Composite material impact probability positioning method and device

Abstract

The invention provides a composite material impact probability positioning method and device, and the method comprises the steps: carrying out the meshing of a square plate of a to-be-detected composite material according to a first interval, and obtaining a plurality of first grids; after an impact object impacts the square plate of the composite material to be detected, determining the time difference of impact signals collected by each pair of sensors; processing the time difference and the position of each first grid by using a Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors; performing weighted summation on the plurality of likelihood values of each first grid to obtain a positioning probability of the first grid; and determining the central point position of the first grid corresponding to the maximum value of the positioning probability as the impact position of the square plate of the to-be-detected composite material. The method improves the impact positioning precision of the composite material, improves the detection precision of the damage generated when the aircraft impacts the composite material, and has the advantages of low cost, time saving and labor saving.

Description

Technical Field

[0001] The present application relates to the technical field of impact damage position detection, and in particular to a composite material impact probability positioning method and device. Background Art

[0002] At present, non-destructive testing methods such as ultrasonic testing, eddy current testing and magnetic particle testing are mainly used to identify damage caused by aircraft collisions. These methods often require expensive and large-scale equipment to be carried out on disassembled structures. In addition, such methods rely on careful planning and regular implementation by ground personnel. Therefore, they have many problems such as large workload, long time consumption, and difficulty in ensuring the safety of aircraft during operation.

[0003] In impact monitoring of composite materials, accurate identification of the impact location helps to determine the location of possible aircraft damage, thereby accelerating the inspection and maintenance of key parts of the aircraft, effectively reducing the time and economic costs of structural maintenance.

[0004] Currently, no effective solution has been proposed for the accurate identification of the impact position of composite materials. Summary of the invention

[0005] In view of this, the present application provides a composite material impact probability positioning method and device to solve the above technical problems.

[0006] In a first aspect, an embodiment of the present application provides a composite material impact probability positioning method, which is used to perform impact positioning on a square plate of a composite material to be tested, wherein a plurality of sensors are arranged around the square plate of the composite material to be tested; the method comprises: Meshing the square plate of the composite material to be tested according to the first spacing to obtain a plurality of first meshes, and determining the position of the center point of each first mesh as the position of the first mesh; After the impact object hits the square plate of the composite material to be tested, determine the time difference of the impact signal collected by each pair of sensors; Using the Gaussian process regression model corresponding to each pair of sensors to process the time difference and the position of each first grid, obtain the likelihood value of each first grid under each pair of sensors; performing weighted summation on multiple likelihood values ​​of each first grid to obtain the positioning probability of the first grid; The center point position of the first grid corresponding to the maximum positioning probability is determined as the impact position of the square plate of the composite material to be tested.

[0007] In a possible implementation, the time difference and the position of each first grid are processed using a Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors, including: For the j-th Gaussian process regression model corresponding to the j-th pair of sensors, at the center point of the first grid The likelihood value at satisfy:

[0008] in, is the time difference of the jth pair of sensors; is the optimal estimate of the hyperparameter vector for the jth pair of sensors is the center point position of the first grid The predicted value at for The mean of for The variance of .

[0009] In one possible implementation, the method further includes: Meshing the square plate of the composite material to be tested according to a second spacing to obtain a plurality of second grids, and determining the position of the center point of each second grid as the position of the second grid; wherein the second spacing is greater than the first spacing; Sequentially drop a steel ball from a preset height onto each second grid of the square plate of the composite material to be tested, and obtain the time difference of the jth pair of sensors of each second grid; Calculate the log-likelihood :

[0010] Where N is the number of the second grid, is the time difference vector, the nth component is the time difference of the jth pair of sensors in the nth second grid, is the time difference vector The kernel matrix, is the center point position vector of N second grids; is the hyperparameter vector of the j-th pair of sensors to be estimated, including: noise variance , signal variance and the length scale parameter ; The optimal estimate of the hyperparameter vector for the jth pair of sensors for:

[0011] The optimal estimate is found through the search algorithm , so that the negative log-likelihood value reaches the minimum value.

