A method for optimizing selection of characteristic strain monitoring points based on generalized correlation coefficient method

Abstract

The application relates to a feature strain monitoring point optimization selection method based on a generalized correlation coefficient, which can be used for determining the strain sensor monitoring position of a typical force structure of an airplane and optimizing the network arrangement thereof. First, a large number of original strain monitoring points are arranged on the airplane structure to be monitored according to simulation and engineering experience, meanwhile, no less than three basic feature working conditions are determined, and the strain response of each original strain monitoring point under each basic feature working condition is obtained, and the feature strain vector thereof is obtained through normalization processing; then, the generalized correlation coefficient provided by the application is used, a group of feature strain monitoring points with the lowest comprehensive correlation is selected according to a set threshold value, and is used for arranging strain sensors, so that the network optimization of the strain sensors of the airplane structure is realized. By using the optimization selection method of the application, the experience dependency of the traditional strain monitoring point arrangement is avoided, the redundant monitoring points can be greatly reduced, and the optimization arrangement of the strain sensors of the airplane structure is realized.

Description

Technical Field

[0001] This invention relates to a method for deploying structural strain monitoring points, specifically to an optimized selection method for aircraft structural characteristic strain monitoring points based on the generalized correlation coefficient method. Background Technology

[0002] Aircraft structural integrity is crucial throughout the entire process of aircraft structural design, manufacturing, and use. Structural life monitoring is a vital component, ensuring that each aircraft can fully realize its lifespan potential while maintaining safe flight. To monitor aircraft structural lifespan, sensors need to be deployed on the aircraft structure to form a corresponding sensor monitoring system. Due to limitations imposed by aircraft structural weight and economic factors, the number of sensors cannot be excessive. Therefore, selecting efficient sensor placement locations, reducing the number of sensors, and optimizing the sensor network become extremely important.

[0003] Li Peng et al. proposed a sensor distribution optimization method for structural health monitoring under multiple operating conditions. The aim is to improve the efficiency of identifying structural anomalies under these conditions by optimizing sensor distribution. The method first obtains structural state data under normal operating conditions through simulation analysis, then calculates the hypersphere clustering index using a support vector data description algorithm. Finally, using both the number of sensors and the hypersphere clustering index as optimization objective functions, an improved non-dominated hierarchical genetic algorithm is used to obtain the non-dominated solution set of this dual-objective optimization problem, providing an optimized sensor distribution scheme for identifying structural anomalies under multiple operating conditions [Chinese Patent Publication No.: 108830407A].

[0004] Luo Xudong et al. proposed a method for determining the fatigue life index of hydraulic conduits. This method uses simulation to predict the natural frequencies, mode shapes, and stress distribution cloud maps of the conduit structure under actual installation conditions, identifying the monitoring points most prone to fatigue failure. The test monitoring points are the points most susceptible to fatigue failure determined by simulation. Strain gauges and acceleration sensors are installed at these monitoring points, and the fatigue life curves of the test specimen are fitted based on stress and cycle number, as well as acceleration and cycle number [Chinese Patent Publication No.: 105651496A].

[0005] Yao Yanfei et al. proposed an aircraft fault monitoring robot group and its path optimization method, including the following steps: locating the position of the monitoring robot group; setting the monitoring point position; locating the monitoring robots near each monitoring point with R as the radius; selecting the monitoring robot with the shortest straight-line distance and obtaining the path scheme between the monitoring robot and the monitoring point; retaining the path scheme and obtaining a new path scheme again; substituting the movement speed of the monitoring robot and analyzing and comparing the arrival time of each path scheme; selecting the path scheme with the shortest arrival time and dispatching the monitoring robot; the monitoring robot arrives at the monitoring point and monitors the monitoring point through a monitoring camera and a thermal imager [Chinese Patent Publication No.: 112180917A].

