A poor sample sensor data expansion method based on mechanical consistency mechanism

Abstract

A poor sample sensor data expansion method based on mechanical consistency mechanism belongs to the technical field of digital twinning. First, in the simulation analysis stage, a real structure parameterized simulation model is established, and a real structure simulation response data set is obtained; second, in the ground test design stage, an equivalent ground test model and a loading scheme are designed; third, in the ground test implementation stage, ground test measuring points are arranged, and a ground test data set is obtained; fourth, in the on-orbit and flight stage, real-time on-orbit and flight measured data are collected, and a measured sensor data set is obtained; fifth, in the data expansion stage, the ground test data set is mapped to the on-orbit and flight sensor data, and an expanded sensor data set is obtained; finally, correlation evaluation and adjustment are carried out on the expanded sensor data. The application can solve the problem of insufficient equipment structure monitoring capability caused by sparse structure sensor number and differences between space and ground states, and realize data expansion for digital twinning high-precision modeling.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of digital twinning, and relates to a poor sample sensor data expansion method based on a mechanical consistency mechanism. BACKGROUND

[0002] The aerospace equipment structures such as space stations and manned spacecraft need to serve for a long time under complex and changeable load environments, and the reliability requirements for structural strength are very high. In order to ensure the normal service of the aerospace equipment structures, the structural mechanical responses need to be monitored in real time, which will provide an important basis for accurately evaluating the service state of the aerospace equipment structures.

[0003] In the process of structural digital twinning monitoring in an on-orbit or flight state, due to the large-scale and complex of the equipment structures, the cost and weight of sensors and other factors, the structural sensors are sparse, the measured data sample size is low, and then the monitoring accuracy of the mechanical responses is affected. At the same time, due to the limitations of model size, loading implementation conditions and other factors, there is a large difference between the ground test and the service state, so that the ground test cannot directly reflect the real structural mechanical responses.

[0004] In view of the above problems, scholars have carried out a lot of research and proposed some solutions. For example, a Chinese invention patent (application number 202510276719.6) provides a method and device for evaluating the consistency of airplane ground test and flight test, fuses the finite element simulation data with the ground test data and the flight test data, evaluates the consistency of the ground data and the flight data by calculating the strain data difference of the same measuring points, but it only uses the interpolation method to directly evaluate and does not propose a technical route considering the difference between the ground and flight mechanical models. A Chinese invention patent (application number 202411391510.6) provides an online ground and flight data fusion method based on a physical information neural network, uses part of the dimensions of the ground test data and flight test data as feature inputs of the physical information neural network model, uses other dimensions of the flight test data as labels, and obtains a real-time monitoring model with high accuracy by using a small amount of high-precision flight data and a large amount of low-precision ground test data, but the training process needs to be based on the consistency of the test and flight conditions. That is, although the existing technical solutions mainly evaluate and correct the errors between the data under the premise that the ground and flight conditions are basically consistent, they are difficult to be directly applied to the case where the structural size or load form is significantly different.

[0005] Therefore, a poor sample sensor data expansion method based on a mechanical consistency mechanism is proposed to solve the problems of sparse structural sensors and significant differences between the ground and flight states, and to realize data expansion and provide sensor data with higher coverage for key positions for digital twinning high-precision modeling. SUMMARY

[0006] The application is to solve the problem of insufficient equipment structure monitoring capability caused by sparse structure sensor number and differences between earth and space states in the prior art, and provides a poor sample sensor data expansion method based on mechanical consistency mechanism.

[0007] In order to achieve the above-mentioned purpose, the technical scheme adopted by the application is:

[0008] A poor sample sensor data expansion method based on mechanical consistency mechanism, the poor sample sensor data expansion method comprises the following steps:

[0009] Step 1: simulation analysis stage, establishing a real structure parameterized simulation model, carrying out simulation analysis, obtaining a real structure simulation response data set, specifically:

[0010] Firstly, a parameterized simulation model is established, and the model parameters include geometric feature parameters and load parameters; then, the real structure geometric features and the real structure use load F are input into the parameterized simulation model as parameters, and the simulation analysis is run to obtain a real structure simulation response data set I f =[X f ,Y f ,Z f ,R f ], wherein X f ,Y f ,Z f are real structure simulation model coordinates, and R f is a real structure simulation response.