[0012] In one possible implementation, the search algorithm adopts a whale optimization algorithm; The optimal estimate is found through the search algorithm , so that the negative log-likelihood value reaches the minimum value, including: Initialize the M search particles of the population, and use the initialized M search particles as the M search particles of the first iteration; wherein the position of each search particle is represented as the hyperparameter vector of the j-th pair of sensors; The search particle with the smallest fitness value among the initialized M search particles is determined as the optimal search particle for the first iteration; Repeat the following steps: Update the positions of the M search particles of the previous iteration to obtain the positions of the M search particles of the current iteration; Substitute the position of each search particle into the fitness function to obtain the fitness value of each search particle; determine the search particle with the smallest fitness value as the optimal search particle for the current number of iterations; Based on the best search particle of the previous iteration and the best search particle of the current iteration, the best search particle of the current iteration is updated; Until the current number of iterations reaches the preset maximum number of iterations; The optimal search particle with the maximum number of iterations is determined as the optimal estimate .

[0013] In a second aspect, an embodiment of the present application provides a composite material impact probability positioning device, which is used to impact position a square plate of a composite material to be tested, wherein a plurality of sensors are arranged around the square plate of the composite material to be tested; the device comprises: A gridding unit, used for gridding the square plate of the composite material to be tested according to a first spacing to obtain a plurality of first grids, and determining the position of the center point of each first grid as the position of the first grid; a determination unit, for determining a time difference between impact signals collected by each pair of sensors after the impact object hits the square plate of the composite material to be tested; A processing unit, configured to process the time difference and the position of each first grid using a Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors; and perform weighted summation on multiple likelihood values ​​of each first grid to obtain a positioning probability of the first grid; The positioning unit is used to determine the center point position of the first grid corresponding to the maximum positioning probability as the impact position of the square plate of the composite material to be tested.

[0014] In a possible implementation, the time difference and the position of each first grid are processed using a Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors, including: For the j-th Gaussian process regression model corresponding to the j-th pair of sensors, at the center point of the first grid The likelihood value at satisfy:

[0015] in, is the time difference of the jth pair of sensors; is the optimal estimate of the hyperparameter vector for the jth pair of sensors is the center point position of the first grid The predicted value at for The mean of for The variance of .

[0016] In a possible implementation, the apparatus further includes a training unit, specifically configured to: Meshing the square plate of the composite material to be tested according to a second spacing to obtain a plurality of second grids, and determining the position of the center point of each second grid as the position of the second grid; wherein the second spacing is greater than the first spacing; Sequentially drop a steel ball from a preset height onto each second grid of the square plate of the composite material to be tested, and obtain the time difference of the jth pair of sensors of each second grid; Calculate the log-likelihood :

[0017] Where N is the number of the second grid, is the time difference vector, the nth component is the time difference of the jth pair of sensors in the nth second grid, is the time difference vector The kernel matrix, is the center point position vector of N second grids; is the hyperparameter vector of the j-th pair of sensors to be estimated, including: noise variance , signal variance and the length scale parameter ; The optimal estimate of the hyperparameter vector for the jth pair of sensors for:

[0018] The optimal estimate is found through the search algorithm , so that the negative log-likelihood value reaches the minimum value.

[0019] In one possible implementation, the search algorithm adopts a whale optimization algorithm; Search the algorithm to find the optimal estimate , so that the negative log-likelihood value reaches the minimum value, including: Initialize the M search particles of the population, and use the initialized M search particles as the M search particles of the first iteration; wherein the position of each search particle is represented as the hyperparameter vector of the j-th pair of sensors; The search particle with the smallest fitness value among the initialized M search particles is determined as the optimal search particle for the first iteration; Repeat the following steps: Update the positions of the M search particles of the previous iteration to obtain the positions of the M search particles of the current iteration; Substitute the position of each search particle into the fitness function to obtain the fitness value of each search particle; determine the search particle with the smallest fitness value as the optimal search particle for the current number of iterations; Based on the best search particle of the previous iteration and the best search particle of the current iteration, the best search particle of the current iteration is updated; Until the current number of iterations reaches the preset maximum number of iterations; The optimal search particle with the maximum number of iterations is determined as the optimal estimate .

[0020] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method of the embodiment of the present application when executing the computer program.

[0021] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions implement the method of the embodiment of the present application when executed by a processor.

