A method and system for detecting tearing strength of textile fabrics

Abstract

The present invention discloses a method and system for detecting the tear strength of textile fabrics, relating to the technical field of fabric detection. The method comprises: synchronously collecting fabric microstructure, macromechanics and composition data through a multimodal sensing system, and dynamically switching detection modes based on real-time recognition results; inputting spatiotemporally correlated defect data into a prediction model enhanced by physical constraints to generate crack propagation trends and tear strength prediction values; wherein the prediction model integrates Griffith's crack propagation theory and a hidden Markov model to achieve dual constraints of material mechanics laws and dynamic defect evolution; adopting a non-contact detection module and a flexible contact detection module to work together, and combining edge and cloud distributed architecture to output a process optimization solution.

Description

Technical Field

[0001] The present application relates to the technical field of fabric testing, and in particular to a method and system for testing the tear strength of textile fabrics. Background Art

[0002] The tear strength of textile fabrics is a core indicator of their durability and performance, directly impacting the quality, safety, and service life of end products. In areas such as apparel, outdoor equipment, and aerospace, fabric tear failure can have serious consequences, such as garment tears that affect the wearing experience and damage to protective fabrics that threaten life. Therefore, accurately measuring fabric tear strength and identifying potential failure hazards are critical steps in ensuring product reliability and optimizing production processes.

[0003] Traditional testing methods mainly use physical destructive tests, such as the Elmendorf method and the trouser tear method, to obtain the maximum tear load of the fabric. This method requires destroying the sample and a single test takes up to 5-10 minutes, which is time-consuming and labor-intensive and has a high sample loss rate.

[0004] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0005] The embodiments of the present application provide a method and system for detecting the tear strength of textile fabrics to solve the above-mentioned technical problems.

[0006] This application provides a method for testing the tear strength of textile fabrics, comprising:

[0007] The multimodal sensing system simultaneously collects fabric microstructure, macromechanics, and composition data, and dynamically switches detection modes based on real-time recognition results.

[0008] Inputting spatiotemporally correlated defect data into a prediction model augmented by physical constraints generates crack growth trends and tear strength predictions. This prediction model integrates Griffith's crack growth theory with a hidden Markov model to achieve dual constraints based on material mechanics and dynamic defect evolution.

[0009] The non-contact detection module and the flexible contact detection module work together, and the edge and cloud distributed architecture are combined to output process optimization solutions.

[0010] Furthermore, the implementation of dynamically switching the detection mode includes:

[0011] The fiber molecular bonding information is obtained through a near-infrared spectrometer. Combined with the fabric surface texture captured by an industrial camera, an adaptive feature pyramid network is constructed for classification to output the fabric type and structure.

[0012] When a knitted fabric is identified, the cyclic tensile load simulation module is activated to simultaneously extract the yarn slip trajectory data and molecular bond strength attenuation data under tensile load. The weights are dynamically assigned through a dual-channel attention mechanism to generate a joint mapping relationship between tensile load and slip displacement.

[0013] When the fabric is identified as coated, the multi-scale residual convolutional network is activated to fuse the porosity distribution data of the penetrating infrared thermal imaging sequence and the laser profile scanning to locate the critical area of fiber breakage under the coating.

[0014] Furthermore, the construction of the prediction model enhanced by physical constraints includes:

[0015] A Griffith energy release rate threshold is embedded in the state transition probability matrix of the hidden Markov model, and when the predicted energy of the crack propagation path exceeds the fracture toughness of the material, the state transition path is adjusted to conform to physical laws;

[0016] A joint loss function is designed to dynamically weight the timing prediction error of the hidden Markov model and the energy constraint error calculated by the Griffith crack growth theory, and the matching degree between the crack evolution timing and the physical law is simultaneously optimized through gradient back propagation;

[0017] According to the optimization result of the joint loss function, a tear strength prediction value that satisfies the dual constraints of stress concentration and energy dissipation is output.

[0018] Furthermore, the implementation of the dual-channel attention mechanism includes:

[0019] A temporal convolutional network is used to extract temporal features from the yarn slip trajectory data to generate a spatiotemporal encoding vector of the slip displacement;

[0020] Using a graph convolutional network to model the molecular interaction relationship of the molecular bond strength attenuation data to generate a topological feature vector of the bond strength;

[0021] The association weights between the spatiotemporal coding vector and the topological feature vector are dynamically calculated through a cross-attention layer to generate an attention distribution map of the joint mapping relationship for indicating parameter adjustment of the cyclic tensile load.

[0022] Furthermore, the adjustment of the state transition path includes:

[0023] Based on Griffith's crack growth theory, the stress intensity factor of the current crack growth direction is calculated to generate the energy release rate constraint boundary;

[0024] In the Viterbi decoding process of the crack growth hidden Markov model, the state transition paths that exceed the constraint boundary are removed;

[0025] The probability of state transition paths that conform to physical laws is redistributed to ensure that the crack propagation trend prediction satisfies both the dynamic defect evolution and material fracture mechanics laws.

[0026] Furthermore, the non-contact detection module includes:

[0027] The air-coupled ultrasonic transducer transmits ultrasonic waves of a set frequency, and the wavelet packet decomposition algorithm is used to extract the yarn breakage characteristic frequency band in the echo signal.

[0028] The acoustic characteristic patterns of fiber debonding and single fiber breakage were distinguished based on the random forest classifier, and the classification results were used as the criterion for defect type.

[0029] A laser profile scanner is used simultaneously to obtain three-dimensional surface morphology data of the fabric, and a curvature mutation detection algorithm is used to locate stress concentration areas where the stiffness gradient exceeds the set threshold.

[0030] Furthermore, the implementation of the curvature mutation detection algorithm includes:

[0031] Calculate Gaussian curvature of 3D topography data and generate a curvature distribution heat map;

[0032] The morphological gradient operator is used to extract the curvature mutation boundary and mark the potential tearing starting point;

[0033] The curvature mutation area is spatially matched with the defect position detected by ultrasonic testing to verify the reliability of the stress concentration area.

