A Smart Detection Method for Strength Uniformity of Ultrafine Polyester Low-Elasticity Yarn

Abstract

This invention discloses an intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn, belonging to the field of intelligent detection technology. The method collects multiple sets of transient excitation response signals under multiple preset tension gradients along the axial direction of the ultrafine polyester low-elasticity yarn under test, constructing a spatiotemporal hybrid response matrix. Adaptive variational mode decomposition is performed on the spatiotemporal hybrid response matrix to separate the eigenmode function components characterizing fiber slip behavior and fracture energy level distribution. A three-dimensional phase space topology map reflecting local mechanical property changes is constructed in the reconstructed phase space. The curvature change trajectory of the high-dimensional manifold in this topology map is tracked to extract the singular attractor feature quantity characterizing local strength defects. Based on the fluctuation distribution of the Lyapunov exponent spectrum of the singular attractor feature quantity along the yarn length direction, the strength uniformity detection result is output. This method can effectively identify strength non-uniformity defects caused by fiber slippage and micro-fracture inside ultrafine polyester low-elasticity yarn.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology, specifically to an intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn. Background Technology

[0002] In textile processing, ultrafine polyester low-elasticity yarn needs to withstand repeated stretching, twisting, and weaving loads. The uniformity of its strength along its length directly determines the quality and scrap rate of the subsequent fabric. Current strength uniformity testing mainly relies on constant-rate tensile testing or elongation at break under constant load, obtaining breaking strength data and calculating the coefficient of variation through segmented sampling. These methods can only obtain the statistical distribution of macroscopic breaking strength and cannot capture the accumulation and abrupt changes of microscopic damage such as internal fiber slippage and microfractures along the yarn length. For microscopic mechanical property inhomogeneities caused by differences in fiber orientation and local crystallinity fluctuations, the force-displacement curves of conventional tensile tests are insufficient to provide spatially resolved anomaly identification, and segmented sampling disrupts the continuity of the yarn, resulting in low testing efficiency.

[0003] In dynamic mechanical testing, acoustic emission (AE) or vibration modal analysis (EMA) techniques have been used for defect detection in continuous fiber materials. AE methods locate damage events by capturing transient elastic waves released when fibers break, but they are insensitive to early damage such as fiber slippage and interfacial debonding before significant fracture. Vibration detection based on natural frequency shifts responds to macroscopic stiffness changes to some extent, but it cannot separate modal components corresponding to various dissipation mechanisms within the fiber assembly. It is also difficult to distinguish signal differences caused by tension fluctuations, environmental noise, and actual structural defects, resulting in a lack of reliable dynamic basis for the microscopic characterization of strength uniformity.

[0004] The strength uniformity test of ultrafine polyester low elastic yarn needs to solve two problems: first, how to separate the weak signal component that is only related to fiber slip behavior and fracture energy level distribution from the overall vibration response under multi-tension state, and eliminate background noise and macro-tension coupling interference; second, how to extract the feature quantity that can locate and quantify micro-strength defects along the yarn length direction from the separated dynamic response component, so as to achieve continuous and non-destructive uniformity discrimination. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide an intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn. By acquiring transient excitation signals under multiple tension gradients and adaptive variational mode decomposition, the modal components characterizing the energy level distribution of fiber slip and fracture are separated. Furthermore, the singular attractor features and Lyapunov exponent spectrum are extracted in the reconstructed phase space, enabling non-destructive continuous detection and localization of local strength defects along the yarn length direction. This overcomes the shortcomings of traditional methods that cannot distinguish microscopic damage inside the fiber and cannot continuously characterize strength uniformity.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn, the method comprising: Multiple sets of transient excitation response signals of the ultrafine polyester low-elasticity yarn under multiple preset tension gradients are collected along the axial direction of the yarn to be tested, forming a spatiotemporal hybrid response matrix. Adaptive variational mode decomposition is performed on the spatiotemporal hybrid response matrix to separate the intrinsic mode function components characterizing the fiber slip behavior and fracture energy level distribution within the ultrafine polyester low-elasticity yarn. Based on the separated intrinsic mode function components, a three-dimensional phase space topology map reflecting the local mechanical property changes of the ultrafine polyester low-elasticity yarn is constructed in the reconstructed phase space. The curvature change trajectory of the high-dimensional manifold is traced in the three-dimensional phase space topology map to extract the singular attractor feature quantity characterizing the local strength defects of the ultrafine polyester low-elasticity yarn. Based on the fluctuation distribution of the Lyapunov exponent spectrum of the singular attractor feature quantity along the length direction of the ultrafine polyester low-elasticity yarn, the strength uniformity detection result of the ultrafine polyester low-elasticity yarn is output.

[0007] As a technical solution of the present invention, multiple sets of transient excitation response signals of the ultrafine polyester low-elasticity yarn under multiple preset tension gradients are collected along the axial direction of the yarn to be tested, forming a spatiotemporal hybrid response matrix. Specifically, the method includes: fixing the ultrafine polyester low-elasticity yarn on an adjustable tension clamp, arranging multiple piezoelectric accelerometers axially at equal intervals; sequentially adjusting the tension of the ultrafine polyester low-elasticity yarn to preset low tension, medium tension, and high tension values, applying pulse excitation to the same position of the ultrafine polyester low-elasticity yarn using an electromagnetic vibrator at each tension value; synchronously collecting the time-domain voltage signals output by the multiple piezoelectric accelerometers, arranging all time-domain voltage signals under the same tension value into a column vector according to the spatial position of the sensors, and splicing the column vectors corresponding to different tension values ​​in order of tension gradient to form the spatiotemporal hybrid response matrix. By using multi-level tension progression and multi-channel spatial synchronous acquisition, the transient dynamic behavior of fiber assemblies under different stress stages can be fully excited and captured, so that the weak characteristic information reflecting the internal fiber slip and fracture energy level distribution can be completely preserved in the hybrid response matrix.

