF) is used to establish an actual load standard deviation prediction model. The historical test load data is standardized and normalized, and the data set is divided by K-fold cross-validation. The mean squared error (MSE) and the coefficient of determination (R 2 ) are used as evaluation indicators for the model prediction ability. The MSE of the actual load standard deviation prediction model based on the RF algorithm is 2.2546, and R 2 is 0.8836. To further improve the prediction accuracy, the genetic algorithm (Genetic Algorithm, GA) is used to optimize the parameters of the RF actual load standard deviation prediction model. After optimization, the MSE of the RF model optimized by the GA algorithm is 2.0154, and R 2 is 0.9018, indicating that the actual load standard deviation prediction model has good prediction performance. Therefore, the optimized GA_RF model is selected as the actual load standard deviation prediction model. Among them, the actual load standard deviation prediction results based on GA_RF are as follows Figure 5 shown.

[0071] Exemplarily, in a specific embodiment, the actual load mean prediction model and the optimized actual load standard deviation prediction model are combined to obtain a target prediction model.

[0072] S2. Construct a reaction force calculation model considering deformation;

[0073] Specifically, based on the coordinates of the constraint points of the statically determinate support system of the full-scale structural strength test system, the load directions of the constraint points, the coordinates of the loading force points, the loading force line angles at each loading level of the loading force points, and the load values, construct a reaction force calculation model considering deformation.

[0074] Exemplarily, in a specific embodiment, the steps of constructing a reaction force calculation model considering deformation are as follows: S21. Based on the coordinates of the constraint points of the statically determinate support system of the full-scale structural strength test system and the load directions of the constraint points, obtain the constraint point coordinate matrix; S22. Based on the coordinates of the loading force points of the statically determinate support system of the full-scale structural strength test system, the loading force line angles at each loading level of the loading force points, and the load values, obtain the load resultant matrix during the test; S23. Based on the constraint point coordinate matrix and the load resultant matrix, obtain the reaction force calculation model considering deformation.

[0075] Exemplarily, in a specific embodiment, in step S21, based on the coordinates of the constraint points of the statically determinate support system of the full-scale structural strength test system and the load directions of the constraint points, obtain the constraint point coordinate matrix; wherein, the expression of the constraint point coordinate matrix is:

[0076]

[0077] In the formula, x in is the x coordinate of the constraint point i at the nth loading level; y in is the y coordinate of the constraint point i at the nth loading level; z in is the z coordinate of the constraint point i at the nth loading level; cosα in is the cosine value of the angle between the load direction of the constraint point i at the nth loading level and the x-axis direction; cosβ in is the cosine value of the angle between the load direction of the constraint point i at the nth loading level and the y-axis direction; cosγ in is the cosine value of the angle between the load direction of the constraint point i at the nth loading level and the z-axis direction; α in is the component of the loading force line angle of the constraint point i at the nth loading level in the x-axis direction (i.e., the angle between the load direction of the constraint point i at the nth loading level and the x-axis direction); β in is the component of the loading force line angle of the constraint point i at the nth loading level in the y-axis direction (i.e., the angle between the load direction of the constraint point i at the nth loading level and the y-axis direction); γ inIt is the component of the loading force line angle of the constraint point i at the nth loading level in the z-axis direction (i.e., the angle between the load direction of the constraint point i at the nth loading level and the z-axis direction), where i = 1, 2, 3... 6 is the number of the constraint points.