[0034] Furthermore, the implementation of the edge and cloud distributed architecture includes:

[0035] A fault-tolerant detection model is deployed at the edge nodes of the weaving process, and a bit-flip fault-tolerant mechanism is used to process the real-time collected warp and weft yarn breakage data.

[0036] After receiving the abnormal data, the cloud platform simulates the tearing life in the actual scenario through the digital twin model, and reversely infers the process parameter combination that meets the target strength.

[0037] Furthermore, the method further comprises an environment adaptive calibration step:

[0038] A mapping relationship table between temperature and humidity and material strength attenuation was established, and the test results were dynamically corrected through Gaussian process regression. When the ambient humidity exceeded the set threshold, a random correction coefficient was applied to the predicted tear strength of cotton fabrics.

[0039] The illumination invariance feature extraction algorithm is used to separate the brightness component in the HSV color space, and the influence of color temperature fluctuation on fiber texture recognition is eliminated through histogram normalization.

[0040] The present application provides a textile fabric tear strength testing system, comprising:

[0041] Detection mode switching module, used to synchronously collect fabric microstructure, macromechanics and composition data through a multimodal sensing system, and dynamically switch detection modes based on real-time recognition results;

[0042] A tear strength prediction module, which is used to input spatiotemporally correlated defect data into a prediction model enhanced by physical constraints to generate crack growth trends and tear strength predictions. This prediction model integrates Griffith's crack growth theory and a hidden Markov model to achieve dual constraints based on material mechanics and dynamic defect evolution.

[0043] The process optimization module is used to use the non-contact detection module and the flexible contact detection module to work together, combining the edge and cloud distributed architecture to output process optimization solutions.

[0044] Based on the embodiments provided herein, a multimodal sensing system is employed to simultaneously acquire fabric microstructure, macromechanics, and compositional data, overcoming the drawback of traditional testing methods that rely on a single physical metric (such as maximum tear load). By dynamically switching detection modes based on real-time recognition results, the system can adaptively adjust detection strategies based on different fabric characteristics, avoiding the blind spots of traditional fixed-process methods for complex fabrics and improving the comprehensiveness and accuracy of detection results. A physically constrained prediction model integrates Griffith's crack growth theory and the Hidden Markov Model, for the first time combining fundamental laws of material mechanics with time series analysis of defect evolution. Compared to traditional "black box" prediction methods that rely solely on statistical laws, this model not only adheres to objective physical mechanisms but also captures the dynamic changes of defects with temporal and spatial correlations (such as the correlation of defect locations during continuous production). This model generates tear strength predictions that include crack growth trends, providing a scientific basis for early identification of potential failure hazards. The collaborative operation of the non-contact and flexible contact detection modules addresses the damage that traditional rigid testing methods can cause to fragile fabrics (such as silk and lace) and the challenges of adapting to complex fabric morphologies (such as curved surfaces and wrinkles). Combining the edge and cloud distributed architecture, edge nodes process high-frequency detection data in real time and respond quickly to anomalies. The cloud platform outputs process optimization solutions based on multi-dimensional data, forming an intelligent closed loop of "detection-analysis-process adjustment", and promoting the upgrade of textile production from an "experience-driven" to a "data-driven" quality control model. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings described herein are used to provide a further understanding of the embodiments of the present invention and constitute a part of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0046] Figure 1 Flowchart of an optional method for detecting tear strength of textile fabrics according to an embodiment of the present application;

[0047] Figure 2 Flowchart of another optional method for testing the tearing strength of textile fabrics according to an embodiment of the present application;

[0048] Figure 3 2 is a structural diagram of an optional textile fabric tear strength detection system according to an embodiment of the present application.

[0049] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0051] Alternatively, as Figure 1 As shown, the present application provides a method for testing the tear strength of textile fabrics, comprising:

[0052] S101, uses a multimodal sensing system to simultaneously collect fabric microstructure, macromechanics, and composition data, and dynamically switches detection modes based on real-time recognition results;

[0053] S102, inputting the spatiotemporally correlated defect data into a prediction model enhanced by physical constraints to generate crack growth trends and tear strength predictions. The prediction model integrates Griffith's crack growth theory and the hidden Markov model to achieve dual constraints of material mechanics and dynamic defect evolution.

[0054] S103 uses a non-contact detection module and a flexible contact detection module to work together, combining edge and cloud distributed architecture to output a process optimization solution.

[0055] Based on the embodiments provided herein, a multimodal sensing system is employed to simultaneously acquire fabric microstructure, macromechanics, and compositional data, overcoming the drawback of traditional testing methods that rely on a single physical metric (such as maximum tear load). By dynamically switching detection modes based on real-time recognition results, the system can adaptively adjust detection strategies based on different fabric characteristics, avoiding the blind spots of traditional fixed-process methods for complex fabrics and improving the comprehensiveness and accuracy of detection results. A physically constrained prediction model integrates Griffith's crack growth theory and the Hidden Markov Model, for the first time combining fundamental laws of material mechanics with time series analysis of defect evolution. Compared to traditional "black box" prediction methods that rely solely on statistical laws, this model not only adheres to objective physical mechanisms but also captures the dynamic changes of defects with temporal and spatial correlations (such as the correlation of defect locations during continuous production). This model generates tear strength predictions that include crack growth trends, providing a scientific basis for early identification of potential failure hazards. The collaborative operation of the non-contact and flexible contact detection modules addresses the damage that traditional rigid testing methods can cause to fragile fabrics (such as silk and lace) and the challenges of adapting to complex fabric morphologies (such as curved surfaces and wrinkles). Combining the edge and cloud distributed architecture, edge nodes process high-frequency detection data in real time and respond quickly to anomalies. The cloud platform outputs process optimization solutions based on multi-dimensional data, forming an intelligent closed loop of "detection-analysis-process adjustment", and promoting the upgrade of textile production from an "experience-driven" to a "data-driven" quality control model.

