Method for positioning embedded defects of curved fiber variable-angle fiber placement component under different working conditions

Abstract

The invention discloses a method for positioning an embedded defect of a curved fiber variable-angle fiber placement component under different working conditions, and belongs to the technical field of material nondestructive testing, and the method comprises the following steps: arranging a distributed fiber grating sensing array and a mechanical sensor on the surface or in the component, and synchronously collecting an input load and a dynamic strain response signal; establishing a reference transfer function of each sensing point in a healthy state; during online service, calculating a real-time transfer function of each point and a distortion amount of the real-time transfer function relative to a reference; generating a continuous distortion field through spatial interpolation based on the spatial coordinates and the distortion of all the sensing points, and calculating a spatial gradient field of the continuous distortion field; and finally, segmenting the gradient field through a self-adaptive threshold value, and extracting a defect contour. According to the method, the variable service load is converted into effective excitation, defect positioning is achieved by analyzing the distortion gradient of the transfer function, the method is particularly suitable for curve fiber layer components with complex anisotropic characteristics, and online, real-time and accurate defect positioning under the real working condition is achieved.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing and measurement technology for materials, and in particular to a method for locating embedded defects in curved fiber variable angle fiber-lay components under different working conditions. Background Technology

[0002] Fiber-reinforced composite materials, especially curved fiber variable-angle fiber-lay components manufactured using automated fiber placement technology, have become a key material system for aerospace main load-bearing structures due to their excellent designability and lightweight potential. However, these components are prone to embedded defects such as delamination and porosity during manufacturing and service, seriously threatening structural safety. Therefore, developing reliable embedded defect detection and location technologies is crucial.

[0003] In existing technologies, defect detection for such components mainly relies on offline non-destructive testing methods such as ultrasonic C-scanning and X-rays. These methods require disassembling the component from the assembly and performing the tests on specific equipment, making it impossible to achieve real-time health monitoring and early warning of the structure under actual service conditions. To achieve online monitoring, existing solutions often employ attaching piezoelectric sensor arrays to the structural surface and actively exciting ultrasonic guided waves, then analyzing the scattering and reflection signals of the guided waves to achieve damage diagnosis.

[0004] However, this method faces inherent bottlenecks when applied to curved fiber variable angle layup components: First, the continuous change in fiber direction within the component leads to strong spatial anisotropy of the material, resulting in complex guided wave propagation paths and wave velocity direction dependence. Defect scattering signals are severely distorted and masked, making accurate interpretation and location difficult. Second, and more importantly, actual engineering structures are always under dynamic conditions such as changing loads and temperatures. These changes in conditions themselves cause significant changes in the structural vibration response, generating strong background noise that severely drowns out and interferes with the weak signal changes caused by tiny defects. This leads to a sharp drop in the reliability of existing methods based on fixed excitation or static benchmark comparison under variable conditions. Summary of the Invention

[0005] This invention overcomes the shortcomings of the prior art and provides a method for locating embedded defects in curved fiber variable angle filament laying components under different working conditions.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions, comprising the following steps:

[0007] S1. On the surface or inside of the curved fiber variable angle filament layup component, a distributed fiber grating sensor array is deployed in a grid pattern or integrated along the critical path to synchronously measure the dynamic strain response signal of all sensing points of the curved fiber variable angle filament layup component in spatial distribution; at the same time, at least one mechanical sensor is installed at the load input point or known excitation source location of the curved fiber variable angle filament layup component to synchronously acquire the input load signal.

[0008] S2. When the curved fiber variable angle filament placement component is confirmed to be in a healthy and defect-free state, the curved fiber variable angle filament placement component is operated within the expected typical operating range, or a dynamic load covering the future service spectrum is applied through a test bench; based on S1, the dynamic strain response signal and input load signal of all sensing points are collected in real time, the dynamic strain response signal and input load signal of each sensing point are selected, the reference transfer function of each sensing point in the healthy and defect-free state is calculated through signal processing, and the health reference value of each sensing point is obtained;

[0009] S3. Based on the reference transfer function of each sensor point, the real-time frequency response value of each sensor point is calculated under the real-time time series, and the real-time frequency response value of each sensor point is compared with the health reference value to calculate the distortion variable.