[0022] The present application improves the impact positioning accuracy of composite materials, thereby improving the damage detection accuracy caused by aircraft impacting composite materials, and has the advantages of low cost, time saving and labor saving. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1A flow chart of a composite material impact probability positioning method provided in an embodiment of the present application; Figure 2 A schematic diagram of an experimental device provided in an embodiment of the present application; Figure 3 A schematic diagram of the arrival time of signals received by four sensors provided in an embodiment of the present application; Figure 4 A schematic diagram comparing five search algorithms provided in the embodiments of the present application; Figure 5 The arrival time difference of each location collected by the experiment provided in the embodiment of the present application; Figure 6 A schematic diagram of a denser mapping of impact positions and arrival time differences learned through a Gaussian process provided in an embodiment of the present application; FIG. 7 ( a ) is a schematic diagram of the likelihood of arrival time differences at various positions collected under sensor 1 in an impact event provided by an embodiment of the present application; FIG. 7 ( b ) is a schematic diagram of the likelihood of arrival time differences at various positions collected under sensor 2 in an impact event provided by an embodiment of the present application; FIG. 7 ( c ) is a schematic diagram of the likelihood of arrival time differences at various positions collected under the sensor 3 in an impact event provided by an embodiment of the present application; FIG. 7 ( d ) is a schematic diagram of the likelihood of arrival time differences at various positions collected under the sensor 4 in an impact event provided by an embodiment of the present application; Figure 8 A schematic diagram of positioning prediction results provided in an embodiment of the present application; Fig. 9 A schematic diagram of comparing the predicted position (most likely position) of each observation sample in the test set with the actual position provided in an embodiment of the present application; Fig.10 A schematic diagram of a square composite material plate with a grid spacing of 100 mm provided in an embodiment of the present application; Fig.11 A functional structure diagram of a composite material impact probability positioning device provided in an embodiment of the present application; Fig.12 A structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0026] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0027] First, the design concept of the embodiments of the present application is briefly introduced.

[0028] Composite materials are new materials that are made of two or more materials of different properties, shapes or qualities through a certain composite process. They have the advantages of light weight, high specific strength, high specific stiffness, and chemical corrosion resistance. They have been widely used in aerospace, shipbuilding, automotive and other engineering fields. The use of composite materials can simplify the molding process of the structure, greatly reduce the weight of the aircraft to improve flight performance, and is an effective way to directly reduce operating costs. The increasingly widespread use of composite materials in the aerospace field has led people to put forward more stringent requirements on the safety of materials.

[0029] Composite materials have the disadvantages of low lateral interface strength, brittle fiber and resin matrix, and are sensitive to impact loads. When the impact energy reaches a certain threshold, the material is prone to matrix cracking, panel delamination, fiber breakage and other damage, which significantly reduces the overall strength and reliability of the structure. Unlike traditional metal materials, the damage inside this composite material is difficult to detect with the naked eye in the early stage. Especially for low-speed impact, if it is not treated, the internal damage will gradually expand during the long-term service process, causing a sharp decline in structural stability, and even weakening the structural strength by 35% - 40%, which may lead to catastrophic consequences. However, during the service of aircraft, it is inevitable that they will be impacted by the outside world, such as bird strikes, hail impacts, sand and gravel impacts during takeoff and landing, and tool fall-off impacts during maintenance. These impacts may not cause immediate overt damage, but they are very likely to cause hidden damage inside the structure, which is more dangerous than overt damage to some extent.

[0030] Aiming at the impact location problem of composite materials with highly nonlinear wave fields, the present application provides a composite material impact probability location method, which uses relative position vectors and corresponding arrival time differences to learn the composite material structure information, and uses the Whale Optimization Algorithm (WOA) to optimize the hyperparameters of GPR to achieve probabilistic location of the impact position for a given observed arrival time difference.

[0031] The method provided in the embodiment of the present application can provide an effective technical means for the safety and stability of the aircraft during use, and adopt a condition-based maintenance strategy to replace the traditional plan-based maintenance strategy in the maintenance process, thereby reducing maintenance costs and having important engineering application value. In impact monitoring, accurate identification of the impact position helps to determine the location where damage may occur, thereby accelerating the inspection and maintenance of key parts, which can effectively reduce the time and economic costs of structural maintenance. Estimation of the impact effect is an important way to determine the degree and mode of damage to composite structures, and is of great significance to structural integrity.

[0032] After introducing the application scenarios and design concepts of the embodiments of the present application, the technical solutions provided by the embodiments of the present application are described below.