[0056] Furthermore, if Figure 2 As shown, the implementation of dynamic switching detection mode includes:

[0057] S201 uses a near-infrared spectrometer to obtain fiber molecular bonding information, combines it with the fabric surface texture captured by an industrial camera, and constructs an adaptive feature pyramid network for classification to output the fabric type and structure;

[0058] S202: When a knitted fabric is identified, a cyclic tensile load simulation module is started to simultaneously extract yarn slip trajectory data and molecular bond strength attenuation data under tensile load, dynamically assign weights through a dual-channel attention mechanism, and generate a joint mapping relationship between tensile load and slip displacement;

[0059] S203, when it is identified as a coated fabric, a multi-scale residual convolutional network is activated to fuse the porosity distribution data of the penetrating infrared thermal imaging sequence and the laser profile scanning to locate the critical area of fiber fracture under the coating.

[0060] In some embodiments of the present application, the fabric type is classified by multimodal data, and a differentiated detection strategy is initiated for knitted / coated fabrics. Specifically,

[0061] Knitted fabric inspection scenarios (such as stretch knit fabrics): When collecting fiber molecular bonding information using a near-infrared spectrometer, the preset scanning wavelength range is 900-1700 nm (covering the characteristic absorption peaks of cotton and spandex), and the industrial camera captures the loop structure texture at a resolution of 12 megapixels. When training the adaptive feature pyramid network, the preset "loop density" classification threshold for knitted fabrics is ≥25 loops / cm (to distinguish it from the warp and weft floating-point structure of woven fabrics).

[0062] When a knitted fabric is identified, the cyclic tensile load simulation module is activated: the preset number of stretching cycles is 50 times (simulating the daily stretching frequency of human joint activities), and the single load range is ±15% of the nominal strength of the fabric (for example, for a fabric with a nominal tear strength of 60N, the load reciprocates between 51N-69N). The yarn slip trajectory is tracked synchronously with a high-speed camera (frame rate 1000fps) with an accuracy of 0.02mm. Combined with the molecular bond strength attenuation data (determined by the change rate of the absorption peak of the amide I band of the near-infrared spectrum), a "load-slip" joint mapping relationship is generated through the dual-channel attention mechanism to accurately locate the weak connection points of the coil structure.

[0063] Coated fabric inspection scenarios (such as outdoor waterproof and breathable fabrics): When a multi-scale residual convolutional network processes penetrating infrared thermal imaging data, the preset detection wavelength is 8-14μm (the transparent spectral range of the coating material), and a laser profile scanner collects surface porosity distribution with 5μm accuracy. When locating the critical area of fiber breakage under the coating, a preset porosity anomaly threshold of ≥3% (i.e., local porosity increases of more than 3% compared to the normal area are marked as defective areas). The authenticity of stress concentration is verified by the temperature gradient change in the thermal imaging sequence (preset ΔT ≥ 2°C) to avoid misjudgment caused by interference from the coating surface. Based on the embodiments provided in this application, through the collaborative data acquisition of a near-infrared spectrometer and an industrial camera, an adaptive feature pyramid network is constructed to achieve accurate classification of fabric type and structure, changing the traditional detection model that relies on manual experience or single-indicator classification.

[0064] A cyclic tensile load simulation module for knitted fabrics simultaneously integrates yarn slip trajectory and molecular bond strength decay data. Dynamically assigning weights through a dual-channel attention mechanism enables the detection system to capture the slip failure mode of knitted fabrics' unique loop structure under cyclic loading. This process links macroscopic mechanical behavior (tensile load-displacement curves) with microscopic molecular bonding characteristics (such as hydrogen bond breakage). A multi-scale residual convolutional network for coated fabrics, integrating penetrating infrared thermal imaging and laser profile scanning data, can penetrate the coating to locate critical areas of fiber breakage. This overcomes the bottleneck of traditional visual inspection, which is limited by the surface condition of the fabric, and provides a new technical path for internal defect detection in functional coated fabrics.

[0065] Furthermore, the construction of the prediction model enhanced by physical constraints includes:

[0066] The Griffith energy release rate threshold is embedded in the state transition probability matrix of the hidden Markov model. When the predicted energy of the crack propagation path exceeds the fracture toughness of the material, the state transition path is adjusted to conform to the physical law.

[0067] In this embodiment, the preset Griffith energy release rate threshold includes but is not limited to 0.5 J / m², 0.7 J / m², etc.

[0068] A joint loss function is designed to dynamically weight the timing prediction error of the hidden Markov model and the energy constraint error calculated by the Griffith crack growth theory, and the matching degree between the crack evolution timing and physical laws is simultaneously optimized through gradient back propagation.

[0069] According to the optimization results of the joint loss function, the tear strength prediction value that satisfies the dual constraints of stress concentration and energy dissipation is output.

[0070] In a specific embodiment, the joint loss function for:

[0071] ;

[0072] ;

[0073] in, is the weighting factor of timing error and physical constraint, with an initial value of 0.6 (prioritizing the timing continuity of crack propagation). When the physical constraint error is Exceeding the relative error of time series forecast When , it is automatically adjusted to 0.5 (balancing the two types of errors); is the relative error of time series prediction of the hidden Markov model; is the total number of time steps; For the The predicted value of the crack extension displacement at the time step is output by the hidden Markov model, with the unit of mm, such as 0.3 mm, reflecting the model's prediction of the crack extension in the time dimension; For the The measured value of the crack extension displacement at the time step is collected by a laser displacement sensor, such as 0.31 mm, which serves as the benchmark data for verifying the prediction results; is the physical constraint error of Griffith theory; is the total number of crack extension paths; For the The measured energy release rate of the crack path is obtained by collecting stress wave energy through an air-coupled ultrasonic transducer and converting it into the fabric thickness. The unit is J / m². For example, the measured value is 0.6 J / m when cotton fiber breaks. The energy release rate threshold of Griffith's crack growth theory is set differently according to the fiber type: 0.5 J / m² for cotton fiber fabrics (critical energy for hydrogen bond rupture) and 2.0 J / m² for polyester fiber fabrics (critical energy for covalent bond rupture); The physical constraint tolerance coefficient allows a reasonable fluctuation of 10% in the energy release rate near the threshold (for example, a threshold of 0.5 J / m² can accept a range of 0.45-0.55 J / m²) to avoid misjudgment caused by detection noise.