[0010] S4. The distortion variable is used as the scalar field value of the position of each sensing point, and combined with the spatial coordinates of all sensing points, a continuous distortion variable field covering the entire curved fiber variable angle filament laying component is generated by the spatial difference algorithm; and spatial gradient calculation is performed on the continuous distortion variable field to calculate the gradient magnitude value of each sensing point.

[0011] S5. Set a gradient threshold, and perform binarization or contour tracking processing on the gradient magnitude to obtain a high gradient region corresponding to the location, size, and contour of the embedded defect.

[0012] In a preferred embodiment of the present invention, the signal processing involves calculating a reference transfer function between the dynamic strain response signal and the input load signal in the frequency domain, based on the dynamic strain response signal and the input load signal; wherein the reference transfer function is the ratio of the spectrum of the dynamic strain response signal to the spectrum of the input load signal; the reference transfer function is expressed in complex form and includes amplitude-frequency characteristics and phase-frequency characteristics; the calculated reference transfer function is stored as a health reference value for each sensing point in a healthy and defect-free state.

[0013] In a preferred embodiment of the present invention, for each sensing point, the difference between the real-time frequency response value and the health benchmark value in a preset key frequency band is calculated in the frequency domain; the distortion variable is obtained by calculating the norm of the difference.

[0014] In a preferred embodiment of the present invention, the calculation process of the distorted variable is as follows:

[0015] After taking the modulus of the complex difference between the real-time frequency response value and the health benchmark value at each frequency point within the preset key frequency band, and then performing integration or sum of squares operation, a non-negative scalar value representing the magnitude of the overall deviation is obtained, which is the distorted variable.

[0016] In a preferred embodiment of the present invention, the generation process of the continuous distorted field includes:

[0017] S401. Use the spatial coordinates of each sensing point and the corresponding distortion as input data;

[0018] S402. The input data is processed using a spatial interpolation algorithm to estimate the distortion value in the area on the surface of the curved fiber variable angle filament laying component where no sensing points are directly set.

[0019] S403. Construct a variogram model based on the spatial autocorrelation of the input data, and use the variogram model as a basis to perform optimal unbiased estimation of the distorted variable values ​​in the region, thereby generating the continuous distorted variable field that is continuously distributed in space.

[0020] In a preferred embodiment of the present invention, the calculation process of the gradient magnitude value of each sensing point includes:

[0021] S411. Perform two-dimensional spatial discrete differentiation on the continuous distorted field;

[0022] S412. At the location of each sensing point, calculate the first-order partial derivatives of the continuous distorted field along the two orthogonal coordinate directions;

[0023] S413. Construct a two-dimensional gradient vector from the two calculated partial derivative components.

[0024] S414. Calculate the magnitude of the two-dimensional gradient vector, which is the gradient magnitude of each sensing point, used to quantify the spatial rate of change of the distorted variable at each sensing point.

[0025] In a preferred embodiment of the present invention, the preset gradient threshold includes:

[0026] S501. Based on the gradient magnitude values ​​of all sensing points calculated in S4, calculate the statistical characteristic value of the gradient magnitude values; the statistical characteristic value is a linear combination of the mean and standard deviation of all gradient magnitude values.

[0027] S502, The result of the linear combination is used as the gradient threshold.

[0028] In a preferred embodiment of the present invention, the deployment principle of the distributed fiber Bragg grating sensing array is as follows:

[0029] S101. Based on the geometry of the curved fiber variable angle fiber placement component and the fiber placement path, determine the structural stress concentration area or defect-prone area as the critical path.

[0030] S102. The sensing points in the distributed fiber optic grating sensing array are arranged along the critical path at a preset interval, and the remaining sensing points are arranged in a regular grid in the main area of ​​the curved fiber variable angle filament laying component, so that the spatial distribution density of the sensing points matches the structural risk level.

[0031] In a preferred embodiment of the present invention, the process of obtaining the health benchmark value includes:

[0032] S201. During the operation of the curved fiber variable angle filament laying component under different typical working conditions, the dynamic strain response signal and the input load signal are repeatedly collected, and multiple reference transfer function samples are calculated.

[0033] S202. For each sensing point, perform statistical analysis on multiple reference transfer function samples under the same working conditions, and calculate the mean or median.

[0034] S203. The calculated mean or median value is used as the health benchmark value for each sensing point in the healthy and defect-free state corresponding to this working condition.