[0008] Preferably, the axial spacing of the plurality of piezoelectric accelerometers is 5mm to 10mm to achieve an optimized balance between spatial resolution and signal-to-noise ratio, ensuring that sufficiently fine local mechanical response differences are captured along the filament axis.

[0009] As a technical solution of the present invention, adaptive variational mode decomposition is performed on the spatiotemporal hybrid response matrix to separate the intrinsic mode function components characterizing the internal fiber slip behavior and fracture energy level distribution of the ultrafine polyester low elastic yarn. Specifically, this includes: taking each column vector of the spatiotemporal hybrid response matrix as the signal to be decomposed, initializing the center frequency and bandwidth constraint parameters of the intrinsic mode function components of each signal to be decomposed; iteratively updating the frequency domain representation of the intrinsic mode function components of each signal to be decomposed using the alternating direction multiplier method, calculating the frequency domain waveform similarity of the mode function components of adjacent two iterations after each iteration; stopping the iteration when the frequency domain waveform similarity is lower than a preset threshold, sorting all converged intrinsic mode function components from low to high center frequency, removing components with center frequencies lower than the low-frequency noise threshold, and retaining the remaining intrinsic mode function components as effective mode components. By employing a dual convergence criterion of adaptive iteration and waveform similarity, the dynamic modes characterizing fiber slippage and fracture can be accurately separated from the noise substrate, avoiding mode aliasing and ensuring that the extracted intrinsic mode function components accurately correspond to the essential physical process of strength degradation.

[0010] Further preferably, the frequency domain waveform similarity of the modal function components in two adjacent iterations is calculated by performing normalized cross-correlation on the frequency domain amplitude sequences output from two adjacent iterations, and the peak value of the normalized cross-correlation is used as the frequency domain waveform similarity. This method can sensitively capture the modal convergence process and improve the stability and accuracy of the decomposition iteration.

[0011] As a technical solution of the present invention, based on the separated intrinsic mode function components, a three-dimensional phase space topology map reflecting the local mechanical property changes of the ultrafine polyester low-elasticity yarn is constructed in the reconstructed phase space. Specifically, this includes: arranging each effective mode component into a time series in chronological order; performing delayed embedding reconstruction on the time series; selecting a delay time and embedding dimension such that the proportion of pseudo-nearest neighbor points in the reconstructed trajectory is lower than a preset threshold; fixing the embedding dimension to three dimensions; calculating the coordinate vector of each time point in the three-dimensional reconstructed phase space to obtain a three-dimensional point cloud set composed of all coordinate vectors; calculating the Euclidean distance between each point in the three-dimensional point cloud set and its neighboring points; constructing a three-dimensional curved surface mesh based on the distribution of neighboring distances of all points; and using the three-dimensional curved surface mesh as the three-dimensional phase space topology map. Through phase space reconstruction and surface approximation, the spatial mechanical property differences of the fiber bundle along the axial direction are mapped to the geometric deformation of the three-dimensional topological surface, making local strength defects visually presented as distorted forms of the phase space manifold, significantly amplifying weak anomaly features.

[0012] As a technical solution of the present invention, the curvature change trajectory of the high-dimensional manifold is traced in the three-dimensional phase space topology diagram to extract singular attractor feature quantities for characterizing local strength defects of the ultrafine polyester low-elasticity yarn. Specifically, this includes: discretizing the three-dimensional phase space topology diagram into multiple local grid cells; calculating the principal curvature and Gaussian curvature values ​​of all points within each local grid cell; arranging the principal curvature and Gaussian curvature values ​​of each local grid cell into a curvature change sequence along the axial order of the ultrafine polyester low-elasticity yarn; identifying anomalous grid cells with abrupt curvature value changes in the curvature change sequence; and using the three-dimensional phase space coordinates corresponding to the anomalous grid cells and the magnitude of their curvature abrupt changes as the singular attractor feature quantities. This method uses differential geometric invariants as a descriptive tool and locates the abrupt changes in fiber slip instability or fracture energy levels by tracing singular jumps in manifold curvature, thus achieving a quantitative nonlinear dynamic characterization of local strength defects in ultrafine polyester low-elasticity yarn.

[0013] Preferably, the anomalous mesh cells with abrupt curvature changes are identified by calculating that the difference between the Gaussian curvature value of the local mesh cell and the mean Gaussian curvature value of the adjacent mesh cells exceeds three standard deviations. This identification strategy can automatically adapt to the slow drift of the background mechanical response along the wire axis and suppress the interference of the global trend term, making the identification of local strength defects more adaptive and sensitive.

[0014] As a technical solution of the present invention, along the axial order of the ultrafine polyester low-elasticity yarn, the principal curvature values ​​and Gaussian curvature values ​​of each local grid unit are arranged into a curvature change sequence. Specifically, this includes: sorting all local grid units according to the physical position coordinates of each local grid unit on the axial direction of the ultrafine polyester low-elasticity yarn to obtain an ordered grid unit list; extracting the principal curvature value of each local grid unit from the ordered grid unit list to form a principal curvature value sequence, extracting the Gaussian curvature value to form a Gaussian curvature value sequence, and using the principal curvature value sequence and the Gaussian curvature value sequence together as the curvature change sequence. By strictly corresponding the phase space geometric features with the actual axial position of the yarn, the subsequently extracted singular attractor features are ensured to have accurate spatial positioning capabilities.