[0074] Based on the above formula, Griffith's crack growth theory is embedded in the machine learning loss function. Using aramid fireproof fabric testing as an example, the prediction error for aramid fabric is reduced by dynamically weighting the temporal error and the energy constraint error (for example, automatically increasing the physical constraint weight to 0.5 when continuous energy anomalies are detected). All predictions meet the Griffith theory threshold (for example, aramid fiber fracture toughness is 5.0 J / m², and the measured energy release rate is ≥4.5 J / m²). In testing a firefighting suit fabric, the model corrected the erroneous path of "sudden fracture without stress concentration" (reducing the proportion of such paths from 35% to 5%), identifying microcracks under the coating caused by high-speed impact, and providing a warning of fabric failure two hours earlier than traditional methods.

[0075] Based on the embodiments provided in this application, the Griffith energy release rate threshold is embedded in the state transition probability matrix of the hidden Markov model, realizing for the first time the deep integration of material fracture mechanics theory and defect evolution time series analysis. Traditional machine learning models often ignore the constraints of physical laws when predicting crack extension, and may output unreasonable results that violate the stress-strain relationship. However, this solution introduces the energy release rate constraint boundary in the state transition path, so that the model prediction process simultaneously satisfies the objective physical mechanism of "stress concentration leading to energy dissipation". The design of the joint loss function dynamically weights the time series prediction error and the physical law constraint error, and synchronously optimizes the time series characteristics of crack evolution and the material mechanics matching degree through gradient back propagation, so that the generated tear strength prediction value not only reflects the dynamic change trend of the defect (such as the spatiotemporal correlation of the defect position in continuous production), but also conforms to the energy conservation principle of Griffith crack extension theory.

[0076] Furthermore, the implementation of the dual-channel attention mechanism includes:

[0077] The temporal convolutional network is used to extract the temporal features of the yarn slip trajectory data and generate the spatiotemporal encoding vector of the slip displacement.

[0078] The graph convolutional network is used to model the molecular interaction relationship based on the molecular bond strength decay data to generate the topological feature vector of the bond strength.

[0079] The association weights between the spatiotemporal coding vector and the topological feature vector are dynamically calculated through the cross-attention layer to generate an attention distribution map of the joint mapping relationship, which is used to indicate the parameter adjustment of the cyclic tensile load.

[0080] In a specific embodiment, the dual-channel attention mechanism association weight is calculated based on the following formula:

[0081] ;

[0082] in, For the Yarn slip trajectory characteristics for each stretching cycle, including displacement (mm), slip velocity (mm / s), and slip acceleration (mm / s²); P is the total number of stretching cycles. For knitted fabrics, this simulates the stretching frequency of daily wear, so the scenario value is 50 (e.g., the knee bending cycle for Lycra stretch fabric); Q is the total number of molecular bonding nodes, corresponding to the number of bonding points in the fiber molecular chain. The cellulose chain of cotton fiber contains approximately 200 hydrogen bond nodes, so the scenario value is 200. is the strength characteristic of the qth molecular bonding node, including bond energy (e.g. 0.5 eV for cotton fiber hydrogen bond), bond length (nm), and dihedral angle (°), detected by near-infrared spectroscopy; is the attention weight matrix of the sliding trajectory feature; is the attention weight matrix of molecular bonding features; is the bias vector, which optimizes the GELU activation function input. It is initially 0 and is adaptively adjusted through training. is the association weight of the qth bonding node in the pth cycle. For example, a spandex node with a weight of 0.85 is determined to be a high-risk failure point.

[0083] Based on the above formula, this method dynamically calculates the correlation weight between yarn slip trajectories and molecular bond strength, achieving cross-modal deep correlation analysis of knitted fabrics from macroscopic slip failure to microscopic bond attenuation. Taking Lycra stretch fabric detection as an example, traditional methods can only detect obvious yarn breaks, resulting in a high rate of missed detection. However, this formula improves the detection accuracy of coil connection failure by focusing on the coupling region of "sudden increase in slip displacement + sudden drop in bond strength" through attention weighting (for example, triggering high-precision detection when the weight of a coil node exceeds 0.8). Specifically, when the weight of bond node No. 150 of the spandex molecular chain suddenly increases by 40% during the 30th stretching cycle, the system can identify the risk of hydrogen bond breakage caused by slip at this node in advance, providing a warning five cycles earlier than traditional single-signal detection, avoiding stretch fabric tearing accidents caused by missed detection.

[0084] Based on the embodiment provided in this application, a dual-channel attention mechanism for knitted fabrics is used to extract the temporal features of the yarn slip trajectory through a temporal convolutional network, and combined with the topological features of the intermolecular interaction modeled by a graph convolutional network, a cross-scale correlation analysis model of "macroscopic slip behavior-microscopic bonding state" is constructed. Traditional detection methods for tear strength analysis of knitted fabrics usually only focus on the macroscopic displacement data under tensile load, while this solution dynamically calculates the correlation weights of the two types of features through a cross-attention layer, which can accurately capture the attenuation law of the molecular bonding strength during yarn slip. For example, when the yarn slip rate in a certain area is strongly correlated with the probability of breaking the adjacent molecular bonds, the system can automatically identify the area as a potential failure starting point. The establishment of this joint mapping relationship provides a more refined feature dimension for the failure mechanism analysis of knitted fabrics, enabling the detection system to perform customized defect diagnosis based on the coil structure characteristics of knitted fabrics, avoiding the problem of insufficient adaptability of the traditional "unified parameter" detection mode to special fabric structures.