[0035] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0036] (1) By using the dynamic load that the curved fiber variable angle filament laying component is subjected to during service as the excitation source, there is no need to add additional external active excitation. Thus, the internal defects of the curved fiber variable angle filament laying component can be continuously monitored and located under the condition that the curved fiber variable angle filament laying component is working normally without interruption. This solves the problems of low efficiency and inability to warn of sudden damage in traditional detection methods, and realizes continuous monitoring of the curved fiber variable angle filament laying component.

[0037] (2) By calculating the distortion of the reference transfer function relative to the healthy reference under the same working conditions, the influence of load amplitude and frequency changes can be effectively removed in principle, thereby significantly improving the stability and reliability of defect feature extraction under complex and variable real working conditions.

[0038] (3) By calculating the spatial gradient of the distorted field, the defect is located, and the model is transformed from relying on a globally uniform material model or wave propagation law to the relative mutation of local properties. This avoids the problem of global anisotropy modeling caused by the curved fiber path, and is therefore more suitable for complex composite material components with continuously changing fiber paths.

[0039] (4) By spatial interpolation and gradient operation, the one-dimensional time-series signal is transformed into a two-dimensional gradient magnitude field binary image, which can then present the defect in a clear high gradient contour form. Furthermore, by adaptive threshold segmentation and contour tracking, the location, projected area, and geometric shape of the defect can be directly output, thus providing a reliable basis for subsequent safety assessment and maintenance decision-making. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart of a preferred embodiment of the positioning method of the present invention; Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0044] like Figure 1 As shown, the method for locating embedded defects in curved fiber variable angle fiber-lay components under different working conditions includes the following steps:

[0045] S1. On the surface or inside of the curved fiber variable angle filament layup component, a distributed fiber grating sensor array is deployed in a grid pattern or integrated along the critical path to synchronously measure the dynamic strain response signal of all sensing points of the curved fiber variable angle filament layup component in spatial distribution; at the same time, at least one mechanical sensor is installed at the load input point or known excitation source location of the curved fiber variable angle filament layup component to synchronously acquire the input load signal.

[0046] In one embodiment, the deployment principle of the distributed fiber Bragg grating sensor array is as follows:

[0047] S101. Based on the geometry of the curved fiber variable angle fiber placement component and the fiber placement path, determine the structural stress concentration area or defect-prone area as the critical path.

[0048] S102. Along the critical path, the sensing points in the distributed fiber grating sensing array are arranged at a preset interval, and the remaining sensing points are arranged in a regular grid in the main area of ​​the curved fiber variable angle filament laying component, so that the spatial distribution density of the sensing points matches the structural risk level.

[0049] It should be noted that the curved fiber variable angle layup component is an advanced composite material component manufactured using an automated layup process. Its core feature is that the laying direction of the reinforcing fiber bundles in each layer is not fixed, but continuously changes along the surface of the component according to a predetermined path. This allows for optimization of the fiber orientation based on the structural stress, thereby achieving superior mechanical properties compared to traditional straight layup.

[0050] A distributed fiber Bragg grating sensor array consists of one or more special optical fibers embedded with multiple fiber Bragg gratings. Each fiber Bragg grating is a miniature sensor (called a sensing point) sensitive to axial strain and temperature. These gratings are inscribed on the optical fiber at specific intervals and can be identified and read by the same demodulation device through multiplexing techniques (such as wavelength division multiplexing).

[0051] Dynamic strain response signal refers to the minute deformation (strain) that changes rapidly with time inside a curved fiber variable angle filament-lay component when it is subjected to changing service loads (such as motor loads and aerodynamic loads).

[0052] When the curved fiber layup component experiences strain, the period of the fiber gratings bonded or embedded within it changes, causing a shift in the center wavelength of the reflected light. By monitoring this wavelength shift in real time using a high-precision demodulator, the continuous signal of strain change over time at each grating location can be simultaneously calculated. This process follows the physical relationship below:

[0053]

[0054] In the formula, This indicates the offset of the center wavelength of the fiber Bragg grating; This is the initial center wavelength of the fiber grating; The effective elastic coefficient; The axial strain to be measured. This is achieved through measurement... The dynamic strain response signal can then be obtained.

[0055] The critical path analysis is mainly based on two types of analysis:

[0056] (1) Through computer finite element simulation, the stress of the curved fiber variable angle filament laying component under the design load is simulated, and the parts with significantly higher stress levels than the surrounding areas (such as the edge of the hole, the geometric change point, and the vicinity of the connection joint) are identified.