[0078] Exemplarily, in a specific embodiment, in step S22, according to the coordinates of the loading force points of the statically determinate support system of the full-scale structural strength test system, the loading force line angles of the loading force points at each loading level, and the load values, the load resultant force matrix during the test is obtained; the expression of the load resultant force matrix is:

[0079]

[0080] In the formula, m represents the total number of loading force points, and the loading force points include the active loading points, the key points, and the total weight of the aircraft; j represents the loading point number, x jn represents the x coordinate of the loading force point j at the nth loading level; y jn represents the y coordinate of the loading force point j at the nth loading level; z jn represents the z coordinate of the loading force point j at the nth loading level; α jn is the component of the loading force line angle of the loading force point j at the nth loading level in the x-axis direction (i.e., the angle between the load direction of the loading force point j at the nth loading level and the x-axis); β jn is the component of the loading force line angle of the loading force point j at the nth loading level in the y-axis direction (i.e., the angle between the load direction of the loading force point j at the nth loading level and the y-axis); γ jn is the component of the loading force line angle of the loading force point j at the nth loading level in the z-axis direction (i.e., the angle between the load direction of the loading force point j at the nth loading level and the z-axis); F jn represents the load value of the loading force point j at the nth loading level; F xn represents the component of the load value of the loading resultant force at the nth loading level in the x-axis direction; F yn represents the component of the load value of the loading resultant force at the nth loading level in the y-axis direction; F zn represents the component of the load value of the loading resultant force at the nth loading level in the z-axis direction; M xn represents the component of the resultant moment of the loading resultant force at the nth loading level in the x-axis direction; M yn represents the component of the resultant moment of the loading resultant force at the nth loading level in the y-axis direction; M zn represents the component of the resultant moment of the loading resultant force at the nth loading level in the z-axis direction, where i = 1, 2, 3... m is the number of the loading force points.

[0081] Exemplarily, in a specific embodiment, in step S23, according to the constraint point coordinate matrix and the resultant load matrix, a reaction force calculation model considering deformation is obtained. The expression of the reaction force calculation model is:

[0082] A n X n =B n n = 1, 2,..., D

[0083] In the formula, X n is the reaction force matrix of the constraint point corresponding to the nth level of load series; A n is the constraint point coordinate matrix; B n is the resultant load matrix during the test; n represents the nth level of load series; D is the total number of loading levels.

[0084] Among them, the expression of the reaction force matrix of the constraint point is:

[0085] X n =[X 1n X 2n X 3n X 4n X 5n X 6n T

[0086] In the formula, X in is the reaction force of the constraint point i at the nth level of loading series, where i = 1, 2, 3... 6 is the number of the constraint point.

[0087] Among them, in this embodiment, according to the known conditions, the parameters shown in Table 1 - Table 5 are set in advance. Among them, the constraint point coordinates and the loading force line angles of the constraint points required for calculating the reaction force are shown in Table 1, the loading point coordinates and the loading force line angles of the loading points are shown in Table 2, the total weight of the aircraft is shown in Table 3, the loading point loads are shown in Table 4; the measurement data of the three-axis angles of the loading force lines are shown in Table 5.

[0088] Table 1

[0089] Constraint point Vertical angle α Spanwise angle β Course angle γ Vertical coordinate x Spanwise coordinate y Course coordinate z Vertical of front main landing gear x1 180 90 90 -3000 0 5070 Vertical of left main landing gear X2 180 90 90 -1500 3795 17710 Vertical of right main landing gear x3 180 90 90 -1500 -3795 17710 Course of front main landing gear x4 90 90 0 -2000 0 5070 Lateral of front main landing gear x5 90 180 90 -2000 0 5070 Lateral of left main landing gear X6 90 0 90 -1500 3795 17710

[0090] Table 2

[0091] Point number Part name Vertical angle α Spanwise angle β Course angle γ Vertical coordinate z Spanwise coordinate y Course coordinate x C1 Front section of fuselage 45.1 69.7 52.4 0 0 3000 C2 Left wing tip 0 90 90 0 17900 17000 C3 Middle of left wing 0 90 90 0 9950 17000 C4 Right wing tip 0 90 90 0 -17900 17000 C5 Middle of right wing 0 90 90 0 -9950 17000 C6 Left wing root 0 90 90 0 2000 17000 C7 Right wing root 0 90 90 0 -2000 17000 C8 Front middle section of fuselage 180 90 90 -1000 0 8000 C9 Middle section of fuselage 180 90 90 -1000 0 20000 C10 Rear middle section of fuselage 180 90 90 -1000 0 32000 C11 Front section of fuselage 0 90 90 -1000 0 2000 C12 Rear section of fuselage 0 90 90 -1000 0 35000