[0085] Furthermore, adjustments to the state transition path include:

[0086] Based on Griffith's crack growth theory, the stress intensity factor of the current crack growth direction is calculated to generate the energy release rate constraint boundary;

[0087] In the Viterbi decoding process of the crack growth hidden Markov model, the state transition paths that exceed the constraint boundary are removed;

[0088] The probability of state transition paths that conform to physical laws is redistributed to ensure that the crack propagation trend prediction satisfies both the dynamic defect evolution and material fracture mechanics laws.

[0089] In some embodiments of the present application, Griffith theory constraints are embedded in the hidden Markov model, and energy thresholds and state transition rules are preset to ensure that the prediction complies with physical laws.

[0090] For example, in the cotton fabric crack prediction scenario, when constructing a hidden Markov model, a Griffith energy release rate threshold of 0.5 J / m² (determined through standard single-filament breakage tests, reflecting the critical energy for breaking hydrogen bonds between cotton fiber molecules) is preset based on the measured fracture toughness of cotton fibers. When the predicted energy of a crack propagation path exceeds this threshold by more than 10% (i.e., ≥0.55 J / m²), a state transition path adjustment is triggered: Constraint boundaries are calculated based on a stress intensity factor formula (not explicitly defined, but simply describing the logic). Unreasonable "sudden fracture" paths that exceed the boundary by 20% (such as predictions of direct fracture without stress concentration) are removed, and the probabilities of paths that conform to energy conservation are redistributed (for example, paths with gradual energy release are retained but their probability weight is increased by 30%).

[0091] When the joint loss function is dynamically weighted, the preset initial weight is "timing prediction error: energy constraint error = 0.6:0.4". When the energy constraint error accounts for more than 60% in five consecutive tests (such as in scenarios where fabric aging causes toughness to decrease), the weight is automatically adjusted to 0.5:0.5 to ensure that the model simultaneously optimizes the matching degree between the crack evolution timing and physical laws.

[0092] Based on the embodiments provided in this application, during the Viterbi decoding process of the hidden Markov model for crack propagation, the stress intensity factor is calculated based on Griffith theory and the energy release rate constraint boundary is generated, effectively filtering out state transition paths that do not conform to the laws of material fracture mechanics. When processing crack propagation data, traditional time-series prediction models may generate prediction results that violate physical common sense (e.g., sudden crack changes occur when stress is not concentrated) due to data noise or model bias. However, this solution ensures that crack propagation trend predictions simultaneously meet the laws of dynamic defect evolution and material fracture toughness requirements by removing paths that exceed the constraint boundary and redistributing the probability of valid paths. For example, when it is detected that the stress intensity factor in a certain area does not reach the material fracture threshold, the system automatically suppresses the unreasonable prediction of "sudden crack propagation" and instead prioritizes paths that conform to the gradual change of energy release rate. This makes the prediction results closer to the actual physical process and improves the reliability and scientific basis of defect evolution prediction.

[0093] Furthermore, the non-contact detection module includes:

[0094] The air-coupled ultrasonic transducer transmits ultrasonic waves of a set frequency, and the wavelet packet decomposition algorithm is used to extract the yarn breakage characteristic frequency band in the echo signal.

[0095] In fiber fabric testing scenarios (such as pure cotton canvas), the set frequency can be 200kHz. In polyester fiber fabric testing scenarios (such as polyester industrial cloth), the set frequency can be 350kHz.

[0096] The acoustic characteristic patterns of fiber debonding and single fiber breakage were distinguished based on the random forest classifier, and the classification results were used as the criterion for defect type.

[0097] A laser profile scanner is used simultaneously to obtain three-dimensional morphology data of the fabric surface, and a curvature mutation detection algorithm is used to locate stress concentration areas where the stiffness gradient exceeds the set gradient threshold.

[0098] In this example, for low-risk fabrics (such as silk), a gradient threshold of 8 N / mm² is set. Because silk fibers are delicate and have low stiffness, even a slight abnormality in fiber alignment can cause a sudden change in stiffness. When detecting slippage defects in silk fabrics, a high-risk slippage point is identified when the stiffness gradient in a region exceeds 8 N / mm² and the curvature heat map appears red (Gaussian curvature > 0.008 mm²). Slippage in silk fabrics is typically caused by a sudden increase in stiffness due to accumulated fiber slippage.

[0099] High-rigidity fabrics (such as aramid fireproof fabrics): Set the gradient threshold to 20N / mm² (aramid fibers are inherently stiff, requiring a larger gradient change to detect defects). When detecting fiber breaks in aramid fabric, if the stiffness gradient reaches 22N / mm² and the ultrasonic test captures a 350kHz high-frequency fracture signal at the same location, the system identifies the area as a stress concentration zone (aramid fiber fracture is accompanied by a significant stiffness change and high-frequency stress waves).

[0100] Based on the embodiments provided herein, the non-contact detection module uses an air-coupled ultrasonic transducer and a laser profiler to collaborate, creating a dual non-destructive testing system combining "acoustic feature analysis and topographic defect location." A wavelet packet decomposition algorithm extracts characteristic frequency bands from the ultrasonic echo signal, and combined with a random forest classifier, distinguishes the acoustic patterns of fiber debonding and single-filament breakage, enabling the system to identify deep structural defects without contacting the fabric. This is crucial for the inspection of fragile materials such as medical non-woven fabrics and fabrics for cultural relics protection, avoiding the physical damage that can be caused by traditional contact inspection. Synchronously collected laser 3D topography data locates stress concentration areas through curvature mutation detection, forming a spatial cross-validation with the ultrasonic inspection results. This addresses the potential for misjudgment of defects with single non-contact inspection technologies (e.g., relying solely on ultrasound) and provides multimodal fusion technical support for internal defect detection in complex structural fabrics (e.g., multi-layer composite fabrics).