[0006] Zhang Ying et al. proposed a differential coverage method based on a hybrid sensor network. This invention is based on the particle swarm optimization (PSO) method, using a single objective function formed by a weighted linear combination of the effective coverage rates of key and general areas as the fitness function. The fitness function is controlled by adjusting the weighting coefficients, thereby guiding mobile sensor nodes to converge towards key areas and ensuring coverage quality in those areas. To improve the computational efficiency of the PSO method, a virtual force velocity component is added to guide mobile nodes to move towards uncovered areas. The magnitude and direction of the virtual force are obtained based on the gravitational field applied to the mobile nodes by monitoring points in the deployment area that do not meet the effective monitoring threshold. This invention is applicable to hybrid sensor networks containing both fixed and mobile nodes [Chinese Patent Publication No.: 103997748B].

[0007] The methods mentioned above can effectively solve some sensor placement problems, but they cannot solve the problem of optimizing the placement of strain sensors on critical load-bearing structures of aircraft. To obtain the flight load on an aircraft structure, it is often necessary to monitor the strain response of the aircraft structure in real time, and then use this strain response to obtain information such as the aircraft's health status. Traditionally, strain sensors on aircraft structures are often placed along the force transmission path. However, the monitoring data from different strain sensors arranged in this way have high correlation, easily leading to a large amount of information redundancy. Therefore, it is necessary to further optimize the strain sensor network layout to reduce redundant information between different strain monitoring points, so as to achieve high-precision acquisition of structural loads using fewer strain sensors. Summary of the Invention

[0008] The purpose of this invention is to provide a method for optimizing the selection of characteristic strain monitoring points based on the generalized correlation coefficient method. This method can optimize the selection of a set of characteristic strain monitoring points with the lowest overall correlation to the load response and the best characterization of the load on the structure from a large number of original strain monitoring points of the aircraft structure. This set of characteristic strain monitoring points can be used for the optimized design of strain sensor network, which greatly reduces redundant monitoring points and thus optimizes the number and location of sensors in the strain sensor monitoring network.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A method for optimizing the selection of characteristic strain monitoring points based on generalized correlation coefficient includes the following steps:

[0011] Step 1: First, based on simulation, real test or engineering experience, place p initial monitoring points on the structure to be monitored on the aircraft. If q monitoring points are ultimately needed, then p needs to be more than 10 orders of magnitude larger than q.

[0012] Step 2: Use finite element simulation software to conduct simulation tests on the aircraft structure to be monitored, determine the basic characteristic working conditions, apply the basic characteristic working conditions to the aircraft structure to be monitored, and obtain the strain values ​​of each original monitoring point under each basic characteristic working condition.

[0013] Step 3: Normalize the strain values ​​of each original monitoring point to obtain a strain vector for each original monitoring point;

[0014] Step 4: Take the original monitoring point with the largest strain value under each basic characteristic working condition as the first characteristic strain monitoring point under that working condition;

[0015] Step 5: Calculate the generalized correlation coefficient between the remaining monitoring points and the selected characteristic strain monitoring points, and use the minimum value of the generalized correlation coefficient as the newly selected characteristic strain monitoring point;

[0016] Step 6: When the Pearson correlation coefficient between the newly selected characteristic strain monitoring point and the previously selected characteristic strain monitoring point exceeds a given threshold, discard the newly selected characteristic strain monitoring point and stop selecting characteristic strain monitoring points under the corresponding working condition.

[0017] Step 7: Merge the characteristic strain monitoring points under different basic characteristic working conditions to obtain the final structural characteristic strain monitoring points.

[0018] Preferably, the number of basic characteristic conditions of the aircraft structure to be monitored is not less than three.

[0019] The normalization process involves setting the maximum strain value among all original strain monitoring points under each basic characteristic working condition to a multiple of 100, preferably 1000 microstrain, and scaling down the strain values ​​of the remaining original strain monitoring points under this working condition proportionally.

[0020] The dimension of the strain vector constructed for each original monitoring point is the same as the number of basic characteristic working conditions, and the vector elements are the normalized results of the strain response of the monitoring point under the basic characteristic working conditions.

[0021] A first characteristic strain monitoring point is set for each basic characteristic working condition.