[0057] (2) Based on manufacturing experience and historical data, identify areas where defects are likely to occur in curved fiber variable angle filament layup components during the filament layup process or use, such as layup boundaries, areas with maximum fiber curvature, and laminate edges.

[0058] By treating the critical path as a key area for monitoring, a higher density of sensors is deployed to capture signals of minute defects that may appear in these high-risk locations.

[0059] A force sensor or accelerometer is installed at the load input point (such as the actuator connection or support point) to directly measure the input load signal acting on the component.

[0060] All fiber Bragg grating sensors and mechanical sensors must be triggered by the same clock source to ensure strict timestamp alignment of all channel data. This is fundamental for subsequent accurate frequency domain analysis and reference transfer function calculations; any time synchronization issues will lead to phase errors, severely impacting positioning accuracy.

[0061] S2. When the curved fiber variable angle filament placement component is confirmed to be in a healthy and defect-free state, the curved fiber variable angle filament placement component is operated within the expected typical operating range, or a dynamic load covering the future service spectrum is applied through a test bench; based on S1, the dynamic strain response signal and input load signal of all sensing points are collected in real time, the dynamic strain response signal and input load signal of each sensing point are selected, the reference transfer function of each sensing point in the healthy and defect-free state is calculated through signal processing, and the health reference value of each sensing point is obtained.

[0062] In one embodiment, signal processing involves calculating a reference transfer function between the dynamic strain response signal and the input load signal in the frequency domain based on the dynamic strain response signal and the input load signal; that is, based on the dynamic strain response signal and the input load signal acquired by S1, the time-domain signals within the same time period are subjected to frequency domain transformation to obtain their respective spectra, and then the ratio of the response spectrum to the load spectrum is calculated to obtain the transfer function.

[0063] The reference transfer function is the ratio of the spectrum of the dynamic strain response signal to the spectrum of the input load signal. The reference transfer function is expressed in complex form and includes amplitude-frequency characteristics and phase-frequency characteristics. The calculated reference transfer function is stored as the health reference value of each sensing point in a healthy and defect-free state.

[0064] For each sensing point, the difference between the real-time frequency response value and the health baseline value in the preset key frequency band is calculated in the frequency domain; the distortion variable is obtained by calculating the norm of the difference.

[0065] It should be further explained that the reference transfer function describes the inherent relationship between the input load and the dynamic strain response at a certain point of output of a curved fiber variable angle filament layup component.

[0066] For a given healthy structure, its reference transfer function is unique and stable under specific operating conditions. Once internal defects (such as delamination) occur, the local stiffness changes, and the reference transfer function at that sensing point will be distorted. Therefore, monitoring changes in the reference transfer function reveals changes in the internal state of the structure more fundamentally and stably than directly monitoring strain amplitude, and is insensitive to changes in the magnitude of the load itself.

[0067] The preset critical frequency band refers to one or more frequency ranges that are most sensitive to local stiffness changes caused by defects among all possible vibration frequencies of the structure.

[0068] Typically obtained through modal testing or finite element analysis of healthy structures, these frequencies usually refer to the first few dominant natural frequencies of the structure and their surrounding region. Focusing the analysis on this frequency band can effectively filter out high-frequency noise and low-frequency drift unrelated to structural damage, significantly improving the signal-to-noise ratio and detection sensitivity of defect features. This preset key frequency band can be expressed as:

[0069] ;

[0070] In the formula, These are the lower and upper limits of the key frequency band, respectively; This is a preset key frequency band.

[0071] The norm is a scalar value used to measure the overall difference between two functions (real-time frequency response value and health baseline value). Calculating the norm of the difference essentially combines the complex differences (both amplitude and phase differences) at all frequency points throughout the entire key frequency band into a number representing the total deviation.

[0072] The process of obtaining health benchmark values ​​includes:

[0073] S201. During the operation of the curved fiber variable angle filament laying component under different typical working conditions, the dynamic strain response signal and input load signal are repeatedly collected, and multiple reference transfer function samples are calculated.

[0074] S202. For each sensing point, perform statistical analysis on multiple reference transfer function samples under the same working conditions, and calculate the mean or median.

[0075] S203. The calculated mean or median value shall be used as the health baseline value for each sensing point in a healthy and defect-free state corresponding to this working condition.