[0092] Table 3

[0093] Point number Vertical angle α Spanwise angle β Course angle γ Vertical coordinate z Spanwise coordinate y Course coordinate x Weight Total weight of aircraft test 180 90 90 0 0 16000 70000

[0094] Table 4

[0095] Number of loading levels C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 Cll C12 10％ 1000 1000 1500 2000 1000 1500 2000 5]67 3333 1500 34000 34000 20％ 2000 2000 3000 4000 2000 3000 4000 10333 6667 3000 34000 34000 30％ 3000 3000 4500 6000 3000 4500 6000 15500 10000 4500 34000 34000 40％ 4000 4000 6000 8000 4000 6000 8000 20667 13333 6000 34000 34000 50％ 5000 5000 7500 10000 5000 7500 10000 25833 16667 7500 34000 34000 60％ 6000 6000 9000 12000 6000 9000 12000 31000 20000 9000 34000 34000 70％ 7000 7000 10500 14000 7000 10500 14000 36167 23333 10500 34000 34000 80％ 8000 8000 12000 16000 8000 12000 16000 41333 26667 12000 34000 34000 90％ 9000 9000 13500 18000 9000 13500 18000 46500 30000 13500 34000 34000 100％ 10000 10000 15000 20000 10000 15000 20000 51667 33333 15000 34000 34000 110％ 11000 11000 16500 22000 11000 16500 22000 56833 36667 16500 34000 34000 120％ 12000 12000 18000 24000 12000 18000 24000 62000 40000 18000 34000 34000 130％ 13000 13000 19500 26000 13000 19500 26000 67167 43333 19500 34000 34000 135％ 13500 13500 20250 27000 13500 20250 27000 69750 45000 20250 34000 34000 140％ 14000 14000 21000 28000 14000 21000 28000 72333 46667 21000 34000 34000 150％ 15000 15000 22500 30000 15000 22500 30000 77500 50000 22500 34000 34000

[0096] Table 5​

[0097] Number of loading levels α1 β1 γ1 α2 β2 γ2 α3 β3 γ3 α4 β4 γ4 10％ 45.01 69.27 52.27 0.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 20％ 45.0t 69.27 52.27 0.08 90.08 90.00 0.21 90.21 90.00 0.08 90.08 90.00 30％ 45.01 69.27 52.27 0.15 90.15 90.00 0.42 90.42 90.00 0.15 90.15 90.00 40％ 45.01 69.27 52.27 0.23 90.23 90.00 0.62 90.62 90.00 0.23 90.23 90.00 50％ 45.01 69.27 52.27 0.30 90.30 90.00 0.83 90.83 90.00 0.30 90.30 90.00 60％ 45.01 69.27 52.27 0.38 90.38 90.00 1.04 91.04 90.00 0.38 90.38 90.00 70％ 45.01 69.27 52.27 0.46 90.46 90.00 1.25 91.25 90.00 0.46 90.46 90.00 80％ 45.01 69.27 52.27 0.53 90.53 90.00 1.46 91.46 90.00 0.53 90.53 90.00 90％ 45.01 69.27 52.27 0.61 90.61 90.00 1.66 91.66 90.00 0.61 90.61 90.00 100％ 45.01 69.27 52.27 0.69 90.69 90.00 1.87 91.87 90.00 0.69 90.69 90.00 Number of loading levels α5 β5 γ5 α6 β6 γ6 α7 β7 γ7 α8 β8 γ8 10％ 0.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 20％ 0.21 90.21 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 30％ 0.42 90.42 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 40％ 0.62 90.62 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 50％ 0.83 90.83 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 60％ 1.04 91.04 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 70％ 1.25 91.25 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 80％ 1.46 91.46 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 90％ 1.66 91.66 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 100％ 1.87 91.87 90.00 0.00 90.00 90.00 0.00 90.00 90.00 180.00 90.00 90.00 Number of loading levels α9 β9 γ9 α10 β10 γ10 α11 β11 γ11 α12 β12 γ12 10％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 20％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 30％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 40％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 50％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 60％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 70％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 80％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 90％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00 100％ 180.00 90.00 90.00 180.00 90.00 90.00 0.00 90.00 90.00 0.00 90.00 90.00