[0033] like Figure 1 As shown, the embodiment of the present application provides a composite material impact probability positioning method, which is used to impact position a square plate of the composite material to be tested, wherein a plurality of sensors are arranged around the square plate of the composite material to be tested; comprising: Step 101: gridding a square plate of the composite material to be tested according to a first spacing to obtain a plurality of first grids, and determining the position of the center point of each first grid as the position of the first grid; Step 102: after the impact object hits the square plate of the composite material to be tested, determine the time difference of the impact signal collected by each pair of sensors; Step 103: using the Gaussian process regression model corresponding to each pair of sensors to process the time difference and the position of each first grid, to obtain a likelihood value of each first grid under each pair of sensors; performing weighted summation on multiple likelihood values ​​of each first grid, to obtain a positioning probability of the first grid; Step 104: Determine the center point position of the first grid corresponding to the maximum positioning probability as the impact position of the square plate of the composite material to be tested.

[0034] Preferably, the number of the sensors is 4, which are respectively arranged at the four corners of the square plate of the composite material to be tested.

[0035] In some embodiments, the time difference and the position of each first grid are processed using a Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors, including: For the j-th Gaussian process regression model corresponding to the j-th pair of sensors, at the center point of the first grid The likelihood value at satisfy:

[0036] in, is the time difference of the jth pair of sensors; is the optimal estimate of the hyperparameter vector for the jth pair of sensors is the center point position of the first grid The predicted value at for The mean of for The variance of .

[0037] In some embodiments, the method further comprises: Meshing the square plate of the composite material to be tested according to a second spacing to obtain a plurality of second grids, and determining the position of the center point of each second grid as the position of the second grid; wherein the second spacing is greater than the first spacing; Sequentially drop a steel ball from a preset height onto each second grid of the square plate of the composite material to be tested, and obtain the time difference of the jth pair of sensors of each second grid; Calculate the log-likelihood :

[0038] Where N is the number of the second grid, is the time difference vector, the nth component is the time difference of the jth pair of sensors in the nth second grid, is the time difference vector The kernel matrix, is the center point position vector of N second grids; is the hyperparameter vector of the j-th pair of sensors to be estimated, including: noise variance , signal variance and the length scale parameter ; The optimal estimate of the hyperparameter vector for the jth pair of sensors for:

[0039] The optimal estimate is found through the search algorithm , so that the negative log-likelihood value reaches the minimum value.

[0040] In some embodiments, the search algorithm adopts a whale optimization algorithm, including: Initialize the M search particles of the population, and use the initialized M search particles as the M search particles of the first iteration; wherein the position of each search particle is represented as the hyperparameter vector of the j-th pair of sensors; The search particle with the smallest fitness value among the initialized M search particles is determined as the optimal search particle for the first iteration; Repeat the following steps: Update the positions of the M search particles of the previous iteration to obtain the positions of the M search particles of the current iteration; Substitute the position of each search particle into the fitness function to obtain the fitness value of each search particle; determine the search particle with the smallest fitness value as the optimal search particle for the current number of iterations; Based on the best search particle of the previous iteration and the best search particle of the current iteration, the best search particle of the current iteration is updated; Until the current number of iterations reaches the preset maximum number of iterations; The specific implementation process of this application is described below in conjunction with a specific application scenario.

[0041] For the 710mm×540mm×50mm glass fiber composite sandwich panel, which consists of two layers of glass fiber reinforced composite material (Glass Fiber Reinforced Plastic, GFRP) surface and polyvinyl chloride foam (PVC) interlayer, the glass fiber fabric is laid in [0° / 90° / 90° / 0°]2S configuration. The grid is divided on the composite panel with a spacing of 50mm. The low-speed impact is generated by a small steel ball with a diameter of 18.3mm and a weight of 30g falling from a height of 10cm. Five impacts are performed at each intersection. The sensor and impact point layout are as follows Figure 2 As shown. Four piezoelectric ceramic sensors are attached to the composite plate to capture the acoustic emission signals generated by low-speed impact. The preamplifier gain is 40dB and the sampling frequency is 2.5MHz. Each channel collects 640 impact acoustic emission signals, and a total of 4 channels of acoustic emission signals are obtained.

[0042] Before training the GPR model, the collected data must be processed. First, the difference in time of arrival of the impact signal at different sensors (dTOA) is extracted. Figure 3 The arrival time of the signal received by the four sensors at the impact point (200,300) is given. Figure 3It can be found that the acoustic emission signal generated by the impact reaches sensor 4 first and sensor 2 last. However, the impact point is closest to sensor 3 in the structure, but the impact signal arrives at sensor 3 later than sensor 4, which also shows that the propagation speed of waves in composite materials is inconsistent. In this anisotropic material, the traditional method is not suitable.