[0101] Furthermore, the implementation of the curvature mutation detection algorithm includes:

[0102] Calculate Gaussian curvature of 3D topography data and generate a curvature distribution heat map;

[0103] The morphological gradient operator is used to extract the curvature mutation boundary and mark the potential tearing starting point;

[0104] The curvature mutation area is spatially matched with the defect position detected by ultrasonic testing to verify the reliability of the stress concentration area.

[0105] In some embodiments of the present application, through collaborative detection of ultrasound and laser, the stress concentration area is located by presetting characteristic frequency bands and stiffness gradient thresholds.

[0106] In a specific embodiment, the multimodal defect matching score is calculated based on the following formula:

[0107] ;

[0108] in, is the total number of defect points to be matched. Each time, high-confidence candidate points are extracted from the laser scanning and ultrasonic testing results. For example, a value of 10 is used to ensure that potential defect areas on the surface and inside of the fabric are covered. For sum index, corresponding to the nth defect point (n=1,2,…,N), for example, the third point satisfies both surface curvature mutation (stiffness gradient 16N / mm²) and ultrasonic signal abnormality (250kHz frequency band energy mutation), and its reliability is verified first; The coordinates of the curvature mutation point located by the nth laser profile scan are collected by a laser profile scanner (accuracy 5μm) and are in mm. They reflect the area where the stiffness gradient of the fabric surface exceeds 15N / mm², such as the micro-deformation position of the warp and weft intersection of woven fabrics (12.5mm, 8.3mm). The coordinates of the defect point located by the nth air-coupled ultrasonic test are located by the ultrasonic transducer (accuracy 0.1 mm), in mm, to capture the acoustic characteristic position of fiber breakage or debonding, such as the fiber breakage point under the coated fabric (12.8 mm, 8.1 mm); is the Gaussian kernel width (spatial matching scale factor), which is dynamically determined according to the fabric type. For knitted fabrics, it is 0.3 mm (to accommodate the natural deformation error of the coil structure), and for coated fabrics, it is 0.8 mm (to compensate for the propagation delay error of the coating on the ultrasonic signal). κ is the multimodal defect matching score, ranging from 0 to 1. A value ≥ 0.8 indicates a high-reliability stress concentration area. For example, when the coordinate difference in a certain area is 0.3 mm, the calculated value κ = 0.91 confirms the presence of a 0.2 mm microcrack under the coating, resolving the problem of missed detection in traditional testing.

[0109] Based on the above formula, a Gaussian kernel function is used to quantitatively calculate the defect location matching between laser scanning and ultrasonic testing. This cross-modal reliability verification system is constructed, addressing the challenge of detecting hidden defects in coated and thick fabrics. For example, traditional single-use ultrasonic testing for outdoor waterproof and breathable fabrics results in a high miss rate due to coating obscuration. However, this formula improves the detection accuracy of even 0.2mm microcracks beneath the coating by setting a Gaussian kernel width σ=0.8mm specific to coated fabrics (compensating for the 0.5mm positioning error caused by the coating). In a specific implementation, when the laser scan detected a stiffness gradient of 16N / mm² (exceeding the threshold of 15N / mm²) at (15.2mm, 8.3mm) and the ultrasonic detection of a fiber break signal at (15.5mm, 8.0mm), the calculated matching degree κ=0.88 was determined. The system identified this area as a critical break zone and guided the process to adjust the warp tension by ±10cN, avoiding waterproofing failures caused by fiber breakage beneath the coating. For canvas with a thickness of 3mm, the interlayer debonding detection rate is increased from 60% to 98% through adaptive adjustment of σ=1.0mm, significantly improving the safety performance detection capability of thick fabrics.

[0110] For example, in yarn breakage and fiber debonding detection scenarios, an air-coupled ultrasonic transducer emits 250kHz ultrasonic waves (the preset frequency band covers the characteristic response of cotton / polyester fiber breakage). A wavelet packet decomposition algorithm extracts the 200-300kHz frequency band as a yarn breakage signature (a breakage signal is identified when the energy in this frequency band accounts for ≥40%), and the 100-150kHz frequency band as a fiber debonding signature (a debonding signal is identified when the phase difference exceeds 180°). The random forest classifier is preset with a classification confidence threshold of 70% (i.e., the final defect type is output when the classification consistency rate of a single decision tree is ≥70%).

[0111] The laser profile scanner collects 3D topography data at a resolution of 10 μm, and the preset threshold for Gaussian curvature calculation is (Corresponding to fiber alignment anomalies at a scale of 0.2mm), a morphological gradient operator uses a 3×3 pixel window to extract the curvature mutation boundary. When the spatial deviation between the curvature mutation area and the ultrasonically detected defect location is ≤0.5mm, it is identified as a high-reliability stress concentration area (for example, an area with both ultrasonic echo anomalies and surface convex deformation is marked as the starting point of a tear).

[0112] Based on the embodiments provided in this application, the curvature mutation detection algorithm performs Gaussian curvature calculation and morphological gradient processing on the three-dimensional morphological data obtained by laser contour scanning, which can accurately locate areas with abnormal stiffness gradients on the surface of the fabric. These areas are often the starting points of tear failure, and traditional visual inspection, which relies on grayscale or texture features, has difficulty capturing the stiffness differences caused by microscopic curvature changes. By spatially matching the curvature mutation boundary with the defect position detected by ultrasonic detection, a "surface morphology abnormality-internal structural defect" correlation verification mechanism is formed. For example, when the curvature mutation boundary of a certain area highly overlaps with the fiber breakage characteristic frequency band in the ultrasonic echo signal, the system can determine that the area is a high-risk tearing point. This spatial registration and cross-validation of multimodal data significantly improves the accuracy of locating stress concentration areas, avoids missed or misjudgments caused by insufficient data dimensions of a single detection technology, and provides a more reliable defect location basis for the accurate prediction of fabric tear strength.