[0011] Step 2: ground test design stage, considering the consistency of the mechanical mechanism of the earth and the space, according to the model size and load state of the real structure and the implementable test, based on the similarity scaling criterion and the parameter optimization method, an equivalent ground test model and a loading scheme are designed. Specifically:

[0012] Step 2.1, for the case that the real structure size is too large and the full-size ground test piece is difficult to process, a scaled test model needs to be designed as a ground test model. First, according to the model size of the implementable test, the geometric similarity constant C L =L real / L scaled is determined, and based on the similarity theory, the model is scaled as a whole, wherein L real represents the real structure size, and L scaled represents the scaled test model size; then, based on the parameter optimization method, the parameters of the scaled test model are adjusted according to the real structure simulation response R f to obtain a scaled test model and a test use load F d. For the case that a full-scale test model can be established, i.e. the geometry and material properties of the test model are consistent with the real structure, the full-scale test model is taken as the ground test model, and F d =F, where F is the real structure service load.

[0013] Further, the design variables of the parameter optimization of the scaled test model include the geometric parameters and the load size of the scaled test model. The design variables are input into the parameterized simulation model of step 1 in each iteration to obtain a scaled test simulation model. The optimization objective is to maximize the correlation between the simulation response of the scaled test model and the simulation response R f of the real structure, and the optimization constraints include the processing feasibility and the sensor arrangement feasibility.

[0014] Step 2.2, considering the ground test loading condition, if the test load F d can be used in step 2.1, the test loading scheme is consistent with F d ; if the loading condition is limited, the test load F d is replaced by N easily implemented loading states or loading levels F i in the linear elastic range of the material:

[0015]

[0016] where subscript i represents the serial number of each loading state or loading level, a is the corresponding coefficient, i = 1, 2, 3, …, N; that is, the actual load distribution is simulated by F i superposition to obtain an equivalent test loading scheme.

[0017] Step three: ground test implementation phase, based on the real structure sensor points and the key areas of the structure, the ground test measuring points are arranged, and the ground test model coordinates are converted into real structure coordinates based on the geometric similarity constant to obtain the ground test data set. Specifically:

[0018] Step 3.1, first, the test measuring point set S1 is arranged at the same positions of the ground test model and the real structure measuring points, which is used to evaluate the correlation between the ground and space data;

[0019] Step 3.2, then, the test measuring point set S2 is arranged in the key areas that are not covered by the real structure measuring points, which is used to provide more mechanical response information. The key areas can be obtained in combination with engineering experience and the simulation response of the real structure.

[0020] Step 3.3, finally, the test is carried out under the loading scheme established in step two, and the response data of each measuring point is recorded. If the full-scale test model is used in step two, the ground test model coordinates and the real structure coordinates X t , Y t , and Z tare the same; if the scale-down test model is used in step two, a scale-down test coordinate system needs to be established on the scale-down test model, the origin position and coordinate axis direction of which are the same as those of the real structure, and the coordinates X ts ,Y ts ,Z ts of the measuring points of the scale-down test model are multiplied by the geometric similarity constant C L of step 2.1 to be transformed into the coordinates X t ,Y t ,Z t of the real structure; and then the ground test data set I t =[X t ,Y t ,Z t ,F i ,R i ] is obtained. Wherein F i is the loading state or loading level, R i is the corresponding test response data, and the subscript i represents the serial number of each loading state or loading level, i = 1, 2, 3, …, N.

[0021] Step four: in the on-orbit and flight phase, real-time collection of the on-orbit and flight measured data of the real structure is performed to obtain a measured sensor data set. Specifically:

[0022] The coordinate information of each sensor of the real structure, real-time response data and real-time load size are collected, and abnormal value detection is performed to obtain the measured sensor data set I a =[X a ,Y a ,Z a ,R a ,F a ], wherein X a , Y a , Z a are the sensor coordinate values, R a is the measured response data, and F a is the measured load value.

[0023] Further, the abnormal value detection method includes the Lévy criterion, threshold setting or the isolated forest algorithm.