[0098] S3. Obtain the loading force line angles and actual loads corresponding to each loading force point in the formal / high-load test;

[0099] Specifically, according to the loading force line angles in the preliminary / low-load test process, establish a loading force line angle prediction model, and based on the loading force line angle prediction model, predict the loading force line angles corresponding to each loading force point in the formal / high-load test; based on the load data and target prediction model in the preliminary / low-load test, predict the mean value and standard deviation of the actual load. Based on the predicted mean value and standard deviation of the actual load, and using Monte Carlo simulation to simulate the load in the formal / high-load test, obtain the actual load corresponding to the formal / high-load test.

[0100] Exemplarily, in a specific embodiment, according to the loading force line angles in the preliminary / low-load test process, use the least squares method to establish a loading force line angle prediction model. Specifically, in this embodiment, according to engineering experience, it can be known that there is a linear relationship between the loading force line angle and the loading level. Therefore, the loading force line angle prediction model can be established by using the loading force line angles measured in real time during the preliminary / low-load test process, so as to obtain the loading force line angles of each loading point in the formal / high-load test. Since the loading force line angle changes linearly with the loading level, the least squares method is selected to establish the loading force line angle prediction model. Taking the triaxial rotation angle data of the loading point C2 in Table 4 as an example, according to the components of the loading force line angles measured in real time in each axial direction during the preliminary / low-load test process, the loading force line angle prediction model is established by using the least squares method. The specific expression of the loading force line angle prediction model is:

[0101] α C2_n = 0.007624 * n - 0.076238

[0102] β C2_n = 0.006518 * n + 89.861348

[0103] γ C2_n = 0.001106 * n + 90.062414

[0104] Wherein, α C2_n is the component of the loading force line angle of the loading point C2 at the nth loading level in the x-axis direction (i.e., the angle between the load direction and the x-axis direction); β C2_n is the component of the loading force line angle of the loading point C2 at the nth loading level in the y-axis direction (i.e., the angle between the load direction and the y-axis direction); γ C2_nis the component of the loading force line angle at the loading force point C2 at the nth loading stage in the z-axis direction (i.e., the angle between the load direction and the z-axis direction); n represents the nth load stage.

[0105] So far, the loading force line angles corresponding to each loading point in the formal / high-load test can be predicted by the loading force line angle prediction model.

[0106] Exemplarily, in a specific embodiment, the steps to obtain the actual load in the formal / high-load test are as follows: Based on the target prediction model established in S2 and combined with the load data of the preliminary / low-load test, predict the actual load mean and actual load standard deviation corresponding to the formal / high-load test load. The actual load means of each loading point at the 110% - 150% loading stage are shown in Table 6, and the actual load standard deviations of each loading point at the 110% - 150% loading stage are shown in Table 7. According to the distribution type analysis result of the historical test load data, it can be known that the load follows a normal distribution. Use Monte Carlo simulation to simulate the load in the formal / high-load test to obtain the corresponding actual load in the formal / high-load test.