[0076] In one embodiment, taking the calculation of the median as an example, if M measurements are performed on the i-th sensing point under a certain operating condition, a reference transfer function sample set is obtained. Then, within the preset key frequency band, calculate the difference norm between each sample and a reference function. Then these M scalars Sort by size and take the median as the health baseline value for that sensor point. .

[0077] S3. Based on the reference transfer function of each sensor point, calculate the real-time frequency response value of each sensor point under real-time time series, and compare the real-time frequency response value of each sensor point with the health reference value to calculate the distortion variable.

[0078] It should be noted that the real-time frequency response value is the transfer function measured at a certain sensing point under real-time load excitation of the structure at the current moment (or within the most recent very short time window). Its calculation method is exactly the same as that for calculating the reference transfer function in S2, but the data source is the service data that is currently in operation.

[0079] The acquisition of real-time frequency response values ​​adopts the same signal processing procedure as in S2 (such as fast Fourier transform), but calculations are performed on the load and strain signals acquired in real time within the sliding time window, thereby ensuring consistency with the health benchmark value in terms of calculation principle and making the two comparable.

[0080] In one embodiment, the calculation process for the distorted variable is as follows:

[0081] For real-time frequency response values With health benchmarks In the preset key frequency band After taking the modulus of the complex differences at each frequency point, and then performing integration or sum of squares, a non-negative scalar value representing the magnitude of the overall deviation is obtained, which is the distorted variable. .

[0082] It should be noted that distorted variables It is a single, non-negative numerical value that quantitatively characterizes the degree of deviation of the real-time structural dynamic characteristics at a given sensing point from its healthy dynamic characteristics. The larger the value, the more significant the change in structural properties (mainly local stiffness) at that location, and the higher the probability of a defect. It serves as the original data source for subsequent spatial defect localization.

[0083] The calculation process for the distorted variable specifically includes:

[0084] Calculate each discrete frequency point within the preset key frequency band. Complex differences and their moduli:

[0085] ;

[0086] ;

[0087] In the formula, The difference is a complex number. The modulus value; To ensure real-time frequency response values ​​within the preset key frequency band Complex differences at various frequency points within the range; The health benchmark value is set in the preset key frequency band. Complex differences at each frequency point within the range; Indicates taking the complex number The real part; Indicates taking the complex number The imaginary part.

[0088] By all Integrating the magnitudes at each frequency point yields the distorted variable:

[0089] ;

[0090] In the formula, Frequency resolution; It is a distorted variable; It is a complex number; For all frequency points; k is the frequency point of the k-th sensing point. Alternatively, the sum of squares can be used:

[0091] ;

[0092] In the formula, It is a negative number; It is a distorted variable; For all frequency points; k is the frequency point of the k-th sensing point. .

[0093] S4. The distortion variable is used as the scalar field value of the position of each sensing point. Combined with the spatial coordinates of all sensing points, a continuous distortion variable field covering the entire curved fiber variable angle filament laying component is generated by the spatial difference algorithm. The spatial gradient operation is performed on the continuous distortion variable field to calculate the gradient magnitude of each sensing point.

[0094] In one embodiment, the generation process of the continuous distorted field includes:

[0095] S401, Set the spatial coordinates of each sensing point and the corresponding distorted variables As input data;

[0096] S402. The input data is processed using a spatial interpolation algorithm to estimate the distortion value in the area on the surface of the curved fiber variable angle filament laying component where no sensing points are directly set.

[0097] S403. Construct a variogram model based on the spatial autocorrelation of the input data, and use the variogram model as a basis to perform optimal unbiased estimation of the regional distorted variable values, generating a continuous distorted variable field that is continuously distributed in space. .

[0098] In one embodiment, the variogram model is a data-driven neural network variogram model. The core of this model lies in using a deep neural network as a general function approximator to autonomously learn the complex nonlinear mapping relationship between spatial distance and semivariance from measured data. This replaces traditional models with pre-defined mathematical forms, thereby more accurately representing the anisotropic and non-stationary spatially correlated structures caused by curved fiber layups and complex damage modes. This provides a crucial foundation for subsequent Kriging interpolation to generate high-fidelity continuous fields.