[0015] As a technical solution of the present invention, based on the fluctuation distribution of the Lyapunov exponent spectrum of the singular attractor characteristic quantity along the length direction of the ultrafine polyester low-elasticity yarn, the strength uniformity detection result of the ultrafine polyester low-elasticity yarn is output. Specifically, this includes: using the three-dimensional phase space coordinates corresponding to each abnormal grid unit as the initial trajectory point, tracking the neighboring trajectories in the three-dimensional phase space topology map, calculating the maximum Lyapunov exponent and the convergence value of the maximum Lyapunov exponent at each abnormal grid unit; arranging the convergence values ​​of the maximum Lyapunov exponent according to the corresponding positions of the abnormal grid units along the axial direction of the ultrafine polyester low-elasticity yarn to form a one-dimensional exponent spectrum curve; performing a difference operation on the one-dimensional exponent spectrum curve, marking the positions where the difference value exceeds a preset variation threshold as strength non-uniformity points, and outputting a detection spectrum marked with the coordinates of all strength non-uniformity points. By employing a quantitative index in chaotic dynamics that is sensitive to initial conditions, the microscopic strength fluctuations inside the fiber bundle are transformed into macroscopic fluctuations of the exponent spectrum curve, obtaining a high spatial resolution strength defect distribution spectrum along the length direction without damaging the sample.

[0016] Preferably, the maximum Lyapunov exponent is calculated using a small data set method by tracking the average divergence rate of the nearest neighbor points in the phase space. This small data set method can robustly estimate the exponent convergence value under small sample phase trajectory conditions, adapting to the relatively sparse data points within local anomaly grid cells of ultrafine polyester low-elasticity yarn.

[0017] As a further preferred embodiment, the convergence value of the maximum Lyapunov exponent is arranged according to the corresponding position of the anomalous grid cell along the axial direction of the ultrafine polyester low-elasticity yarn to form a one-dimensional exponential spectrum curve. Specifically, this includes: establishing a mapping relationship between each anomalous grid cell and the axial position of the ultrafine polyester low-elasticity yarn, and obtaining the axial distance coordinates corresponding to each anomalous grid cell; plotting discrete points in a two-dimensional coordinate system with the axial distance coordinates as the horizontal axis and the convergence value of the maximum Lyapunov exponent as the vertical axis; performing curve fitting on the discrete points using cubic spline interpolation, and using the resulting continuous curve as the one-dimensional exponential spectrum curve. Through this continuous processing, the exponential spectrum curve can smoothly reflect the evolution trend of intensity uniformity along the axial direction, facilitating subsequent differential analysis and precise location of anomalous positions.

[0018] As a further preferred approach, a difference operation is performed on the one-dimensional exponential spectrum curve. Positions where the difference value exceeds a preset variation threshold are marked as intensity non-uniformity points. A detection map is output, marked with the coordinates of all intensity non-uniformity points. Specifically, this includes: calculating a forward difference value for each curve point on the one-dimensional exponential spectrum curve, i.e., the difference between the ordinate value of the current curve point and the ordinate value of the previous curve point; comparing the absolute value of the forward difference value with the preset variation threshold; when the absolute value of the forward difference value is greater than the preset variation threshold, recording the axial distance coordinate corresponding to the current curve point as a candidate non-uniformity position; sorting all candidate non-uniformity positions in ascending order of axial distance coordinates, removing duplicate candidate points with a spacing smaller than a preset minimum defect spacing, and using the remaining candidate points as the intensity non-uniformity points; marking the coordinates of the intensity non-uniformity points on the axial schematic diagram of the ultrafine polyester low-elasticity yarn to generate the detection map. This method effectively suppresses false markings caused by trajectory tracking errors or interpolation oscillations, ensuring that the final output intensity non-uniformity points are all physically meaningful real defect positions, achieving highly reliable and comprehensive detection.

[0019] Through the above technical solution, the present invention transforms the internal fiber slippage and fracture precursors that are difficult to capture in the traditional uniformity measurement of ultrafine polyester low elastic yarn into quantitative changes in phase space manifold geometry and chaotic index spectrum. It obtains high-precision strength defect distribution along the axial direction under non-destructive conditions, and the detection process is highly intelligent and the results are intuitive and reliable.

[0020] The technical effects and advantages provided by the present invention in the above technical solution are as follows: By collecting multiple sets of transient excitation response signals under preset tension gradients along the axial direction of ultrafine polyester low-elasticity yarn and constructing a spatiotemporal hybrid response matrix, the dynamic response of the yarn under different static load backgrounds is organized into structured data containing both spatial position and tension state dimensions. This allows for the synchronous recording of slip and micro-fracture information activated by tension level changes within the fiber bundle through multiple channels. Adaptive variational mode decomposition iteratively separates each column of this matrix in the frequency domain, and after each iteration, convergence is controlled by the frequency domain waveform similarity of adjacent modal components, retaining only modal components with center frequencies higher than the low-frequency noise threshold. The eigenmode function components separated in this way directly correspond to the dynamic characteristics excited by interfiber frictional slip and local fracture events, avoiding the defects of traditional filtering or fixed-scale decomposition methods that submerge the slip response in fundamental frequency vibration or noise. The extracted effective modal components have high purity in the frequency domain and strong characterization power for fiber slip behavior and fracture energy level distribution.