[0107] Table 6

[0108] Number of loading levels C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 110％ 11000.77 11000.81 16506.68 22007.97 11000.83 16506.58 22007.98 56820.75 36666.26 16506.55 33997.20 33997.35 120％ 12000.89 12000.95 18000.14 23999.89 12001.01 17999.96 23999.87 61984.67 39996.15 17999.89 33998.88 33998.56 130％ 12995.10 12995.11 19497.72 25990.81 12995.11 19497.72 25990.80 67144.44 43308.74 19497.74 33989.76 33989.35 135％ 13496.07 13496.08 20240.69 26990.23 13496.08 20240.61 26990.00 69746.25 44992.83 20240.61 33998.42 33998.43 140％ 13994.84 13994.87 20991.12 27987.03 13994.84 20991.16 27987.17 72308.85 46650.43 20991.21 33992.71 33992.73 145％ 14494.73 14494.74 21741.67 28987.23 14494.70 21742.04 28987.82 74892.91 48317.29 21741.66 33992.41 33992.83 150％ 14995.17 14995.01 22494.32 29991.19 14995.21 22494.41 29991.26 77479.36 49991.43 22494.14 33994.24 33994.81

[0109] Table 7

[0110] Number of loading levels C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 110％ 0.75 0.76 2.54 4.36 0.78 2.52 4.35 13.39 3.29 2.53 5.56 5.41 120％ 1.04 1.03 4.10 4.77 1.04 3.81 4.91 18.56 11.50 3.77 7.54 7.28 130％ 0.46 0.45 0.78 2.27 0.45 0.81 2.26 12.90 4.82 0.78 7.83 7.55 135％ 0.56 0.57 4.09 6.00 0.59 4.06 6.31 16.48 10.84 3.99 7.22 7.04 140％ 1.69 1.67 3.10 4.75 1.63 2.99 4.95 13.38 5.27 3.17 7.25 7.47 145％ 1.75 1.81 4.58 4.68 1.78 4.56 4.94 13.90 4.57 4.74 6.37 6.49 150％ 1.85 1.81 4.42 5.83 1.85 4.41 6.07 9.15 3.96 4.28 5.66 6.17

[0111] S4. Predict the reaction forces during the formal / high-load test;

[0112] Specifically, based on the actual load corresponding to the formal / high-load test, the loading force line angles corresponding to each loading force point in the formal / high-load test, and use the reaction force calculation model to predict the reaction forces during the formal / high-load test.

[0113] Exemplarily, in a specific embodiment, according to the reaction force calculation model, the reaction force prediction results of the formal / high-load test are obtained, as shown in Figure 6 And the distribution of the reaction forces of each loading point and its confidence interval at the 95% confidence level (i.e., the upper and lower limits of the reaction force distribution) are predicted. Taking the reaction force prediction results at the 140% loading stage as an example, the reaction force distribution diagram of each loading point is shown in Figure 7 shown, and the high-load reaction force prediction confidence interval is shown in Table 10.

[0114] Table 10

[0115]

[0116] As shown in Figure 6 are the reaction force prediction results of the formal / high-load test.Figure 7 For the predicted distribution of the reaction force at the 140% loading level, Table 10 gives the confidence intervals for the predicted reaction forces in the formal / high-load tests, verifying the accuracy and practicality of the reaction force prediction of this system in the full-scale strength test.

[0117] A reaction force prediction system for full-scale strength tests based on data fusion, comprising: a historical parameter acquisition module, a target prediction model construction module, a reaction force calculation model construction module, a loading force line angle prediction module, an actual load acquisition module, and a reaction force prediction module. Among them, the historical parameter acquisition module is used to acquire the historical test load data collected in the full-scale structural strength test system; the target prediction model construction module is used to construct a target prediction model for predicting the mean value and standard deviation of the actual load based on the historical test load data; the reaction force calculation model construction module is used to construct a reaction force calculation model considering deformation according to the coordinates of the constraint points of the statically determinate support system of the full-scale structural strength test system, the load directions of the constraint points, the coordinates of the loading force points, the loading force line angles and load values of the loading force points at each loading level; the loading force line angle prediction module is used to establish a loading force line angle prediction model based on the loading force line angles during the pretest / low-load test, and predict the loading force line angles corresponding to each loading force point in the formal / high-load test based on the loading force line angle prediction model; the actual load acquisition module is used to predict the mean value and standard deviation of the actual load based on the pretest / low-load test load data and the target prediction model, and simulate the load in the formal / high-load test using Monte Carlo simulation based on the predicted mean value and standard deviation of the actual load to obtain the actual load corresponding to the formal / high-load test; the reaction force prediction module is used to predict the reaction force during the formal / high-load test based on the actual load corresponding to the formal / high-load test, the loading force line angles corresponding to each loading force point in the formal / high-load test, and by using the reaction force calculation model.