[0024] Step five: data expansion phase, based on the linear superposition principle and the model coordinate mapping method, the ground test data set obtained in step three is mapped to the on-orbit and flight sensor data to obtain an expanded sensor data set. Specifically:

[0025] Step 5.1, if the test loading scheme in step two is the same as the test load used, the test response value is directly taken as the response of the real structure under the on-orbit and flight working condition If step two adopts an equivalent test loading scheme, then based on the principle of linear superposition, the actual structure's response under on-orbit and flight conditions will be... It can be represented as the response R under each experimental loading state or loading level. i The linear combination of the load F used in step two of the experiment d Represented as F i The linear combination forms are the same:

[0026]

[0027] If the actual structure uses load F and the measured load value F in step four... a If a deviation exists, it should be corrected according to the load ratio k. This leads to the acquisition of a ground-based equivalent response dataset of the actual structure under on-orbit and flight conditions. The load ratio k is defined as follows:

[0028]

[0029] in, This represents the measured load value. This indicates that the actual structure uses loads. It represents the vector norm or the absolute value of a scalar.

[0030] Step 5.2: Based on the consistency between the experimental model and the actual structural coordinates after coordinate transformation in Step 3, the experimental measurement points can be mapped to on-orbit and flight measurement points, and the equivalent response dataset I from the ground test can be generated. e Incorporate measured sensor dataset I a The expanded sensor dataset is obtained. .

[0031] Step Six: Data Verification Phase. Correlation assessment and adjustment are performed on the expanded sensor data. Specifically:

[0032] Based on step 5.1, the equivalent response dataset I from the ground test. e Response data R of ground test measurement point set S1 c , and the measured response data R a Correlation analysis was conducted, and the evaluation indicators included R. 2 , RRMSE. If the correlation evaluation results do not meet the requirements, data adjustment is carried out, including: correcting and re-conducting ground tests, removing outliers using numerical simulation, and correcting ground test responses based on spatial interpolation. The final result is the expanded and corrected sensor dataset I. final This enables data expansion for high-precision modeling of digital twins.

[0033] Furthermore, the correlation evaluation result not meeting the requirements means that: R 2If the RRMSE is below a set threshold, which is a dimensionless number less than 1; or if the RRMSE is above a set threshold, which is a dimensionless number greater than 0.

[0034] The beneficial effects of this invention are as follows:

[0035] (1) This invention is aimed at monitoring the mechanical response of equipment structure. By designing an equivalent ground test model and loading scheme, it solves the problem of differences between the ground and ground states caused by limiting factors such as model size and loading implementation conditions, and realizes the ground test design based on the mechanical consistency mechanism.

[0036] (2) Based on the principle of linear superposition and the model coordinate mapping method, the ground test dataset is mapped to on-orbit and flight sensor data, thereby expanding the ground test data to the real structure. This further solves the problem of insufficient equipment structure monitoring capability caused by the sparse number of sensors in the real structure, and realizes data expansion for high-precision modeling of digital twins. Attached Figure Description

[0037] Figure 1 Flowchart for implementing a data expansion method for poor-sample sensors based on the mechanical consistency mechanism;

[0038] Figure 2 A schematic diagram of the geometric features of the actual structure of a spacecraft segment and a scaled-down experimental model; Figure 2 (a) in the model is the full-size model; Figure 2 (b) in the model is a scaled-down version;

[0039] Figure 3 This is a schematic diagram illustrating the principle of ground test data mapping.

[0040] Figure 4 A schematic diagram of sensor measurement point expansion;

[0041] Figure 5 A schematic diagram showing the stress field monitoring results of a spacecraft module before and after data expansion, along with an error comparison diagram; Figure 5 (a) in the image is the cloud map before data expansion; Figure 5 (b) in the image is the cloud map after data expansion; Figure 5 (c) in the figure represents the error comparison. Detailed Implementation

[0042] To make the problem solved by the present invention, the method adopted, and the effect achieved by the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present invention are shown in the accompanying drawings, not all of them.