[0043] After calculating the arrival time of the impact signal on each sensor, the arrival time difference of each pair of sensors is calculated separately. Using 4 sensors can produce 6 different pairing combinations, and a vector containing 6 arrival time differences will be generated for each impact. After the data set is established, the Gaussian process representation of the impact position and arrival time difference mapping is learned for each pair of sensors. When using the sum of the SE with anisotropic kernel and the Matérn3 / 2 covariance function as the kernel function of the Gaussian process model, the parameter range of the model and the results of optimization using WOA are shown in Table 1: Table 1: Model parameters and optimization results

[0044] Figure 4 The optimization performance of WOA is compared with that of BFGS algorithm, Scale Conjugate Gradient algorithm (SCG), Gray Wolf Optimizer (GWO) and Genetic Algorithm (GA). The parameters of WOA, GA and GWO are shown in Table 2. Figure 4 It can be found that WOA can converge to a better result than the genetic algorithm and the gray wolf optimization algorithm due to its good global search ability and local search ability. Although the BFGS algorithm and the conjugate gradient algorithm can also converge to a lower value, WOA is more efficient and converges to a better result using fewer iterations.

[0045] Table 2: WOA, GA, GWO algorithm parameters

[0046] Figure 5 and Figure 6 The learning results of the spatial position and arrival time difference mapping for the combination of sensor 1 and sensor 4 are given. Figure 5 The arrival time difference of each location collected in the experiment is used to learn the Gaussian process model. Figure 6It is a denser mapping of impact position and arrival time difference learned through Gaussian process. It can be found that the denser mapping is not obtained based on linear interpolation calculation, but is obtained based on the training mechanism of GPR model, using training data to learn the complex characteristics of composite material structure, which is also the principle of GPR to achieve high-resolution positioning. This is also the reason why the first spacing is smaller than the second spacing. When impact positioning, the grid of the composite square plate can be divided smaller to achieve high-resolution positioning.

[0047] Next, the positioning prediction is carried out. In the experiment, the collected data is divided into a training set and a test set, and the training set is used to train the hyperparameters of the Gaussian process model corresponding to each pair of sensors. Figures 7 (a), 7 (b), 7 (c), and 7 (d) show the likelihood of the arrival time difference collected at each location under sensor 1 (Sensor_1), sensor 2 (Sensor_2), sensor 3 (Sensor_3), and sensor 4 (Sensor_4) for a randomly selected impact event in the test set. Among them, "×" is the true position, and the brighter the color, the greater the likelihood value, that is, the greater the possibility of it being the impact position. In each pair of sensors, it can be found that the true position is on the bright strip, however, the area span of the bright strip is large, and it is difficult to determine a relatively accurate impact position. According to the acoustic emission theory, when the acoustic emission source is determined based on the fastest wave arrival time difference, if the acoustic emission source is not on the connection path of the two sensors, at least three sensors are required to locate the acoustic emission source.

[0048] use Figure 2 The data in the above table is combined with six pairs of sensors to calculate the coordinates of the predicted points. The prediction results are as follows: Figure 8 As shown in the figure. The "×" on the figure indicates the location of the real impact, and the brighter the color, the greater the likelihood value. It can be found that there is an area with a large likelihood value near the real impact location, that is, the location where the impact occurs is likely to be in this area, which also shows that the impact location is accurately predicted. In addition, the figure shows the probability of impact at various locations of the structure. During the maintenance process, maintenance personnel can choose to perform maintenance within a certain confidence area based on experience, thereby avoiding the time and economic costs caused by large-scale maintenance.

[0049] Compare the predicted position (most likely position) of each observation in the test set with the true position, and the results are as follows: Fig. 9As shown. The figure shows the area (99.5% confidence area) where the likelihood value of each test point is between the upper 0.005 quantile and the upper bound, and the area of ​​this area is 0.5% of the structural area. From the results, it can be found that the true position basically falls within the 99.5% confidence range of the predicted results. The area of ​​these predicted areas is only 99.5% of the area of ​​the overall structure, and maintenance personnel can detect the structure at a lower cost. In addition, it can be found from the figure that there is a significant difference between some predicted positions and the true positions. In this case, estimating a possible area where an impact occurs is more valuable for reference than a certain impact position coordinate.

[0050] In order to evaluate the overall performance of the algorithm, the root mean square error of the predicted position of the test sample is calculated according to the following formula, where N represents the number of test samples.