[0118] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A prediction method for reaction force in full-scale strength test based on data fusion, characterized in that Including: Obtain historical test load data collected in the full-scale structural strength test system; Based on the historical test load data, construct a target prediction model for predicting the actual load mean and the actual load standard deviation; According to the coordinates of the constraint points of the statically determinate support system of the full-scale structural strength test system, the load directions of the constraint points, the coordinates of the loading force points, the loading force line angles and the load values of the loading force points at each loading stage, construct a reaction force calculation model considering deformation; According to the loading force line angles during the preliminary test / low-load test, establish a loading force line angle prediction model, and based on the loading force line angle prediction model, predict the loading force line angles corresponding to each loading force point in the formal / high-load test; Based on the load data of the preliminary test / low-load test and the target prediction model, predict the actual load mean and the actual load standard deviation. Based on the predicted actual load mean and the actual load standard deviation, and using Monte Carlo simulation to simulate the loads in the formal / high-load test, obtain the actual loads corresponding to each loading point in the formal / high-load test; Based on the actual loads and the loading force line angles corresponding to each loading point in the formal / high-load test, and use the reaction force calculation model to predict the reaction forces during the formal / high-load test.

2. The full-scale strength test reaction force prediction method based on data fusion according to claim 1, characterized in that The steps to obtain the historical test load data collected in the full-scale structural strength test system are: Take the historical test load data collected in the full-scale structural strength test system as the initial test load data; Perform data processing on the initial test load data to obtain the historical test load data. The data processing includes: missing value processing, outlier detection and outlier processing.

3. A full-scale strength test reaction force prediction method based on data fusion according to claim 1, characterized in that The steps to construct a target prediction model for predicting the actual load mean and the actual load standard deviation based on the historical test load data are: Based on the historical test load data, and use the least squares method to construct an actual load mean prediction model; Based on the historical test load data, and use the random forest algorithm to establish an actual load standard deviation prediction model; Use the genetic algorithm to optimize the model parameters of the actual load standard deviation prediction model to obtain the optimized actual load standard deviation prediction model; Combine the actual load mean prediction model and the optimized actual load standard deviation prediction model to obtain the target prediction model.

4. A full-scale strength test reaction force prediction method based on data fusion according to claim 3, characterized in that The expression of the actual load mean prediction model is: y = 0.999708 * x + 1.1778 In the formula, x represents the historical test load data; y represents the predicted actual load mean.

5. A full-scale strength test reaction force prediction method based on data fusion according to claim 1, characterized in that The steps to construct a reaction force calculation model considering deformation are: According to the coordinates of the constraint points of the statically determinate support system of the full-scale structural strength test system and the load directions of the constraint points, obtain the constraint point coordinate matrix; According to the coordinates of the loading force points of the statically determinate support system of the full-scale structural strength test system, the loading force line angles and the load values of the loading force points at each loading stage, obtain the load resultant matrix during the test; According to the constraint point coordinate matrix and the load resultant matrix, obtain the reaction force calculation model considering deformation.