[0021] Based on the reconstruction of a three-dimensional phase space topology map using effective modal components, delayed embedding is used to map the time-domain signal into a high-dimensional manifold in phase space. Local folding and stretching of the manifold are tracked through changes in the principal curvature and Gaussian curvature of discrete grid cells. A curvature change sequence is constructed along the fiber axis, and the deviation between the Gaussian curvature of local grid cells and the mean of adjacent regions is calculated. Anomalous grid cells with abrupt curvature changes are identified as singular attractor features. This method of capturing geometrical abrupt changes in dynamic trajectories in phase space can amplify abrupt changes in local mechanical properties caused by fiber slip instability or microfractures into significant changes in manifold curvature, allowing early strength defects that are difficult to distinguish in the original signal amplitude to be sensitively captured in the phase space topology. Furthermore, the maximum Lyapunov exponent is calculated using anomalous cells as initial trajectory points and arranged along the axis. Intensity inhomogeneities are marked by differential abrupt changes in the exponent spectrum, outputting a detection spectrum containing position coordinates. This process enables continuous scanning and axial positioning of microscopic strength defects. The detection results are not limited by the selection of segmented sampling positions and can reflect the degree of local dynamic instability at any position along the fiber bundle length. It provides non-destructive, high spatial resolution detection output for evaluating the strength uniformity of ultrafine polyester low-elasticity yarn. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0023] Figure 1 This is a flowchart of an intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn; Figure 2 This is a schematic diagram of the construction of the spatiotemporal hybrid response matrix for multi-tension gradient vibration testing of ultrafine polyester low-elasticity yarn; Figure 3 This is a flowchart of the singular attractor feature extraction based on the curvature of local grid cells. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0025] See Figure 1This invention provides an intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn. Multiple sets of transient excitation response signals of the ultrafine polyester low-elasticity yarn under multiple preset tension gradients are collected along the axial direction of the yarn to be tested, forming a spatiotemporal hybrid response matrix. Adaptive variational mode decomposition is performed on the spatiotemporal hybrid response matrix to separate the intrinsic mode function components characterizing the fiber slip behavior and fracture energy level distribution within the ultrafine polyester low-elasticity yarn. Based on the separated intrinsic mode function components, a three-dimensional phase space topology map reflecting the local mechanical property changes of the ultrafine polyester low-elasticity yarn is constructed in the reconstructed phase space. The curvature change trajectory of the high-dimensional manifold is traced in the three-dimensional phase space topology map to extract singular attractor feature quantities characterizing local strength defects in the ultrafine polyester low-elasticity yarn. Based on the fluctuation distribution of the Lyapunov exponent spectrum of the singular attractor feature quantities along the length direction of the ultrafine polyester low-elasticity yarn, the strength uniformity detection result of the ultrafine polyester low-elasticity yarn is output.

[0026] In specific implementation, please refer to Figure 2 One end of a microfiber polyester low-elasticity yarn is fixed to the fixed clamp of an adjustable tension clamp, and the other end is fixed to the movable clamp of the same clamp. The axial tension applied to the microfiber polyester low-elasticity yarn is changed by adjusting the position of the movable clamp. Multiple piezoelectric accelerometers are arranged along the axial direction of the microfiber polyester low-elasticity yarn, with equal spacing between adjacent sensors ranging from 5mm to 10mm. The sensitive axis of each piezoelectric accelerometer is perpendicular to the axial direction of the microfiber polyester low-elasticity yarn. The piezoelectric accelerometers are attached to the surface of the microfiber polyester low-elasticity yarn using miniature clamps, the mass of which is less than one percent of the mass per unit length of the microfiber polyester low-elasticity yarn. The tension of the microfiber polyester low-elasticity yarn is then sequentially adjusted to preset low, medium, and high tension values. The low tension value is set to 5% of the breaking strength of microfiber polyester low elastic yarn, the medium tension value is set to 15% of the breaking strength of microfiber polyester low elastic yarn, and the high tension value is set to 25% of the breaking strength of microfiber polyester low elastic yarn.

[0027] At each tension value, an electromagnetic vibrator applies pulse excitation to the same location on the ultrafine polyester low-elasticity yarn. The pulse width is 0.5 milliseconds, and the amplitude of the pulse excitation causes the ultrafine polyester low-elasticity yarn to produce micro-vibrations, the amplitude of which does not exceed one-tenth of the diameter of the ultrafine polyester low-elasticity yarn. The excitation head of the electromagnetic vibrator contacts the surface of the ultrafine polyester low-elasticity yarn, and the contact position is located at the mid-section of the axial direction of the ultrafine polyester low-elasticity yarn.

[0028] The time-domain voltage signals output by multiple piezoelectric accelerometers are acquired simultaneously. The sampling frequency of the time-domain voltage signals is set to five times the resonant frequency of the piezoelectric accelerometers. The sampling duration of the time-domain voltage signals covers the complete process of vibration decay of the ultrafine polyester low-elasticity yarn to the noise level after pulse excitation.

[0029] The time-domain voltage signals output by all piezoelectric accelerometers under the same tension value are arranged into a column vector according to the spatial position order of the piezoelectric accelerometers. The spatial position order is along the axis of the ultrafine polyester low elastic yarn from one end to the other.

[0030] The column vectors corresponding to low tension values, medium tension values, and high tension values ​​are concatenated according to the tension gradient order to form a spatiotemporal hybrid response matrix. Each row in the spatiotemporal hybrid response matrix corresponds to the time-domain voltage signal of a piezoelectric accelerometer under the three tension gradients, and each column in the spatiotemporal hybrid response matrix corresponds to the time-domain voltage signals of all piezoelectric accelerometers under the same tension gradient within the complete sampling time.

[0031] In the spatiotemporal hybrid response matrix, the position located at the... Line 1 The elements of a column are denoted as , Indicates the first The piezoelectric accelerometer was in the first... The time-domain voltage signal sequence obtained by downsampling a tension gradient. The value ranges from 1 to integers, This represents the total number of piezoelectric accelerometers. The value is determined by the length of the ultrafine polyester low-elasticity yarn and the spacing between adjacent piezoelectric accelerometers. The values ​​are 1, 2, and 3, which correspond to low tension, medium tension, and high tension values, respectively.

[0032] In practice, each column vector of the spatiotemporal hybrid response matrix is ​​extracted as an independent signal to be decomposed. The column vectors of the spatiotemporal hybrid response matrix correspond to the signal sequence formed by arranging the time-domain voltage signals of all piezoelectric accelerometers under the same tension gradient within the complete sampling time.