[0043] Figure 1 This invention provides a flowchart of a method for expanding low-sample sensor data based on a mechanical consistency mechanism, as illustrated in this embodiment. The method includes:

[0044] Step 1: Simulation Analysis Phase. Establish a parameterized simulation model of the real structure, conduct simulation analysis, and obtain a dataset of the real structure's simulation response. Specifically:

[0045] First, a parametric simulation model is established. The actual structure is the spacecraft's on-orbit module, made of aluminum alloy with an elastic modulus of 70 GPa and a Poisson's ratio of 0.33. The finite element method is used, with a mesh size of approximately 120,000. Model parameters include geometric feature parameters and load parameters. Geometric feature parameters include the total length of the module, the maximum diameter of the module, the skin thickness, the rib spacing and dimensions. The load distribution of the on-orbit module is uniform internal pressure, and the load parameter is the pressure magnitude. Then, the geometric features and loads of the on-orbit module are input as parameters into the parametric simulation model, and the simulation analysis is run. The actual structural dimensions are defined as the total length L of the on-orbit module. real =10m, maximum diameter is 7m, the load used in the on-orbit section is a uniformly distributed internal air pressure F=100kPa, such as Figure 2 As shown in (a) above, the on-orbit module simulation response dataset I was obtained. f =[X f ,Y f Z f ,R f ], where X f ,Y f Z f For the coordinates of the on-orbit module simulation model, R f This is a simulation response for the on-orbit module.

[0046] Step Two: Ground Test Design Phase. Considering the consistency of terrestrial and terrestrial mechanical mechanisms, and based on the actual structure and the dimensions and load conditions of the testable model, an equivalent ground test model and loading scheme are designed using similarity scaling criteria and parameter optimization methods. Specifically:

[0047] Step 2.1: Due to the excessively large size of the on-orbit module, a full-size ground test specimen is difficult to manufacture; therefore, a scaled-down test model needs to be designed as the ground test model. First, based on the model size L to be tested... scaled =2.5m, determine the geometric similarity constant C L =L real / L scaled =4, based on similarity theory, the overall model is scaled down, and then based on parameter optimization methods, the simulation response R of the on-orbit module is calculated. fAdjust the parameters of the scaled-down test model to obtain a scaled-down test model that meets the test conditions. The design results are shown in (b) of Figure 2 , and the test load F d = 778 kPa.

[0048] Further, the parameter optimization design variables of the scaled-down test model include the skin thickness of the scaled-down test model, the strip spacing, the strip size, and the load size, and the load distribution form is the same as that of the on-orbit cabin section (uniform internal air pressure). In each iteration, the design variables are input into the parameterized simulation model of step one to obtain the scaled-down test simulation model. The optimization objective is to maximize the correlation between the von-Mises stress simulation response of the scaled-down test model and the von-Mises stress simulation response R f of the on-orbit cabin section. The optimization constraints include the processing feasibility of the cabin thin-walled stiffened structure and the sensor arrangement feasibility. A genetic algorithm optimizer is used. The design results of the scaled-down test piece and the on-orbit cabin section at corresponding points have a von-Mises stress simulation response R 2 = 0.98.

[0049] Step 2.2, considering the ground test loading conditions, since it is not possible to directly implement air pressure loading, a water pressure step loading scheme is used instead. Ten loading levels are set, with each level having an incremental load of 40 kPa. The test load F d is replaced by 10 water pressure loading levels F i :

[0050]

[0051] where subscript i represents the serial number of each water pressure loading level, a is the corresponding coefficient, and i = 1, 2, 3, …, 10. By superimposing F i , the uniform internal air pressure load distribution of the on-orbit cabin section is simulated, and an equivalent test loading scheme is obtained.

[0052] Step three: ground test implementation phase, based on the real structure sensor points and the key areas of the structure, arrange the ground test measuring points, and based on the geometric similarity constant, convert the ground test model coordinates to the real structure coordinates to obtain the ground test data set. Specifically:

[0053] Step 3.1, first, arrange the test measuring point set S1 at the same positions as the on-orbit cabin section measuring points on the ground test model, which is used to evaluate the correlation between the terrestrial and space data. The on-orbit cabin section has 8 strain measuring points, each using a three-piece right-angle strain gauge, with a total of 24 strain acquisition channels. The test measuring point set S1 is arranged at the same positions as the on-orbit cabin section measuring points on the ground test model, containing 24 channels (8 strain measuring points).