6. The full-scale strength test reaction force prediction method based on data fusion according to claim 5, wherein, The expression of the constraint point coordinate matrix is: where x in is the x - coordinate of the constraint point i at the n - th loading level; y in is the y - coordinate of the constraint point i at the n - th loading level; z in is the z - coordinate of the constraint point i at the n - th loading level; cosα in is the cosine value of the angle between the load direction and the x - axis direction of the constraint point i at the n - th loading level; cosβ in is the cosine value of the angle between the load direction and the y - axis direction of the constraint point i at the n - th loading level; cosγ in is the cosine value of the angle between the load direction and the z - axis direction of the constraint point i at the n - th loading level; α in is the component of the loading force line angle of the constraint point i at the n - th loading level in the x - axis direction; β in is the component of the loading force line angle of the constraint point i at the n - th loading level in the y - axis direction; γ in is the component of the loading force line angle of the constraint point i at the n - th loading level in the z - axis direction; where i = 1, 2, 3…6 is the number of the constraint point.

7. A full-scale strength test reaction force prediction method based on data fusion according to claim 6, characterized in that, The expression of the load resultant matrix is: In the formula, m represents the total number of loading force points, and the loading force points include active loading points, key points, and the total weight of the aircraft; j represents the loading point number, x jn represents the x - coordinate of the loading force point j at the n - th loading stage; y jn represents the y - coordinate of the loading force point j at the n - th loading stage; z jn represents the z - coordinate of the loading force point j at the n - th loading stage; α jn is the component of the loading force line angle of the loading force point j at the n - th loading stage in the x - axis direction; β jn is the component of the loading force line angle of the loading force point j at the n - th loading stage in the y - axis direction; γ jn is the component of the loading force line angle of the loading force point j at the n - th loading stage in the z - axis direction; F jn represents the load value of the loading force point j at the n - th loading stage; F xn represents the component of the load value of the resultant loading force at the n - th loading stage in the x - axis direction; F yn represents the component of the load value of the resultant loading force at the n - th loading stage in the y - axis direction; F zn represents the component of the load value of the resultant loading force at the n - th loading stage in the z - axis direction; M xn represents the component of the resultant moment of the resultant loading force at the n - th loading stage in the x - axis direction; M yn represents the component of the resultant moment of the resultant loading force at the n - th loading stage in the y - axis direction; M zn represents the component of the resultant moment of the resultant loading force at the n - th loading stage in the z - axis direction, where i = 1, 2, 3…m is the number of the loading force points.

8. A full-scale strength test reaction force prediction method based on data fusion according to claim 7, characterized in that The expression of the reaction force calculation model considering deformation is: A n X n = B n n = 1, 2, …, D Where, X n is the reaction force matrix of the constraint point corresponding to the nth load level; A n is the constraint point coordinate matrix; B n is the resultant load matrix during the test; n represents the nth load level; D is the total number of loading levels.

9. A method for predicting reaction forces in full-scale strength tests based on data fusion according to claim 1, characterized in that According to the loading force line angle during the preliminary test / low-load test, a prediction model of the loading force line angle is established using the least squares method.

10. A full-scale strength test reaction force prediction system based on data fusion, characterized in that, Including: A historical parameter acquisition module for acquiring historical test load data collected in the full-scale structural strength test system; A target prediction model construction module for constructing a target prediction model for predicting the actual load mean and actual load standard deviation based on the historical test load data; A reaction force calculation model construction module for constructing a reaction force calculation model considering deformation according to the coordinates of the constraint points of the statically determinate support system in the full-scale structural strength test system, the load directions of the constraint points, the coordinates of the loading force points, the loading force line angles and load values of the loading force points at each loading stage; A loading force line angle prediction module for establishing a loading force line angle prediction model according to the loading force line angle during the preliminary test / low-load test, and predicting the loading force line angles corresponding to each loading force point in the formal / high-load test based on the loading force line angle prediction model; An actual load acquisition module for predicting the actual load mean and actual load standard deviation based on the preliminary test / low-load test load data and the target prediction model, and simulating the load of the formal / high-load test using Monte Carlo simulation based on the predicted actual load mean and actual load standard deviation to obtain the actual load corresponding to the formal / high-load test; A reaction force prediction module for predicting the reaction force during the formal / high-load test based on the actual load corresponding to the formal / high-load test, the loading force line angles corresponding to each loading force point in the formal / high-load test, and using the reaction force calculation model.

Images