[0099] The variogram model employs a multilayer perceptron architecture. The input layer receives a scalar Euclidean distance between sensor pairs after maximum normalization. The network contains four fully connected hidden layers with 64, 128, 64, and 32 neurons respectively, using dense connections between layers. Each hidden layer is followed by a batch normalization layer and a linear rectified unit activation function to accelerate training convergence and enhance the model's nonlinear expressive power. The output layer is a single neuron using the Softplus activation function. This ensures that the output value is always positive, satisfying the mathematical constraint of non-negativity of semivariance. The architecture achieves an end-to-end mapping from distance to semivariance by progressively condensing and refining features.

[0100] It needs to be explained that, This represents the weighted sum of the outputs of the neurons in the previous layer (i.e., the net input). This function is the Softplus activation function, a smooth approximation of the ReLU function. It ensures that the output value is always positive, perfectly meeting the mathematical constraint that the semivariance must be non-negative.

[0101] The training data for the variogram model comes directly from the discrete distorted variable field output by step S3. For models containing... Data set of sensor points A sample set is constructed by computing all unique pairs of points. Each sample consists of independent variables. and dependent variable Composition, co-generation of approximately 100 sample points. Data preprocessing includes... Divide by the global maximum distance ,Will Divide by the population variance of the distorted variable To achieve scale normalization, the preprocessed dataset was randomly divided into training, validation, and test sets in a 7:2:1 ratio.

[0102] Based on the above, further explanation is needed: Let x and y represent the Euclidean distance between the i-th and j-th sensing points; x and y are the spatial coordinates of the sensing points. Let represent the distorted variables at the i-th and j-th sensing points, respectively; This represents the experimental semivariance between point i and point j; This represents the maximum distance between all sensor point pairs, used for maximum value normalization of the distance. / Unify the input scale to The interval is used to improve the stability and convergence speed of neural network training; This represents the overall variance of the distorted variable D across all sensing points.

[0103] The training of the variogram model aims to minimize the prediction error. The loss function is defined as the sum of the mean squared error and the L2 regularization term: Batch size Regularization coefficient The optimizer uses an adaptive moment estimation algorithm, with its key parameter set as the first-order moment decay factor. Second-order moment attenuation factor Initial learning rate The training employs a dynamic learning rate scheduling strategy. When the validation set loss fails to decrease for five consecutive epochs, the learning rate is multiplied by a factor of 0.5. The maximum number of training epochs is set to 200, and an early stopping mechanism is used: if the validation loss fails to improve for 15 consecutive epochs, training is terminated to prevent overfitting.

[0104] Based on the above, further explanation is needed: , These represent the semivariance predicted by the neural network model and the true semivariance calculated from the data, respectively. Refers to all weight parameters in a neural network.

[0105] The calculation process for the gradient magnitude of each sensing point includes:

[0106] S411. Perform two-dimensional spatial discrete differentiation on a continuous distorted field;

[0107] S412, Location of each sensing point Calculate the first-order partial derivatives of the continuous distorted field along the two orthogonal coordinate directions respectively;

[0108] S413. Construct a two-dimensional gradient vector from the two calculated partial derivative components.

[0109] S414. Calculate the magnitude of the two-dimensional gradient vector. The magnitude is the gradient magnitude at each sensing point, which is used to quantify the spatial rate of change of the distorted variable at each sensing point.

[0110] In one embodiment, a continuous distorted field Represented on a regular grid. For grid points Its first-order partial derivatives along the x and y directions can be approximated using the central difference method:

[0111] ;

[0112] ;

[0113] In the formula, and These are continuous distorted variable fields At grid points At that point, the numerical approximations of the first-order partial derivatives along the x and y directions are: and These represent the grid spacing in the x and y directions, respectively; for each sensing point... Its partial derivative and It can be obtained from bilinear interpolation of its grid cell. and get; , For target grid points The distorted values ​​of adjacent right and left points in the x-direction, and adjacent upper and lower points in the y-direction.

[0114] The gradient vector is defined as: Gradient magnitude of the k-th sensing point The Euclidean norm of the gradient vector is:

[0115] .

[0116] In the formula, It is a two-dimensional vector, called the gradient vector at the k-th sensing point; and It is a continuous distorted variable field The true location of the k-th sensor point The values ​​of the first-order partial derivatives along the x and y directions at that point.

[0117] To explain further, due to Typically, since the point is not located on a regular grid, its derivative cannot be directly calculated using the finite difference formula. It requires bilinear interpolation to obtain the derivative from the four corner points of the grid cell containing that point, which have already been calculated. and The value is obtained by difference estimation.