[0033] For each signal to be decomposed, the center frequency and bandwidth constraint parameters of the intrinsic mode function (IMF) components are initialized. The initial value of the center frequency is uniformly selected within the Nyquist frequency range of the signal to be decomposed, and the bandwidth constraint parameter is initialized to the reciprocal of the number of sampling points of the signal to be decomposed multiplied by the variance of the signal to be decomposed. The frequency domain representation of the IMF components of each signal to be decomposed is iteratively updated using the alternating direction multiplier method. In each iteration, the center frequency of the current IMF component is fixed, and the frequency domain amplitude of the IMF component is updated using Wiener filtering; the center frequency of the IMF component is recalculated based on the updated frequency domain amplitude, completing one update of the frequency domain representation.

[0034] After each iteration, the frequency domain amplitude sequences output from two adjacent iterations are subjected to normalized cross-correlation calculation. The peak value of the normalized cross-correlation calculation is used as the frequency domain waveform similarity. The frequency domain waveform similarity is calculated as follows: in, Indicates the first Frequency domain waveform similarity after the next iteration This is the iteration number. The value of is an integer greater than or equal to 2; Indicates the first The frequency domain magnitude sequence output in the nth iteration is at the 1st iteration. The amplitude at each frequency point For frequency point number, The value ranges from 1 to integers, The total number of frequency points contained in the frequency domain amplitude sequence. Determined by the frequency domain discretization accuracy; Indicates the first The frequency domain magnitude sequence output in the nth iteration is at the 1st iteration. The amplitude at each frequency point This is the frequency offset. The value of is made Get the integer offset of the maximum value. The search scope is arrive The integers between 0.95 and 0.95 are used. Iteration stops when the frequency domain waveform similarity falls below a preset threshold (0.95). All converged intrinsic mode function (IMF) components are sorted from low to high center frequency. The low-frequency noise threshold is set to 1.5 times the lower limit of the piezoelectric accelerometer's frequency response, which is determined by the sensor's factory parameters. IMF components with center frequencies below the low-frequency noise threshold are discarded, and the remaining IMF components are retained as valid modal components.

[0035] In the specific implementation, each effective modal component is arranged into a time series in chronological order, and each data point in the time series corresponds to the modal component amplitude at a sampling time. Delayed embedding reconstruction is then performed on the time series. When selecting the delay time, the mutual information method is used to calculate the mutual information between the time series and its delayed version. The delay value corresponding to the first point where the mutual information drops to a minimum is taken as the delay time. When selecting the embedding dimension, the calculation starts from embedding dimension 1 and increments incrementally. A reconstruction trajectory is constructed under each embedding dimension, and the Euclidean distance between each point in the reconstruction trajectory and its nearest neighbor is calculated. The Euclidean distance between the point and its nearest neighbor after increasing the embedding dimension by one dimension is also calculated. Points where the ratio of the two Euclidean distances exceeds a distance ratio threshold are identified as pseudo-nearest neighbors. The proportion of pseudo-nearest neighbors to all points in the reconstruction trajectory is taken as the pseudo-nearest neighbor ratio. When the pseudo-nearest neighbor ratio is lower than a preset ratio threshold, the current embedding dimension is determined as the final embedding dimension. The preset ratio threshold is 0.05. The embedding dimension is fixed at three dimensions. The three-dimensional embedding dimension is set as follows: when the embedding dimension determined by the proportion of pseudo-nearest neighbors is greater than three dimensions, the embedding dimension that makes the proportion of pseudo-nearest neighbors fall below 0.2 for the first time is selected and compared with three dimensions. If the proportion of pseudo-nearest neighbors corresponding to three dimensions is less than 0.2, then three dimensions are used as the fixed embedding dimension. When the embedding dimension determined by the proportion of pseudo-nearest neighbors is less than or equal to three dimensions, the embedding dimension is directly used for subsequent reconstruction.

[0036] In the 3D reconstructed phase space, the coordinate vector of each time point consists of the value at the current moment in the time series, the value after a delay of one time, and the value after a delay of two time. The coordinate vector in the 3D reconstructed phase space is calculated for each valid time point in the time series, resulting in a 3D point cloud set. In the 3D point cloud set, the Euclidean distance between each point and its neighboring points is calculated. A neighboring point is defined as a point whose Euclidean distance to that point is less than a neighboring distance threshold, which is one-tenth of the average Euclidean distance between all point pairs in the 3D point cloud set. The Euclidean distance is calculated as follows: in, Represents the 3D point cloud set. The point and the first Euclidean distance between points and These are all point number identifiers. The value ranges from 1 to integers, The value ranges from 1 to integers and Not equal to , This represents the total number of points in the 3D point cloud set. , , They represent the first The coordinate values ​​of a point on the three coordinate axes in the three-dimensional reconstructed phase space; , , They represent the first The coordinates of a point on the three coordinate axes in the three-dimensional reconstructed phase space.

[0037] A 3D curved surface mesh is constructed based on the proximity distribution of all points. The space containing the 3D point cloud is divided into hexahedral mesh cells, with the side length of each mesh cell being twice the proximity distance threshold. Within each mesh cell, the average coordinates of all points falling within the cell are calculated as the mesh node coordinates. Adjacent non-empty mesh nodes are connected to form triangular patches. All triangular patches together constitute the 3D curved surface mesh, which serves as a 3D phase space topology graph.