[0022] The generalized correlation coefficient is the average of the squares of the Pearson correlation coefficients between the strain vectors of the remaining monitoring points and the strain vectors of all the previously selected characteristic strain monitoring points.

[0023] A threshold s is used to control the number of characteristic strain monitoring points selected through optimization. The threshold s is set according to the complexity of the structure, and is preferably set to 0.98.

[0024] Compared with the prior art, the present invention has the following advantages:

[0025] (1) The method of the present invention monitors the strain data on the aircraft structure. The strain response of the aircraft structure is directly caused by the load on the aircraft and has a more direct mapping relationship.

[0026] (2) The overall correlation between the strain monitoring points selected by the method of the present invention is very small, which can effectively reduce a large amount of redundant data. The number of sensors required under the same structural load inversion accuracy is much smaller than that of the traditional arrangement method. Attached Figure Description

[0027] Figure 1 This is a flowchart of the optimized selection process for characteristic strain monitoring points based on the generalized correlation coefficient.

[0028] Figure 2 This is a distribution map of the original monitoring points on the wing box section model.

[0029] Figure 3 This is a distribution map (small squares) of characteristic strain monitoring points on the wing box section.

[0030] Figure 4 This is a map showing the original monitoring point distribution on the wing-shaped aluminum alloy plate.

[0031] Figure 5 This is a distribution diagram (circles) of characteristic strain monitoring points on an aluminum alloy plate shaped like an airfoil. Detailed Implementation

[0032] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0033] According to the Pearson correlation coefficient:

[0034]

[0035] In the formula, Cov(X,Y) is the covariance of variables X and Y; Var[X] is the variance of variable X; Var[Y] is the variance of variable Y.

[0036] While the Pearson correlation coefficient can effectively characterize the correlation between two variables, it cannot characterize the correlation between a variable and multiple variables. To characterize the comprehensive correlation between a variable and multiple variables, this invention proposes a generalized correlation coefficient R as follows:

[0037]

[0038] In the formula, R is the generalized correlation coefficient; r1, r2, ... r k The Pearson correlation coefficients between the specified variable and variables 1, 2, ..., k, where k is the number of variables in the series.

[0039] The generalized correlation coefficient is the average of the squared Pearson correlation coefficients between a variable and multiple variables, and it can characterize the comprehensive correlation between a variable and multiple variables. Based on the definition of the generalized correlation coefficient, this invention proposes a method for optimizing the selection of strain monitoring points for aircraft structures.

[0040] like Figure 1 As shown, firstly, a large number of original strain monitoring points are arranged on the aircraft structure to be monitored based on simulation experiments, real experiments, or experience. Then, finite element simulation software is used to conduct simulation experiments on the aircraft structure, applying n (n≥3) basic characteristic conditions. The strain values ​​of each original strain monitoring point under the n basic characteristic conditions are calculated, and each original strain monitoring point corresponds to n strain values. The strain values ​​obtained under the n basic characteristic conditions are then normalized. After normalization, each original strain monitoring point can obtain a characteristic strain vector with n elements. Finally, the original strain monitoring point with the largest strain under the first basic characteristic condition is taken as the first characteristic strain monitoring point under that condition, and the strain values ​​are calculated separately. Calculate the generalized correlation coefficient between the strain vector of the remaining original strain monitoring points and the strain vector of the first characteristic monitoring point under this operating condition. The original strain monitoring point with the smallest generalized correlation coefficient is selected as the second characteristic strain monitoring point under this operating condition. Then, calculate the generalized correlation coefficient between the strain vector of the remaining original strain monitoring points and the strain vectors of the determined first and second characteristic strain monitoring points respectively. The original strain monitoring point with the smallest generalized correlation coefficient is selected as the undetermined monitoring point. It is then determined whether the Pearson correlation coefficient between the strain vector of this monitoring point and the strain vectors of the previous two characteristic strain monitoring points is less than a specified threshold s. If both are less than the threshold s, the undetermined monitoring point is selected as the third characteristic strain monitoring point under this basic operating condition. This process is repeated until the Pearson correlation coefficient between the strain vector of the undetermined monitoring point and the strain vector of a selected characteristic strain monitoring point exceeds the specified threshold s. The selection of characteristic strain monitoring points under this operating condition is then stopped. The same procedure is then used to select characteristic strain monitoring points under other basic characteristic operating conditions. Finally, the characteristic strain monitoring points under different basic characteristic operating conditions are merged to obtain all characteristic strain monitoring points for the aircraft structure.