[0051]

[0052] The advantage of the probability-based method is that it allows the identification of areas greater than a certain probability threshold. In addition, it is more desirable to detect small areas during the detection process. In order to evaluate the performance of the model from this perspective, two evaluation indicators for the probability positioning method are defined: ACCP and Smin, where ACCP represents the accuracy of the predicted area where the true impact position falls within a given threshold p. , represents the quantile of the prediction result. For example, when p=0.975, ACC0.975 represents the accuracy of the true impact position in the prediction area with a confidence level greater than 97.5% among all test samples. The minimum detection threshold Smin represents the percentage of the detection structure area required to detect the true impact position, that is, the minimum upper quantile of the impact position detected.

[0053] It can be found that the prediction results are related to the results of the covariance function. Here, a total of four covariance functions are explored, namely SE, Matérn3 / 2, SE + Matérn3 / 2, and SE × Matérn3 / 2, while considering the influence of various homogeneous kernels and heterogeneous kernels. The test points on the test set are predicted, and the results are shown in Table 3.

[0054] It can be seen from the table that the performance of the anisotropic kernel function is better than that of the isotropic kernel function, because the anisotropy of the composite material causes each dimension to be asymmetric. In addition, the square exponential covariance function SE has the worst performance, because this strong smoothness assumption is unrealistic for composite impact location modeling. The composite covariance function of SE and Matérn3 / 2 covariance function inherits the advantages of these two kernel functions, and the kernel function formed by their addition has the best performance. The root mean square error on the anisotropic kernel is 8.94mm. When the threshold is set to 97.5%, the true location of the impact can be 100% identified. On average, only 0.15% of the structural area needs to be detected to identify the true impact location.

[0055] Table 3: Performance comparison of four kernel functions

[0056] In data-driven methods, the amount of data in the training set has an important impact on the performance of the model. In order to study the sensitivity of the positioning algorithm to the number of data points used in the training phase, a training set with different grid spacings is constructed to learn the Gaussian process representation, and the test points on the test set are predicted and the RMSE under each grid spacing is calculated as the evaluation index. The grid spacing is defined as the distance of each training point in the x and y directions. In addition to the 50mm grid spacing used in the above analysis, a 100mm grid spacing is also defined, such as Fig.10 shown.

[0057] In addition, in order to evaluate the impact of the grid on the model, the accuracy of the true impact position falling into the 97.5% confidence interval and the minimum detection threshold were calculated. ,The model performance of each grid spacing training is shown in Table 4.

[0058] In the 100 mm grid spacing, the root mean square error is 31.62 mm, and the accuracy is affected to a certain extent. If the traditional coordinate-based positioning method is used, it is difficult to determine the true impact location. Using the GPR-based method, the probability results of impact positioning are provided. The number of training grid points using a 100 mm grid spacing is 36, which is only 28% of the 50 mm grid spacing, but on average, at least 1.27% of the structure area needs to be tested to find the true impact location. In addition, this probability-based method also provides maintenance personnel with guiding maintenance strategies, which can locate the impact damage location more quickly.

[0059] Table 4: Effects of different grid spacings

[0060] In addition to the impact location problem, this method can also solve the location of certain damage, provided that the damage needs to generate an acoustic emission event when it occurs. In addition, in addition to anisotropic composite materials, this method is also applicable to complex structures with special parts such as holes and bosses. Through offline learning of the model, the probabilistic location of impact or damage can be achieved in real time, providing guidance for further inspection and maintenance.

[0061] Based on the same inventive concept, the embodiment of the present application provides a composite material impact probability positioning device for impact positioning of a square plate of the composite material to be tested, wherein a plurality of sensors are arranged around the square plate of the composite material to be tested, Fig.11 As shown, the composite material impact probability positioning device 200 provided in the embodiment of the present application at least includes: A gridding unit 201 is used to grid the square plate of the composite material to be tested according to a first spacing to obtain a plurality of first grids, and determine the position of the center point of each first grid as the position of the first grid; A determination unit 202, for determining a time difference between impact signals collected by each pair of sensors after the impact object hits the square plate of the composite material to be tested; The processing unit 203 is used to process the time difference and the position of each first grid using the Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors; and perform weighted summation on multiple likelihood values ​​of each first grid to obtain a positioning probability of the first grid; The positioning unit 204 is used to determine the center point position of the first grid corresponding to the maximum positioning probability as the impact position of the square plate of the composite material to be tested.