[0038] In specific implementation, please refer to Figure 3 The 3D phase space topology map is discretized into multiple local mesh cells. The local mesh cells are divided as follows: within the 3D space occupied by the 3D phase space topology map, equidistant divisions are made along the three coordinate axes with a fixed step size, which is one-fifth of the average Euclidean distance between all point pairs in the 3D point cloud set. Each local mesh cell is a cubic region, containing the vertices of the triangular facets in the 3D phase space topology map that fall within the cubic region. For each local mesh cell, the principal curvature and Gaussian curvature values ​​of all points within the local mesh cell are calculated. Within each local mesh cell, the coordinates of all points within the local mesh cell are selected to construct a local surface fitting equation, which adopts a quadratic surface form. in, The coordinates in the local surface fitting equation are... The surface height value at that location; This represents the coordinate variable in the first horizontal direction within the local coordinate system. The value is taken as the first horizontal coordinate of a point within a local grid cell; This represents the coordinate variable in the second horizontal direction within the local coordinate system. The value is taken as the second horizontal coordinate value of the point within the local grid cell; In the equation for fitting a quadratic surface The coefficient of the term, The solution is obtained through least squares fitting. The range of values ​​depends on the spatial distribution of points within the local grid cell; In the equation for fitting a quadratic surface The coefficient of the term, The solution is obtained through least squares fitting. The range of values ​​depends on the spatial distribution of points within the local grid cell; In the equation for fitting a quadratic surface The coefficient of the cross term, The solution is obtained through least squares fitting. The range of values ​​depends on the spatial distribution of points within the local grid cell; In the equation for fitting a quadratic surface The coefficient of the linear term, The solution is obtained by least squares fitting. In the equation for fitting a quadratic surface The coefficient of the linear term, The solution is obtained by least squares fitting. This represents the constant term in the equation for fitting the quadratic surface. The solution is obtained through least squares fitting.

[0039] The principal curvature values ​​and Gaussian curvature values ​​are calculated based on the coefficients of the local surface fitting equation. The principal curvature values ​​include the first principal curvature value and the second principal curvature value, which are the maximum and minimum values ​​of the normal curvature of the surface corresponding to the local surface fitting equation at a certain point; the Gaussian curvature value is the product of the first principal curvature value and the second principal curvature value.

[0040] Based on the physical position coordinates of each local grid cell along the axial direction of the ultrafine polyester low-elasticity yarn, all local grid cells are sorted. The method for determining the physical position coordinates of each local grid cell along the axial direction of the ultrafine polyester low-elasticity yarn is as follows: obtain the piezoelectric accelerometer numbers corresponding to all points within the local grid cell. The piezoelectric accelerometer numbers correspond to equally spaced positions along the axial direction of the ultrafine polyester low-elasticity yarn. Multiply the average value of the piezoelectric accelerometer numbers within the local grid cell by the sensor spacing to obtain the axial physical position coordinates. All local grid cells are sorted in ascending order of their axial physical position coordinates to obtain an ordered grid cell list. The principal curvature values ​​of each local grid cell are extracted from the ordered grid cell list to form a principal curvature value sequence. Each element in the principal curvature value sequence is the first principal curvature value of the corresponding local grid cell. The Gaussian curvature values ​​of each local grid cell are extracted from the ordered grid cell list to form a Gaussian curvature value sequence. The principal curvature value sequence and the Gaussian curvature value sequence are combined as the curvature change sequence.

[0041] Identify anomalous grid cells with abrupt curvature changes in curvature variation sequences. Calculate the difference between the Gaussian curvature value of each local grid cell and the mean Gaussian curvature value of its neighboring grid cells. Neighboring grid cells are defined as the two adjacent local grid cells in the ordered grid cell list, one before and one after the current local grid cell, for a total of four neighboring local grid cells. The mean Gaussian curvature value of the neighboring grid cells is the arithmetic mean of the Gaussian curvature values ​​of the four neighboring local grid cells. Calculate the standard deviation of the Gaussian curvature values ​​of the four neighboring local grid cells. When the difference between the Gaussian curvature value of a local grid cell and the mean Gaussian curvature value of its neighboring grid cells exceeds three times the standard deviation, the local grid cell is identified as an anomalous grid cell with abrupt curvature changes. Use the three-dimensional phase space coordinates corresponding to the anomalous grid cell and the curvature change amplitude of the anomalous grid cell as singular attractor characteristic quantities. The three-dimensional phase space coordinates corresponding to the anomalous grid cell are taken as the coordinates of the center point of the anomalous grid cell, which is the average of the start and end coordinates of the anomalous grid cell in the three coordinate axes. The curvature change amplitude is the absolute value of the difference between the Gaussian curvature value of the anomalous grid cell and the mean Gaussian curvature value of its neighboring grid cells.

[0042] In a specific implementation, all the singular attractor features output by the above embodiment are obtained. Each singular attractor feature includes the three-dimensional phase space coordinates corresponding to the anomalous mesh cell and the curvature abrupt change amplitude of the anomalous mesh cell.

[0043] The three-dimensional phase space coordinates corresponding to each anomalous grid cell are used as the initial trajectory point. For each initial trajectory point, L points with the smallest Euclidean distance to the initial trajectory point are determined in the three-dimensional phase space topology graph as nearest neighbors, where L is set as the integer part of the square root of the total number of points N in the three-dimensional point cloud set. Using the initial trajectory point and the L nearest neighbors as the starting point for tracking, the neighboring trajectories in the three-dimensional phase space topology graph are tracked. The tracking method for neighboring trajectories is as follows: starting from the tracking starting point, extend forward along the tangent direction of the manifold surface in the three-dimensional phase space topology graph by one evolution step, where the evolution step is one-tenth of the average Euclidean distance between all point pairs in the three-dimensional point cloud set, until a new trajectory point is reached; at the new trajectory point, the extension along the tangent direction of the manifold surface is repeated, and the number of repetitions is set as the upper limit of the evolution step, which is twice L.