[0041] The method is applicable to the optimization selection of strain monitoring points for aircraft structures.

[0042] The number of original strain monitoring points is much greater than the number of characteristic strain monitoring points.

[0043] The number of the basic characteristic operating conditions shall not be less than three.

[0044] The dimension of the strain vector constructed from each original strain monitoring point is the same as the number of basic characteristic working conditions, and the vector elements are the normalized results of the strain response of the monitoring point under the basic characteristic working conditions.

[0045] The normalization process involves setting the maximum strain value among all original strain monitoring points under each basic characteristic working condition to a multiple of 100, preferably 1000 microstrain, and scaling down the strain values ​​of the remaining original strain monitoring points under this working condition proportionally.

[0046] Each of the basic characteristic operating conditions has a first characteristic strain monitoring point.

[0047] The generalized correlation coefficient is the average of the squares of the Pearson correlation coefficients between the strain vector of the specified monitoring point and the strain vectors of all previously selected characteristic strain monitoring points.

[0048] The threshold s can be changed according to the actual situation, and is generally preferred to be 0.98. Specific Implementation Example 1

[0050] This example, taken with reference to the accompanying drawings and using the wing box section as an example, provides a more detailed description of the invention through specific embodiments.

[0051] Step 1: Determine the initial strain monitoring points on the wing box section:

[0052] The wing box segment model was imported into ANSYS Workbench for simulation. Common operating conditions were applied to the wing box segment, and the strain contour distribution of the wing box segment was obtained, such as... Figure 2 As shown, and based on engineering experience, original strain monitoring points were set up in areas with large strain distribution. Sixteen monitoring zones were set up, with 40 monitoring points evenly distributed in each monitoring zone, thus obtaining a total of 640 original strain monitoring points.

[0053] Step 2: Determine the basic characteristic working conditions:

[0054] The selected basic characteristic load cases need to best reflect most actual load conditions. Aircraft structural load cases are generally combinations of multiple loads, typically combinations of bending moment, shear force, and torque. The most basic characteristic load cases can be considered as pure bending moment loading, pure shear force loading, and pure torque loading. These three basic characteristic load cases can best "characterize" the load characteristics of all conditions.

[0055] Step 3: Construct the feature strain vector:

[0056] Simulation tests were conducted on the wing box section using ANSYS Workbench, applying pure bending moment, pure shear force, and pure torque conditions respectively. Each original strain monitoring point yielded three strain values. The strain vector elements of the original strain monitoring point with the maximum strain response under the three basic characteristic conditions were set to 1000. The strain responses of the other original strain monitoring points under this condition were scaled proportionally, resulting in a characteristic strain vector with three elements for each original strain monitoring point.

[0057] Step 4: Optimization of characteristic strain monitoring points based on generalized correlation coefficient:

[0058] First, the original strain monitoring point with the largest equivalent strain value under the pure bending moment loading condition is selected as the first characteristic strain monitoring point under this condition. Second, the generalized correlation coefficient between the strain vectors of the remaining original strain monitoring points and the strain vector of the first characteristic strain monitoring point under the determined pure bending moment loading condition is calculated. The original strain monitoring point corresponding to the minimum value of the generalized correlation coefficient is selected as the second characteristic strain monitoring point under this basic condition. Third, the generalized correlation coefficient between the strain vectors of the remaining original strain monitoring points and the strain vectors of the determined first and second characteristic strain monitoring points is calculated respectively. The monitoring point corresponding to the minimum value of the generalized correlation coefficient is selected as the undetermined monitoring point. Then, it is determined whether the Pearson correlation coefficient between the strain vector of the undetermined monitoring point and the strain vectors of the first and second characteristic strain monitoring points is less than the threshold s (here, s is taken as 0.98). If they are all less than the threshold s, then the undetermined monitoring point can be selected as the third characteristic strain monitoring point under this basic condition. Similarly, the remaining characteristic strain monitoring points are selected until the Pearson correlation coefficient between the strain vector of the undetermined monitoring point and the strain vector of any determined characteristic strain monitoring point exceeds the threshold s. Then, the selection of characteristic strain monitoring points under this working condition is stopped. The characteristic strain monitoring points under the pure bending moment working condition are monitoring points No. 4, 118, 236, 265, 277, 340 and 533, respectively.