[0062] It should be noted that the principle of solving the technical problem by the composite material impact probability positioning device 200 provided in the embodiment of the present application is similar to the method provided in the embodiment of the present application. Therefore, the implementation of the composite material impact probability positioning device 200 provided in the embodiment of the present application can refer to the implementation of the method provided in the embodiment of the present application, and the repeated parts will not be repeated.

[0063] like Fig.12 As shown, the electronic device 300 provided in the embodiment of the present application includes at least: a processor 301, a memory 302, and a computer program stored in the memory 302 and executable on the processor 301. When the processor 301 executes the computer program, the composite material impact probability positioning method provided in the embodiment of the present application is implemented.

[0064] The electronic device 300 provided in the embodiment of the present application may further include a bus 303 connecting different components (including the processor 301 and the memory 302). The bus 303 represents one or more of several types of bus structures, including a memory bus, a peripheral bus, a local bus, and the like.

[0065] The memory 302 may include a readable medium in the form of a volatile memory, such as a random access memory (RAM) 3021 and / or a cache memory 3022 , and may further include a read-only memory (ROM) 3023 .

[0066] The memory 302 may also include a program tool 3024 having a set (at least one) of program modules 3025, the program modules 3025 including but not limited to: an operating subsystem, one or more application programs, other program modules, and program data, each of which or some combination may include the implementation of a network environment.

[0067] The electronic device 300 may also communicate with one or more external devices 304 (e.g., keyboards, remote controls, etc.), and may also communicate with one or more devices that enable a user to interact with the electronic device 300 (e.g., mobile phones, computers, etc.), and / or communicate with any device that enables the electronic device 300 to communicate with one or more other electronic devices 300 (e.g., routers, modems, etc.). Such communication may be performed through an input / output (I / O) interface 305. Furthermore, the electronic device 300 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 306. Fig.12 As shown, the network adapter 306 communicates with other modules of the electronic device 300 via the bus 303. It should be understood that although Fig.12 Not shown, other hardware and / or software modules may be used in conjunction with the electronic device 300, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, disk arrays (Redundant Arrays of Independent Disks, RAID) subsystems, tape drives, and data backup storage subsystems.

[0068] It should be noted that Fig.12 The electronic device 300 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0069] The embodiment of the present application further provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the composite material impact probability positioning method provided in the embodiment of the present application is implemented.

[0070] In addition, although the operations of the method of the present application are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0071] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A composite material impact probability positioning method, used for impact positioning of a square plate of a composite material to be tested, wherein a plurality of sensors are arranged around the square plate of the composite material to be tested; characterized in that: include: Meshing the square plate of the composite material to be tested according to the first spacing to obtain a plurality of first meshes, and determining the position of the center point of each first mesh as the position of the first mesh; After the impact object hits the square plate of the composite material to be tested, determine the time difference of the impact signal collected by each pair of sensors; Using the Gaussian process regression model corresponding to each pair of sensors to process the time difference and the position of each first grid, obtain the likelihood value of each first grid under each pair of sensors; performing weighted summation on multiple likelihood values ​​of each first grid to obtain the positioning probability of the first grid; The center point position of the first grid corresponding to the maximum positioning probability is determined as the impact position of the square plate of the composite material to be tested.

2. The composite material impact probability positioning method according to claim 1, characterized in that: The time difference and the position of each first grid are processed using a Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors, including: For the j-th Gaussian process regression model corresponding to the j-th pair of sensors, at the center point of the first grid The likelihood value at satisfy: in, is the time difference of the jth pair of sensors; is the optimal estimate of the hyperparameter vector for the jth pair of sensors is the center point position of the first grid The predicted value at for The mean of for The variance of .

3. The composite material impact probability positioning method according to claim 2, characterized in that: The method further comprises: Meshing the square plate of the composite material to be tested according to a second spacing to obtain a plurality of second grids, and determining the position of the center point of each second grid as the position of the second grid; wherein the second spacing is greater than the first spacing; Sequentially drop a steel ball from a preset height onto each second grid of the square plate of the composite material to be tested, and obtain the time difference of the jth pair of sensors of each second grid; Calculate the log-likelihood : Where N is the number of the second grid, is the time difference vector, the nth component is the time difference of the jth pair of sensors in the nth second grid, is the time difference vector The kernel matrix, is the center point position vector of N second grids; is the hyperparameter vector of the j-th pair of sensors to be estimated, including: noise variance , signal variance and the length scale parameter ; The optimal estimate of the hyperparameter vector for the jth pair of sensors for: The optimal estimate is found through the search algorithm , so that the negative log-likelihood value reaches the minimum value.