[0044] The maximum Lyapunov exponent at each anomalous grid cell was calculated using a small data set method. During the tracking process, the Euclidean distances between corresponding points on all tracking trajectories and the initial trajectory point were recorded at each evolution step. The average of these Euclidean distances was calculated as the average divergence distance. A divergence curve was plotted with the evolution step count on the x-axis and the natural logarithm of the average divergence distance on the y-axis. A least-squares line was fitted to the linearly increasing portion of the divergence curve, and the slope of the fitted line was taken as the convergence value of the maximum Lyapunov exponent.

[0045] A mapping relationship between each anomalous mesh element and the axial position of the ultrafine polyester low-elasticity yarn is established. This mapping is determined by querying the piezoelectric accelerometer serial numbers corresponding to the vertices of the triangular facets in the 3D phase space topology diagram contained within the anomalous mesh element. Specifically, the serial numbers of the piezoelectric accelerometers corresponding to the source effective modal components of all triangular facet vertices within the anomalous mesh element are statistically analyzed, and the axial physical position coordinates corresponding to the most frequently occurring piezoelectric accelerometer serial number are taken as the axial distance coordinates corresponding to the anomalous mesh element. Discrete points are plotted in a 2D coordinate system with the axial distance coordinates as the horizontal axis and the convergence value of the maximum Lyapunov exponent as the vertical axis, with each discrete point corresponding to an anomalous mesh element. Cubic spline interpolation is used to fit the discrete points to a curve. The boundary conditions for cubic spline interpolation are set to natural boundary conditions, i.e., the second derivative of the fitted curve is zero at the first and last discrete points. The resulting continuous curve is used as a one-dimensional exponential spectrum curve.

[0046] The forward difference value is calculated for each point on the one-dimensional exponential spectrum curve. The points on the curve are obtained by sampling the curve at fixed intervals, where the fixed sampling interval is one-thousandth of the total span of the axial distance coordinate. The forward difference value is calculated as follows: in, Represents the first exponential curve on a one-dimensional exponential spectrum. Forward difference values ​​of each curve point; The curve point number, The value of is an integer greater than or equal to 2; Represents the first exponential curve on a one-dimensional exponential spectrum. The ordinate value of each curve point, i.e., the convergence value of the maximum Lyapunov exponent; Represents the first exponential curve on a one-dimensional exponential spectrum. The ordinate values ​​of the curve points.

[0047] The absolute value of the forward difference is compared with a preset variation threshold. The preset variation threshold is set based on the standard deviation of the ordinate values ​​of all curve points on the one-dimensional exponential spectrum curve, multiplied by a scaling factor of 3 to obtain the preset variation threshold. When the absolute value of the forward difference is greater than the preset variation threshold, the axial distance coordinate corresponding to the current curve point is recorded as a candidate non-uniform location. The specific operation of marking the location corresponding to the current curve point as a candidate non-uniform location includes: recording the axial distance coordinate value corresponding to the current curve point, and recording the absolute value of the forward difference as the variation intensity value of the candidate non-uniform location.

[0048] All candidate non-uniform locations are sorted in ascending order of axial distance coordinates. Duplicate candidate points with a spacing smaller than the preset minimum defect spacing are removed. The preset minimum defect spacing is set to 3 mm, based on approximately twice the diameter of microfiber polyester low-elasticity yarn. The diameter of microfiber polyester low-elasticity yarn ranges from 0.1 mm to 3 mm, and the upper limit of 3 mm is used as the preset minimum defect spacing to ensure that adjacent defects belong to different physical locations. The removal method is as follows: traverse the sorted list of candidate non-uniform locations. When the difference in axial distance coordinates between two adjacent candidate non-uniform locations is less than the preset minimum defect spacing, retain the candidate non-uniform location with the larger variation intensity value and remove the candidate non-uniform location with the smaller variation intensity value.

[0049] The remaining candidate points are designated as points of intensity non-uniformity. The coordinates of these points are marked on an axial schematic diagram of the ultrafine polyester low-elasticity yarn. This is done by drawing a perpendicular line to the axial direction at the corresponding axial distance coordinate on the schematic diagram, and then labeling the axial distance coordinate and variation intensity value of each point next to the line. The axial schematic diagram with the coordinates of all intensity non-uniformity points is then output as the detection spectrum.

[0050] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for intelligently detecting the strength uniformity of ultrafine polyester low-elasticity yarn, characterized in that, Includes the following steps: Multiple sets of transient excitation response signals of the ultrafine polyester low elastic yarn under multiple preset tension gradients are collected along the axial direction of the yarn to be tested, forming a spatiotemporal hybrid response matrix; Adaptive variational mode decomposition is performed on the spatiotemporal hybrid response matrix to separate the intrinsic mode function components that characterize the fiber slip behavior and fracture energy level distribution inside the ultrafine polyester low elasticity yarn; Based on the separated intrinsic mode function components, a three-dimensional phase space topology map reflecting the local mechanical property changes of the ultrafine polyester low elasticity yarn is constructed in the reconstructed phase space. Tracing the curvature change trajectory of the high-dimensional manifold in the three-dimensional phase space topology diagram, we extract the singular attractor feature quantity used to characterize the local strength defects of the ultrafine polyester low elasticity yarn. Based on the fluctuation distribution of the Lyapunov exponent spectrum of the strange attractor characteristic quantity along the length direction of the ultrafine polyester low elastic yarn, the strength uniformity test result of the ultrafine polyester low elastic yarn is output.

2. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 1, characterized in that, Multiple sets of transient excitation response signals of the ultrafine polyester low-elasticity yarn under multiple preset tension gradients are collected along the axial direction of the yarn to be tested, forming a spatiotemporal hybrid response matrix, specifically including: The ultrafine polyester low-elasticity yarn is fixed on an adjustable tension clamp, and multiple piezoelectric accelerometers are arranged axially at equal intervals. The tension of the ultrafine polyester low elastic yarn is sequentially adjusted to a preset low tension value, medium tension value and high tension value. At each tension value, an electromagnetic vibrator is used to apply pulse excitation to the same position of the ultrafine polyester low elastic yarn. The time-domain voltage signals output by the multiple piezoelectric accelerometers are acquired synchronously. All time-domain voltage signals under the same tension value are arranged into column vectors according to the spatial position of the sensors. The column vectors corresponding to different tension values ​​are spliced ​​together in order of tension gradient to form the spatiotemporal hybrid response matrix.

3. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 2, characterized in that, The axial spacing of the multiple piezoelectric accelerometers is 5mm to 10mm.

4. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 2, characterized in that, Adaptive variational mode decomposition is performed on the spatiotemporal hybrid response matrix to separate the eigenmode function components characterizing the fiber slip behavior and fracture energy level distribution within the ultrafine polyester low-elasticity yarn, specifically including: Each column vector of the spatiotemporal hybrid response matrix is ​​taken as the signal to be decomposed, and the center frequency and bandwidth constraint parameters of the eigenmode function components of each signal to be decomposed are initialized. The frequency domain representation of the intrinsic mode function component of each signal to be decomposed is updated iteratively by alternating direction multiplier method, and the frequency domain waveform similarity of the mode function components of two adjacent iterations is calculated after each iteration. When the frequency domain waveform similarity is lower than a preset threshold, the iteration stops. All converged intrinsic mode function components are sorted from low to high according to their center frequency. Components with center frequencies lower than the low-frequency noise threshold are removed, and the remaining intrinsic mode function components are retained as valid mode components.

5. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 4, characterized in that, Based on the separated intrinsic mode function components, a three-dimensional phase space topology map reflecting the local mechanical property changes of the ultrafine polyester low-elasticity yarn is constructed in the reconstructed phase space, specifically including: Each effective modal component is arranged into a time series in chronological order. The time series is then reconstructed by delay embedding. The delay time and embedding dimension are selected such that the proportion of pseudo-nearest neighbor points of the reconstructed trajectory is lower than a preset proportion threshold. By fixing the embedding dimension to three dimensions, the coordinate vector of each time point in the three-dimensional reconstructed phase space is calculated to obtain a three-dimensional point cloud set composed of all coordinate vectors. For each point in the three-dimensional point cloud set, calculate its Euclidean distance to neighboring points, construct a three-dimensional curved surface mesh based on the distribution of neighboring distances of all points, and use the three-dimensional curved surface mesh as the topological graph of the three-dimensional phase space.

6. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 5, characterized in that, Tracing the curvature change trajectory of the high-dimensional manifold in the three-dimensional phase space topology diagram, and extracting singular attractor feature quantities to characterize the local strength defects of the ultrafine polyester low-elasticity yarn, specifically including: The three-dimensional phase space topology map is discretized into multiple local grid cells, and the principal curvature and Gaussian curvature values ​​of all points in each local grid cell are calculated. Along the axial sequence of the ultrafine polyester low-elasticity yarn, the principal curvature values ​​and Gaussian curvature values ​​of each local grid unit are arranged into a curvature variation sequence; In the curvature change sequence, abnormal grid cells with abrupt curvature value changes are identified, and the three-dimensional phase space coordinates corresponding to the abnormal grid cells and the curvature change amplitude are used as the strange attractor feature quantities.

7. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 6, characterized in that, Abnormal mesh cells with abrupt curvature values ​​are identified by calculating that the difference between the Gaussian curvature value of the local mesh cell and the mean Gaussian curvature value of the adjacent mesh cells exceeds three times the standard deviation.

8. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 6, characterized in that, Based on the fluctuation distribution of the Lyapunov exponent spectrum of the strange attractor characteristic quantity along the length direction of the ultrafine polyester low-elasticity yarn, the strength uniformity detection result of the ultrafine polyester low-elasticity yarn is output, specifically including: Using the three-dimensional phase space coordinates corresponding to each anomalous grid cell as the initial trajectory point, the neighboring trajectories in the three-dimensional phase space topology map are tracked, and the maximum Lyapunov exponent and the convergence value of the maximum Lyapunov exponent at each anomalous grid cell are calculated. The convergence values ​​of the maximum Lyapunov exponent are arranged according to the corresponding positions of the abnormal grid cells along the axial direction of the ultrafine polyester low elastic yarn to form a one-dimensional exponential spectrum curve. The one-dimensional exponential spectrum curve is subjected to a difference operation. The positions where the difference value exceeds the preset variation threshold are marked as intensity non-uniformity points, and the detection spectrum marked with the coordinates of all intensity non-uniformity points is output.

9. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 8, characterized in that, The maximum Lyapunov exponent is calculated using a small data method by tracking the average divergence rate of the nearest neighbors in the phase space.

10. The intelligent detection method for the strength uniformity of ultrafine polyester low-elasticity yarn according to claim 4, characterized in that, The frequency domain representation of the intrinsic mode function components of each signal to be decomposed is updated iteratively using the alternating direction multiplier method. After each iteration, the frequency domain waveform similarity of the mode function components between two adjacent iterations is calculated, specifically including: In each iteration, the center frequency of the current intrinsic mode function component is fixed, and the frequency domain amplitude of the intrinsic mode function component is updated by Wiener filtering; The center frequency of the intrinsic mode function component is recalculated based on the updated frequency domain amplitude, thus completing one update of the frequency domain representation; Normalized cross-correlation is performed on the frequency domain amplitude sequences output from two adjacent iterations, and the peak value obtained from the normalized cross-correlation calculation is used as the frequency domain waveform similarity.

Images