[0054] Step 3.2, then, in the key areas of the on-orbit cabin segment measurement point is not covered, arrange test measurement point set S2, for providing more mechanical response information. According to the engineering experience, on-orbit cabin segment simulation data to determine the key area, including the cabin skin, weld area, arrange test measurement point set S2, which contains 75 strain measurement points, each strain measurement point adopts three pieces of right angle strain gauge, a total of 225 strain acquisition channel data.

[0055] Step 3.3, finally, under the water pressure loading level established in step two, respectively, carry out the test, record the response data of each measurement point. Because step two adopts the scaled-down test model, it is necessary to establish the scaled-down test coordinate system on the scaled-down test model, the origin position, coordinate axis direction and on-orbit cabin segment are the same, and the measurement point coordinates X ts ,Y ts ,Z ts of the scaled-down test model are multiplied by the geometric similarity constant C L =4 of step 2.1, which are transformed into on-orbit cabin segment coordinates X t ,Y t ,Z t ; Further, the ground test data set I t =[X t ,Y t ,Z t ,F i ,R i ] is obtained, containing 249 channels (83 strain measurement points). Among them, F i represents 10 water pressure loading levels, R i is the corresponding test response data, and subscript i represents the serial number of each water pressure loading level, i=1,2,3,…,10.

[0056] Step four: in the on-orbit, flight phase, real-time collection of real structure on-orbit, flight measured data, obtain measured sensor data set. Specifically:

[0057] Collect on-orbit cabin segment sensor coordinate information, real-time response data, real-time load size, and carry out abnormal value detection. In a certain period of time, the air pressure sensor, on-orbit cabin segment strain channel has collected 157,000 time points of steady-state strain data. According to the Ljapunov criterion, the abnormal data deviating from the mean value by more than 3 times the sample standard deviation is removed for the air pressure sensor and each strain channel, and then the air pressure or strain mean value is recalculated as the measured response value of the channel, and the measured sensor data set I a =[X a ,Y a ,Z a ,R a ,F a ] is obtained, containing 24 channels (8 strain measurement points), wherein X a , Y a , Za R is the sensor coordinate value a F is the measured response data a =102kPa is the measured load value.

[0058] Step five: data expansion phase, based on the linear superposition principle and model coordinate mapping method, the ground test data set obtained in step three is mapped to the on-orbit, flying sensor data, and the expanded sensor data set is obtained. Specifically:

[0059] Step 5.1, since step two adopts the equivalent test loading scheme, based on the linear superposition principle, the response of the on-orbit cabin under the uniform internal air pressure working condition can be expressed as the linear combination of the response R i of each water pressure loading level in the ground test, and F d is expressed in the same linear combination form as F i :

[0060]

[0061] The strain data of 249 channels and 10 loading levels in the ground water pressure test are regarded as the strain of the test piece under the condition of only filling water (triangular distribution internal pressure load), and the strain of the test piece under the condition of step-by-step loading (uniform internal pressure load) is superimposed, wherein the latter is the same as the uniform internal air pressure load of the on-orbit cabin, and can be expressed by the strain difference between the water pressure loading levels, as shown in Figure 3 According to the difference method, the coefficients of each water pressure loading level are determined:

[0062]

[0063] Since there is a deviation between the on-orbit cabin load F=100kPa and the measured load value F a =102kPa in step four, the load ratio k is calculated:

[0064]

[0065] Therefore, the value of is corrected to 1.02 times the original value. Further, the equivalent response data set I of the ground test under the on-orbit, flying working condition of the on-orbit cabin is obtained.

[0066] Step 5.2, based on the coordinate consistency of the test model after step three coordinate conversion and the on-orbit cabin, the test measuring points can be mapped to the on-orbit, flying measuring points, the equivalent response data set I e of the ground test is incorporated into the measured sensor data set I a , and the expanded sensor data set I , including 273 channels (91 measurement points, 8 of which are redundant for evaluating the correlation between ground and space data) data, such as Figure 4 .