[0113] Furthermore, the implementation of edge and cloud distributed architecture includes:

[0114] A fault-tolerant detection model is deployed at the edge nodes of the weaving process, and a bit-flip fault-tolerant mechanism is used to process the real-time collected warp and weft yarn breakage data.

[0115] After receiving the abnormal data, the cloud platform simulates the tearing life in the actual scenario through the digital twin model, and reversely infers the process parameter combination that meets the target strength.

[0116] In some embodiments of the present application, edge nodes perform real-time fault-tolerant processing and reverse infer process parameters in the cloud, involving preset abnormality determination thresholds and parameter adjustment ranges.

[0117] For example, in the real-time monitoring scenario of the weaving process, the fault-tolerant detection model deployed on the edge node has the following preset abnormal judgment rules for warp and weft yarn breakage: the tension load fluctuation at five consecutive sampling points (sampling frequency 100Hz) exceeds ±20% of the nominal value (for example, if the nominal tension is 100cN, an alarm is triggered when >120cN or <80cN is continuously detected), and the bit flip fault-tolerant mechanism automatically filters short-term noise (such as instantaneous fluctuations caused by equipment vibration, and abnormal signals with a duration of <50ms are ignored).

[0118] After receiving abnormal data, the cloud-based digital twin model presets simulation parameters: simulating the "stretching-bending" composite stress scenario of the fabric in actual use (such as the sudden impact condition of automobile airbag fabric) with a time step of 0.1 seconds. When inferring process parameters, the warp tension adjustment range is limited to ±10cN of the current value (to avoid exceeding the stable adjustment range of the loom tension system), ensuring the engineering feasibility of the optimization plan.

[0119] Based on the embodiments provided herein, the design of an edge and cloud distributed architecture achieves a layered collaboration of "real-time detection, exception handling, and process optimization" in the textile production process. Edge nodes in the weaving process deploy a fault-tolerant detection model, employing a bit-flipping fault-tolerant mechanism to process the frequently collected warp and weft breakage data. This model can perform data noise reduction and anomaly identification within milliseconds, ensuring real-time detection even during high-speed production line operation. This addresses the quality control lag caused by network latency in traditional cloud-based centralized processing. After receiving abnormal data uploaded by edge nodes, the cloud platform uses a digital twin model to simulate the tear life of fabrics under actual usage scenarios (such as complex outdoor stress environments) and infers the process parameter combinations (such as warp tension and weft density) that meet the target strength. This transforms the detection data from a "basis for quality determination" into an "input for production process optimization." For example, if the predicted tear strength of a batch of fabric fluctuates, the cloud system can automatically trace the abnormality to the loom tension parameters and generate a specific adjustment plan.

[0120] Furthermore, the method further includes an environment adaptive calibration step:

[0121] A mapping relationship table between temperature and humidity and material strength attenuation was established, and the test results were dynamically corrected through Gaussian process regression. When the ambient humidity exceeded the preset humidity sensitivity threshold, a random correction coefficient was applied to the predicted tear strength value of cotton fabrics.

[0122] The preset humidity sensitivity threshold may include but is not limited to 80% RH, 70% RH, etc.

[0123] The illumination invariance feature extraction algorithm is used to separate the brightness component in the HSV color space, and the influence of color temperature fluctuation on fiber texture recognition is eliminated through histogram normalization.

[0124] In some embodiments of the present application, the influence of environmental interference on the detection results is dynamically corrected by presetting temperature and humidity thresholds and light processing algorithms.

[0125] For example, in a high-humidity workshop detection scenario (such as the finishing section of a cotton spinning mill): when establishing a temperature-humidity mapping relationship table, the preset humidity sensitivity threshold for cotton fabrics is 80% RH (when this threshold is exceeded, the cotton fiber absorbs moisture, resulting in a significant decrease in strength). The Gaussian process regression model dynamically updates the correction coefficient based on the historical data of the past 24 hours (including parameters such as temperature and humidity, fabric strength, and production batches). For example, when the measured humidity is 85%, the predicted strength value is automatically multiplied by a correction coefficient of 0.9 to reflect the decrease in strength after moisture absorption.

[0126] In the illumination invariance processing, when separating the brightness component in the HSV color space, the preset color temperature fluctuation tolerance range is 5000K-7000K (covering the common color temperature range of workshop LED lighting and natural light), and the brightness value is normalized to the grayscale range of 120-180 through histogram normalization (eliminating the impact of brightness differences under different light sources on fiber texture recognition, for example, to prevent shadow areas from being mistakenly identified as fiber breaks).

[0127] Based on the embodiments provided herein, the environmental adaptive calibration step uses Gaussian process regression to establish a dynamic mapping between temperature and humidity and material strength attenuation. For moisture-sensitive materials such as cotton fabrics, a correction factor is automatically applied when the ambient humidity exceeds a threshold, addressing the problem of test result deviations caused by environmental parameter fluctuations in traditional detection methods. For example, in a high-humidity workshop environment, the hygroscopic expansion of cotton fibers can alter their mechanical properties. This dynamic correction mechanism compensates for these changes in real time, ensuring the stability of detection results. The illumination-invariant feature extraction algorithm separates the luminance component in the HSV color space and normalizes it to a histogram, eliminating the impact of color temperature fluctuations from different light sources (such as sunlight, LEDs, and sodium lamps) on fiber texture recognition. This prevents defect misidentification caused by varying lighting conditions in traditional visual inspection (e.g., shadow areas being mistakenly identified as fiber breaks). The combination of these two technologies enables the detection system to maintain stable performance in complex industrial environments (e.g., temperature and humidity fluctuations of ±20% and light color temperature variations of >1000K), eliminating the need for the constant temperature and humidity laboratory environment required for traditional testing. This significantly improves the device's adaptability to different scenarios and the reliability of detection results.