[0118] S5. Preset gradient threshold, perform binarization or contour tracking on the gradient magnitude to obtain the high gradient region corresponding to the location, size and contour of the embedded defect.

[0119] Furthermore, the preset gradient threshold includes:

[0120] S501. Based on the gradient magnitude values ​​of all sensing points calculated by S4, calculate the statistical characteristic value of the gradient magnitude values; the statistical characteristic value is a linear combination of the mean and standard deviation of all gradient magnitude values.

[0121] S502, Use the result of the linear combination as the gradient threshold.

[0122] In one embodiment, the mean of the gradient magnitudes of all N sensing points is calculated. and standard deviation :

[0123] ;

[0124] ;

[0125] In the formula, The mean; is the standard deviation; N is the total number of sensing points; k is the k-th sensing point; It is the gradient magnitude of the k-th sensing point, and it is the gradient vector. The Euclidean norm.

[0126] To further explain, It is a non-negative scalar value that discards directional information and only quantifies the degree of drastic change in the total space of the distortion at the sensing point. Furthermore, at the defect boundary, due to abrupt changes in material properties, the distortion changes drastically, thus the gradient magnitude... Significant peaks will appear, which will ultimately serve as the direct input signal for subsequent threshold segmentation to identify and delineate defect contours.

[0127] Then, the mean and standard deviation are linearly combined to obtain the gradient threshold T:

[0128] ;

[0129] in, A positive constant coefficient is used to control the strictness of the threshold; T is the gradient threshold. The mean; The standard deviation is denoted as .

[0130] Finally, the gradient magnitude field is binarized using a gradient threshold to generate a binary image. :

[0131] ;

[0132] in, It is a binary image; For the gradient magnitude The interpolated continuous gradient field. Connected regions with a value of 1 represent high-gradient regions corresponding to the location, size, and contour of the embedded defect. The contour tracking algorithm can be run directly on this binary image to extract the boundaries of each connected region as the defect contour.

[0133] Specifically The gradient magnitude at each discrete sensing point is calculated in step S4. After a spatial interpolation process similar to that generated in step S4, a continuous function covering the entire surface of the curved fiber variable angle filament layup component is obtained.

[0134] T is the gradient threshold, a scalar criterion calibrated based on statistical laws, calculated in steps S501-S502. This gradient threshold is the boundary distinguishing between background noise / normal changes and suspected defect edges. It is adaptively determined based on the actual distribution of the real-time gradient field in each monitoring session, with the aim of ensuring the accuracy of monitoring different curved fiber angle layup components and different degrees of damage.

[0135] If the gradient value at a certain sensing point If so, the sensor point is determined to be a foreground point. .

[0136] If the gradient value at a certain sensing point If the sensor point is not identified as a background point, then the sensor point is determined to be a background point. .

[0137] In simple terms, this judgment operation is image segmentation, which transforms the surface of the curved fiber-angled wire-laying component into a black-and-white image. The set of white pixels with a value of 1 constitutes the high-gradient region for preliminary judgment, which is the defect region.

[0138] The outer boundary of the high-gradient region can be extracted using contour tracing algorithms (such as Moore's neighborhood tracing, Freeman chain code, etc.), thus obtaining a set of ordered boundary point coordinates. This set of coordinates precisely defines the geometric contour of the defect.

[0139] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for locating embedded defects in curved fiber variable angle fiber-lay components under different working conditions, characterized in that, Includes the following steps: S1. On the surface or inside of the curved fiber variable angle filament layup component, a distributed fiber grating sensor array is deployed in a grid pattern or integrated along the critical path to synchronously measure the dynamic strain response signal of all sensing points of the curved fiber variable angle filament layup component in spatial distribution; at the same time, at least one mechanical sensor is installed at the load input point or known excitation source location of the curved fiber variable angle filament layup component to synchronously acquire the input load signal. S2. When the curved fiber variable angle filament placement component is confirmed to be in a healthy and defect-free state, the curved fiber variable angle filament placement component is operated within the expected typical operating range, or a dynamic load covering the future service spectrum is applied through a test bench. Based on the real-time acquisition of the dynamic strain response signal and input load signal of all sensing points by S1, the dynamic strain response signal and input load signal of each sensing point are selected, and the reference transfer function of each sensing point in a healthy and defect-free state is calculated through signal processing, and the health reference value of each sensing point is obtained. S3. Based on the reference transfer function of each sensor point, the real-time frequency response value of each sensor point is calculated under the real-time time series, and the real-time frequency response value of each sensor point is compared with the health reference value to calculate the distortion variable. S4. The distortion variable is used as the scalar field value of the position of each sensing point, and combined with the spatial coordinates of all sensing points, a continuous distortion variable field covering the entire curved fiber variable angle filament laying component is generated by the spatial difference algorithm; and spatial gradient calculation is performed on the continuous distortion variable field to calculate the gradient magnitude value of each sensing point. S5. Set a gradient threshold, and perform binarization or contour tracking processing on the gradient magnitude to obtain a high gradient region corresponding to the location, size, and contour of the embedded defect.