[0067] Step six: data verification phase, correlation evaluation and adjustment of expanded sensor data. Specifically:

[0068] According to step 5.1 ground test equivalent response data set I e Response data R c of ground test measurement point set S1 a Correlation analysis is carried out between the measured response data R c RRMSE=0.17 of the 24 channels does not meet the requirements, data adjustment is carried out, including: correcting and redeveloping ground test, combining numerical simulation to exclude abnormal values, ground test response correction based on spatial interpolation. First, compare the response data R a of ground test measurement point set S1 final , for channels with large errors (>50με), re-patch and re-test to exclude test errors; then, based on radial basis function interpolation, build a test-measured error model for correcting ground test data, after correction RRMSE=0.06, finally get the expanded and corrected sensor data set I final .

[0069] Based on this data set, digital twin monitoring is carried out, and the full-field mechanical response information R=f(X,Y,Z) is obtained, and the von-Mises stress field is shown in Figure 5 The monitoring average absolute error is 4.7MPa, compared with the digital twin monitoring using the unexpanded sensor data set (the average absolute error is 11.5MPa), the accuracy is obviously improved, and the proposed method can realize high-precision modeling for digital twin.

[0070] The above-described embodiments only express the implementation of the present application, but cannot be interpreted as limiting the scope of the present application. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of protection of the present application.

Claims

1. A method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism, characterized in that, The method for expanding sensor data from impoverished samples includes the following steps: Step 1: Simulation analysis stage, establish a parameterized simulation model of the real structure, carry out simulation analysis, and obtain a dataset of the simulation response of the real structure; Step 2: Ground test design stage. Based on the actual structure and the size and load state of the model that can be tested, and based on the similarity scaling criterion and parameter optimization method, design an equivalent ground test model and loading scheme. Step 3: Ground test phase. Based on the actual structural sensor locations and key structural areas, ground test measurement points are arranged, and the coordinates of the ground test model are converted into the coordinates of the actual structure based on the geometric similarity constant to obtain the ground test dataset. Step 4: During the on-orbit and flight phases, collect real-time on-orbit and flight measurement data of the actual structure to obtain the measured sensor dataset; Step 5: Data expansion stage. Based on the principle of linear superposition and the model coordinate mapping method, the ground test dataset obtained in Step 3 is mapped to on-orbit and flight sensor data to obtain the expanded sensor dataset. Step Six: Data Verification Phase. Conduct correlation assessment and adjustment on the expanded sensor data.

2. The method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism according to claim 1, characterized in that, Step one specifically involves: First, a parametric simulation model is established, with parameters including geometric feature parameters and load parameters. Then, the geometric features of the real structure and the load F on the real structure are input into the parametric simulation model, and the simulation analysis is run to obtain the simulation response dataset I of the real structure. f =[X f ,Y f Z f ,R f ], where X f ,Y f Z f R represents the coordinates of the actual structural simulation model. f This is a simulation response of the real structure.

3. The method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism according to claim 1, characterized in that, Step two specifically involves: Step 2.1: Design a scaled-down test model as a ground test model; First, determine the geometric similarity constant C based on the model dimensions used in the experiment. L =L real / L scaled Based on similarity theory, the model is scaled down overall, where L real L represents the actual structural dimensions. scaled This represents the scaled-down experimental model dimensions; then, based on parameter optimization methods, the simulation response R of the actual structure is calculated. f Adjust the parameters of the scaled-down test model to obtain a scaled-down test model that meets the test conditions and the test load F. d For cases where a full-scale experimental model can be established, meaning the geometry and material properties of the test specimen are consistent with the actual structure, the full-scale experimental model serves as the ground-based experimental model, and F is taken as... d =F, where F is the actual load used in the structure; Step 2.2, considering the ground test loading conditions, if the load F used in the test in Step 2.1 can be... d Then the test loading scheme and F d Maintain consistency; if loading conditions are limited, then within the linear elastic range of the material, the test load F will be kept constant. d Replace with N easily implementable loading states or loading levels F i : Where the subscript i represents the sequence number of each loading state or loading level, α is the corresponding coefficient, i=1,2,3,…,N; that is, through F i By superimposing and simulating the actual load distribution, an equivalent test loading scheme is obtained.

4. The method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism according to claim 3, characterized in that, In step 2.1, the parameter optimization design variables of the scaled-down test model include the geometric characteristic parameters and load magnitude of the scaled-down test model. In each iteration, the design variables are input into the parameterized simulation model from step one to obtain the scaled-down test simulation model. The optimization objective is to optimize the ratio of the simulated response of the scaled-down test model to the simulated response of the actual structure, R. f To maximize relevance, optimization constraints include manufacturing feasibility and sensor placement feasibility.