[0059] The selection process for characteristic strain monitoring points under pure shear force and pure torque loading conditions is the same as that under pure bending moment loading conditions. The characteristic strain monitoring points under pure shear force conditions are monitoring points 4, 118, 236, 265, 277, 340, and 533. The characteristic strain monitoring points under pure torque conditions are monitoring points 66, 120, 236, 326, 500, 533, and 572.

[0060] Finally, the characteristic strain monitoring points under the three operating conditions were merged to obtain all the characteristic strain monitoring points on the wing box section, such as... Figure 3 As shown, these are monitoring points numbered 4, 66, 118, 120, 236, 265, 277, 326, 340, 500, 533, and 572, respectively. Specific Implementation Example 2

[0062] This example, taken with reference to the accompanying drawings and using an aluminum alloy plate in the shape of an airfoil as an example, provides a more detailed description of the present invention through specific embodiments.

[0063] Step 1: Determine the initial strain monitoring points on the aluminum alloy of the wing shape:

[0064] The wing-shaped aluminum alloy plate model was imported into ANSYS Workbench for simulation. A concentrated force of the same magnitude was applied to different locations on the wing-shaped aluminum alloy plate. Figure 4 As shown, the strain cloud map distribution on the wing-shaped aluminum alloy plate was obtained. Original strain monitoring points were arranged in areas with large strain distribution. Five monitoring zones were set up, and several monitoring points were evenly arranged in each monitoring zone, resulting in a total of 129 original strain monitoring points.

[0065] Step 2: Determine the characteristic operating conditions:

[0066] The selected characteristic load cases need to best reflect the concentrated force loading conditions of most aluminum alloy plates. Load case 11 is the load case with the smallest x-coordinate and the largest y-coordinate among all loading points; load case 43 is a load case with both large x-coordinates and y-coordinates and located near an inflection point; and load case 57 is a load case with the largest x-coordinate and the smallest y-coordinate among all loading points. These three basic characteristic load cases can best "characterize" the load-bearing characteristics of all load cases.

[0067] Step 3: Construct the feature strain vector:

[0068] Simulation tests were conducted on aluminum alloy plates using ANSYS Workbench, applying three characteristic working conditions. Each original strain monitoring point yielded three strain values. The strain vector element of the original strain monitoring point with the maximum strain response under the three basic characteristic working conditions was set to 1000. The strain responses of the other original strain monitoring points under this condition were scaled proportionally, resulting in a characteristic strain vector with three elements for each original strain monitoring point.

[0069] Step 4: Optimization of characteristic strain monitoring points based on generalized correlation coefficient:

[0070] First, the original strain monitoring point with the largest equivalent strain value under operating condition 11 is selected as the first characteristic strain monitoring point under this operating condition. Second, the generalized correlation coefficient between the strain vectors of the remaining original strain monitoring points and the strain vector of the first characteristic strain monitoring point under the determined basic operating condition is calculated. The original strain monitoring point corresponding to the minimum value of the generalized correlation coefficient is selected as the second characteristic strain monitoring point under this basic operating condition. Third, the generalized correlation coefficient between the strain vectors of the remaining original strain monitoring points and the strain vectors of the first and second characteristic strain monitoring points is calculated respectively. The monitoring point corresponding to the minimum value of the generalized correlation coefficient is selected as the undetermined monitoring point. Then, it is determined whether the Pearson correlation coefficient between the strain vector of the undetermined monitoring point and the strain vectors of the first and second characteristic strain monitoring points is less than the threshold s (here, s is taken as 0.98). If they are all less than the threshold s, then the undetermined monitoring point can be selected as the third characteristic strain monitoring point under this basic operating condition. This process continues until the Pearson correlation coefficient between the strain vector of the undetermined monitoring point and the strain vector of any of the already determined characteristic strain monitoring points exceeds a threshold s. At this point, the selection of characteristic strain monitoring points for that operating condition is stopped. Figure 5 As shown, the characteristic strain monitoring points under working condition 11 are monitoring points No. 28, 31, 50, 61, 87 and 121.