4. The composite material impact probability positioning method according to claim 3, characterized in that: The search algorithm adopts the whale optimization algorithm; The optimal estimate is found through the search algorithm , so that the negative log-likelihood value reaches the minimum value, including: Initialize the M search particles of the population, and use the initialized M search particles as the M search particles of the first iteration; wherein the position of each search particle is represented as the hyperparameter vector of the j-th pair of sensors; The search particle with the smallest fitness value among the initialized M search particles is determined as the optimal search particle for the first iteration; Repeat the following steps: Update the positions of the M search particles of the previous iteration to obtain the positions of the M search particles of the current iteration; Substitute the position of each search particle into the fitness function to obtain the fitness value of each search particle; determine the search particle with the smallest fitness value as the optimal search particle for the current number of iterations; Based on the best search particle of the previous iteration and the best search particle of the current iteration, the best search particle of the current iteration is updated; Until the current number of iterations reaches the preset maximum number of iterations; The optimal search particle with the maximum number of iterations is determined as the optimal estimate .

5. A composite material impact probability positioning device, used for impact positioning of a square plate of a composite material to be tested, wherein a plurality of sensors are arranged around the square plate of the composite material to be tested; characterized in that: include: A gridding unit, used for gridding the square plate of the composite material to be tested according to a first spacing to obtain a plurality of first grids, and determining the position of the center point of each first grid as the position of the first grid; a determination unit, for determining a time difference between impact signals collected by each pair of sensors after the impact object hits the square plate of the composite material to be tested; A processing unit, configured to process the time difference and the position of each first grid using a Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors; Performing weighted summation on multiple likelihood values ​​of each first grid to obtain a positioning probability of the first grid; The positioning unit is used to determine the center point position of the first grid corresponding to the maximum positioning probability as the impact position of the square plate of the composite material to be tested.

6. The composite material impact probability positioning device according to claim 5, characterized in that: The time difference and the position of each first grid are processed using a Gaussian process regression model corresponding to each pair of sensors to obtain a likelihood value of each first grid under each pair of sensors, including: For the j-th Gaussian process regression model corresponding to the j-th pair of sensors, at the center point of the first grid The likelihood value at satisfy: in, is the time difference of the jth pair of sensors; is the optimal estimate of the hyperparameter vector for the jth pair of sensors is the center point position of the first grid The predicted value at for The mean of for The variance of .

7. The composite material impact probability positioning device according to claim 6, characterized in that: The device also includes a training unit, which is specifically used for: Meshing the square plate of the composite material to be tested according to a second spacing to obtain a plurality of second grids, and determining the position of the center point of each second grid as the position of the second grid; wherein the second spacing is greater than the first spacing; Sequentially drop a steel ball from a preset height onto each second grid of the square plate of the composite material to be tested, and obtain the time difference of the jth pair of sensors of each second grid; Calculate the log-likelihood : Where N is the number of the second grid, is the time difference vector, the nth component is the time difference of the jth pair of sensors in the nth second grid, is the time difference vector The kernel matrix, is the center point position vector of N second grids; is the hyperparameter vector of the j-th pair of sensors to be estimated, including: noise variance , signal variance and the length scale parameter ; The optimal estimate of the hyperparameter vector for the jth pair of sensors for: The optimal estimate is found through the search algorithm , so that the negative log-likelihood value reaches the minimum value.

8. The composite material impact probability positioning device according to claim 7, characterized in that: The search algorithm adopts the whale optimization algorithm; The optimal estimate is found through the search algorithm , so that the negative log-likelihood value reaches the minimum value, including: Initialize the M search particles of the population, and use the initialized M search particles as the M search particles of the first iteration; wherein the position of each search particle is represented as the hyperparameter vector of the j-th pair of sensors; The search particle with the smallest fitness value among the initialized M search particles is determined as the optimal search particle for the first iteration; Repeat the following steps: Update the positions of the M search particles of the previous iteration to obtain the positions of the M search particles of the current iteration; Substitute the position of each search particle into the fitness function to obtain the fitness value of each search particle; determine the search particle with the smallest fitness value as the optimal search particle for the current number of iterations; Based on the best search particle of the previous iteration and the best search particle of the current iteration, the best search particle of the current iteration is updated; Until the current number of iterations reaches the preset maximum number of iterations; The optimal search particle with the maximum number of iterations is determined as the optimal estimate .

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 4 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images