5. The method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism according to claim 3, characterized in that, Step three specifically involves: Step 3.1: Arrange the test measurement point set S1 at the same location as the measurement points of the ground test model and the real structure to evaluate the correlation between the ground and space data; Step 3.2: Arrange test measurement point set S2 in key areas not covered by the actual structure measurement points to provide mechanical response information. The key areas can be determined by combining engineering experience and the simulation response of the actual structure. Step 3.3: Conduct experiments under the loading scheme established in Step 2, and record the response data at each measuring point. If a full-scale test model is used in Step 2, the coordinates of the ground test model and the X coordinates of the actual structure are... t ,Y t Z t The same applies; if a scaled-down test model is used in step two, a scaled-down test coordinate system needs to be established on the scaled-down test model, with its origin and coordinate axis directions being the same as the real structure, and the X coordinates of the measurement points on the scaled-down test model being... ts ,Y ts Z ts Multiply by the geometric similarity constant C from step 2.1 L Transformed into true structural coordinates X t ,Y t Z t ; and thus obtain the ground test dataset I t =[X t ,Y t Z t ,F i ,R i ]; where F i For loading status or loading level, R i For the corresponding test response data, the subscript i represents the sequence number of each loading state or loading level, i=1,2,3,…,N.

6. The method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism according to claim 5, characterized in that, Step four specifically involves: Collect the coordinate information of each sensor on the real structure, real-time response data, and real-time load magnitude, and perform outlier detection to obtain the measured sensor dataset I. a =[X a ,Y a Z a ,R a ,F a ], where X a Y a Z a R represents the sensor coordinates. a For measured response data, F a This represents the measured load value.

7. The method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism according to claim 6, characterized in that, The outlier detection methods include the Laida criterion, setting thresholds, or the isolated forest algorithm.

8. The method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism according to claim 5, characterized in that, Step five specifically involves: Step 5.1: If the test loading scheme in Step 2 is the same as the test load, then the test response value is directly taken as the actual structural response under on-orbit and flight conditions. If step two adopts an equivalent test loading scheme, then based on the principle of linear superposition, the actual structure's response under on-orbit and flight conditions will be... It can be represented as the response R under each experimental loading state or loading level. i The linear combination of the load F used in step two of the experiment d Represented as F i The linear combination forms are the same: If the actual structure uses load F and the measured load value F in step four... a If a deviation exists, it should be corrected according to the load ratio k. This leads to the acquisition of a dataset of equivalent ground-based responses to the actual structure under on-orbit and flight conditions. ; Step 5.2: Based on the consistency between the experimental model and the actual structural coordinates after coordinate transformation in Step 3, the experimental measurement points can be mapped to on-orbit and flight measurement points, and the equivalent response dataset I from the ground test can be generated. e Incorporate measured sensor dataset I a The expanded sensor dataset is obtained. .

9. A method for expanding the capacity of low-sample sensor data based on the mechanical consistency mechanism according to claim 8, characterized in that, In step 5.1, the load ratio k is defined as: ,in This represents the measured load value. This indicates that the actual structure uses loads. It represents the vector norm or the absolute value of a scalar.

10. A method for expanding the capacity of poor-sample sensor data based on the mechanical consistency mechanism according to claim 8, characterized in that, Step six specifically involves: Based on step 5.1, the equivalent response dataset I from the ground test. e Response data R of ground test measurement point set S1 c , and the measured response data R a Correlation analysis was conducted, and the evaluation indicators included R. 2 , RRMSE; If the correlation evaluation results do not meet the requirements, data adjustment will be carried out, including: correcting and re-conducting ground tests, removing outliers by combining numerical simulation, and correcting ground test responses based on spatial interpolation. Finally, the expanded and corrected sensor dataset I is obtained. final To achieve data expansion for high-precision modeling of digital twins; The correlation evaluation result does not meet the requirements, meaning that: R 2 If the RRMSE is below a set threshold, which is a dimensionless number less than 1; or if the RRMSE is above a set threshold, which is a dimensionless number greater than 0.

Images