[0071] The selection process for characteristic strain monitoring points under operating conditions 43 and 57 is the same as that under operating condition 11; the characteristic strain monitoring points under operating condition 43 are monitoring points 28, 50, 61 and 129 respectively. The characteristic strain monitoring points under operating condition 57 are monitoring points 28, 37, 50, 61, 87 and 121 respectively.

[0072] Finally, the characteristic strain monitoring points under the three working conditions were merged to obtain all the characteristic strain monitoring points on the aluminum alloy plate of the wing shape, namely monitoring points No. 28, 31, 37, 50, 61, 87, 121 and 129.

Claims

1. A method for optimizing selection of characteristic strain monitoring points based on a generalized correlation coefficient, characterized in that, The method comprises the following steps: First step: first, arrange p original monitoring points on the aircraft structure to be monitored according to simulation or real test or engineering experience, and suppose that q monitoring points are finally needed, then p should be more than 10 orders of magnitude than q; Second step: use finite element simulation software to simulate the aircraft structure to be monitored, determine the basic characteristic working conditions, and apply the basic characteristic working conditions to the aircraft structure to be monitored to obtain the strain values of each original monitoring point under each basic characteristic working condition; Third step: normalize the strain values of each original monitoring point, and obtain a strain vector for each original monitoring point; Fourth step: take the original monitoring point with the maximum strain value under each basic characteristic working condition as the first characteristic strain monitoring point under the working condition; Fifth step: calculate the generalized correlation coefficient between the remaining monitoring points and the selected characteristic strain monitoring points, and take the minimum value of the generalized correlation coefficient as the newly selected characteristic strain monitoring point; Sixth step: when the Pearson correlation coefficient between the newly selected characteristic strain monitoring point and the selected characteristic strain monitoring point exceeds a given threshold value, discard the newly selected characteristic strain monitoring point, and stop selecting the characteristic strain monitoring point under the corresponding working condition; Seventh step: combine the characteristic strain monitoring points under different basic characteristic working conditions to obtain the final structural characteristic strain monitoring points. 2.The method of claim 1, wherein: The number of basic characteristic working conditions of the aircraft structure to be monitored is not less than 3.

3. The method of claim 1, wherein the method is characterized by: The normalization processing mode is to set the maximum strain value of all original strain monitoring points under each basic characteristic working condition as a multiple of 100, and scale the strain values of the remaining original strain monitoring points under the working condition in proportion.

4. The method of claim 1, wherein the method is characterized by: The dimension of the strain vector constructed for each original monitoring point is the same as the number of basic characteristic working conditions, and the vector elements are the normalized processing results of the strain responses of the monitoring point under the basic characteristic working conditions.

5. The method of claim 1, wherein the method is characterized by: A first characteristic strain monitoring point is set under each basic characteristic working condition.

6. The method of claim 1, wherein the method is characterized by: The generalized correlation coefficient is the average value of the square of the Pearson correlation coefficient between the strain vector of the remaining monitoring point and the strain vector of all previously selected characteristic strain monitoring points.

7. The method of claim 1, wherein the method is characterized by: The threshold value s is used to control the number of characteristic strain monitoring points obtained by optimization selection, and the threshold value s is set according to the complexity of the structure.

8. The method of claim 7, wherein the method is characterized by: The threshold value s is set to 0.98.

Images