2. The method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions according to claim 1, characterized in that: The signal processing involves calculating a reference transfer function between the dynamic strain response signal and the input load signal in the frequency domain, based on the dynamic strain response signal and the input load signal. The reference transfer function is the ratio of the spectrum of the dynamic strain response signal to the spectrum of the input load signal. The reference transfer function is expressed in complex form and includes amplitude-frequency and phase-frequency characteristics. The calculated reference transfer function is stored as a health reference value for each sensing point under a healthy, defect-free state.

3. The method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions according to claim 2, characterized in that: For each sensing point, the difference between the real-time frequency response value and the health benchmark value in the preset key frequency band is calculated in the frequency domain; the distortion variable is obtained by calculating the norm of the difference.

4. The method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions as described in claim 3, characterized in that: The calculation process for the distorted variable is as follows: After taking the modulus of the complex difference between the real-time frequency response value and the health benchmark value at each frequency point within the preset key frequency band, and then performing integration or sum of squares operation, a non-negative scalar value representing the magnitude of the overall deviation is obtained, which is the distorted variable.

5. The method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions according to claim 4, characterized in that: The generation process of the continuous distorted field includes: S401. Use the spatial coordinates of each sensing point and the corresponding distortion as input data; S402. The input data is processed using a spatial interpolation algorithm to estimate the distortion value in the area on the surface of the curved fiber variable angle filament laying component where no sensing points are directly set. S403. Construct a variogram model based on the spatial autocorrelation of the input data, and use the variogram model as a basis to perform optimal unbiased estimation of the distorted variable values ​​in the region, thereby generating the continuous distorted variable field that is continuously distributed in space.

6. The method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions according to claim 5, characterized in that: The calculation process for the gradient magnitude at each sensing point includes: S411. Perform two-dimensional spatial discrete differentiation on the continuous distorted field; S412. At the location of each sensing point, calculate the first-order partial derivatives of the continuous distorted field along the two orthogonal coordinate directions; S413. Construct a two-dimensional gradient vector from the two calculated partial derivative components. S414. Calculate the magnitude of the two-dimensional gradient vector, which is the gradient magnitude of each sensing point, used to quantify the spatial rate of change of the distorted variable at each sensing point.

7. The method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions according to claim 1, characterized in that: The preset gradient threshold includes: S501. Based on the gradient magnitude values ​​of all sensing points calculated in S4, calculate the statistical characteristic value of the gradient magnitude values; the statistical characteristic value is a linear combination of the mean and standard deviation of all gradient magnitude values. S502, The result of the linear combination is used as the gradient threshold.

8. The method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions according to claim 1, characterized in that: The deployment principle of the distributed fiber Bragg grating sensing array is as follows: S101. Based on the geometry of the curved fiber variable angle fiber placement component and the fiber placement path, determine the structural stress concentration area or defect-prone area as the critical path. S102. The sensing points in the distributed fiber optic grating sensing array are arranged along the critical path at a preset interval, and the remaining sensing points are arranged in a regular grid in the main area of ​​the curved fiber variable angle filament laying component, so that the spatial distribution density of the sensing points matches the structural risk level.

9. The method for locating embedded defects in curved fiber variable angle fiber-laying components under different working conditions according to claim 2, characterized in that: The process of obtaining the health benchmark values ​​includes: S201. During the operation of the curved fiber variable angle filament laying component under different typical working conditions, the dynamic strain response signal and the input load signal are repeatedly collected, and multiple reference transfer function samples are calculated. S202. For each sensing point, perform statistical analysis on multiple reference transfer function samples under the same working conditions, and calculate the mean or median. S203. The calculated mean or median value is used as the health benchmark value for each sensing point in the healthy and defect-free state corresponding to this working condition.

Images