[0128] Optionally, as shown in 3, the present application provides a textile fabric tear strength detection system, comprising:

[0129] Detection mode switching module 301, used to synchronously collect fabric microstructure, macromechanics and composition data through a multimodal sensing system, and dynamically switch detection modes based on real-time recognition results;

[0130] The tear strength prediction module 302 is used to input the spatiotemporally correlated defect data into a prediction model enhanced by physical constraints to generate crack growth trends and tear strength predictions. The prediction model integrates Griffith's crack growth theory and the hidden Markov model to achieve dual constraints of material mechanics and dynamic defect evolution.

[0131] The process optimization module 303 is used to adopt the non-contact detection module and the flexible contact detection module to work together, and output a process optimization solution in combination with the edge and cloud distributed architecture.

[0132] It should be noted that, in the present application, the embodiments implemented on the textile fabric tearing strength detection system side can be referenced with the embodiments implemented on the textile fabric tearing strength detection method side, and this application will not describe them one by one.

[0133] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for testing the tearing strength of textile fabrics, characterized in that: include: The multimodal sensing system simultaneously collects fabric microstructure, macromechanics, and composition data, and dynamically switches detection modes based on real-time recognition results. Inputting spatiotemporally correlated defect data into a prediction model augmented by physical constraints generates crack growth trends and tear strength predictions. This prediction model integrates Griffith's crack growth theory with a hidden Markov model to achieve dual constraints based on material mechanics and dynamic defect evolution. The non-contact detection module and the flexible contact detection module work together, and the edge and cloud distributed architecture are combined to output process optimization solutions.

2. The method for testing the tearing strength of textile fabrics according to claim 1, characterized in that: The implementation of dynamic switching detection mode includes: The fiber molecular bonding information is obtained through a near-infrared spectrometer. Combined with the fabric surface texture captured by an industrial camera, an adaptive feature pyramid network is constructed for classification to output the fabric type and structure. When a knitted fabric is identified, the cyclic tensile load simulation module is activated to simultaneously extract the yarn slip trajectory data and molecular bond strength attenuation data under tensile load. The weights are dynamically assigned through a dual-channel attention mechanism to generate a joint mapping relationship between tensile load and slip displacement. When the fabric is identified as coated, the multi-scale residual convolutional network is activated to fuse the porosity distribution data of the penetrating infrared thermal imaging sequence and the laser profile scanning to locate the critical area of fiber breakage under the coating.

3. The method for testing the tearing strength of textile fabrics according to claim 1, wherein: Construction of prediction models enhanced by physical constraints, including: A Griffith energy release rate threshold is embedded in the state transition probability matrix of the hidden Markov model, and when the predicted energy of the crack propagation path exceeds the fracture toughness of the material, the state transition path is adjusted to conform to physical laws; A joint loss function is designed to dynamically weight the timing prediction error of the hidden Markov model and the energy constraint error calculated by the Griffith crack growth theory, and the matching degree between the crack evolution timing and the physical law is simultaneously optimized through gradient back propagation; According to the optimization result of the joint loss function, a tear strength prediction value that satisfies the dual constraints of stress concentration and energy dissipation is output.

4. The method for testing the tearing strength of textile fabrics according to claim 2, wherein: The implementation of the dual-channel attention mechanism includes: A temporal convolutional network is used to extract temporal features from the yarn slip trajectory data to generate a spatiotemporal encoding vector of the slip displacement; Using a graph convolutional network to model the molecular interaction relationship of the molecular bond strength attenuation data to generate a topological feature vector of the bond strength; The association weights between the spatiotemporal coding vector and the topological feature vector are dynamically calculated through a cross-attention layer to generate an attention distribution map of the joint mapping relationship for indicating parameter adjustment of the cyclic tensile load.

5. The method for testing the tearing strength of textile fabrics according to claim 3, characterized in that: The adjustment of the state transition path includes: Based on Griffith's crack growth theory, the stress intensity factor of the current crack growth direction is calculated to generate the energy release rate constraint boundary; In the Viterbi decoding process of the crack growth hidden Markov model, the state transition paths that exceed the constraint boundary are removed; The probability of state transition paths that conform to physical laws is redistributed to ensure that the crack propagation trend prediction satisfies both the dynamic defect evolution and material fracture mechanics laws.

6. The method for testing the tearing strength of textile fabrics according to claim 1, characterized in that: The non-contact detection module includes: The air-coupled ultrasonic transducer transmits ultrasonic waves of a set frequency, and the wavelet packet decomposition algorithm is used to extract the yarn breakage characteristic frequency band in the echo signal. The acoustic characteristic patterns of fiber debonding and single fiber breakage were distinguished based on the random forest classifier, and the classification results were used as the criterion for defect type. A laser profile scanner is used simultaneously to obtain three-dimensional morphology data of the fabric surface, and a curvature mutation detection algorithm is used to locate stress concentration areas where the stiffness gradient exceeds the set gradient threshold.

7. The method for testing the tearing strength of textile fabrics according to claim 6, characterized in that: The implementation of the curvature mutation detection algorithm includes: Calculate Gaussian curvature of 3D topography data and generate a curvature distribution heat map; The morphological gradient operator is used to extract the curvature mutation boundary and mark the potential tearing starting point; The curvature mutation area is spatially matched with the defect position detected by ultrasonic testing to verify the reliability of the stress concentration area.

8. The method for testing the tearing strength of textile fabrics according to claim 1, wherein: The implementation of the edge and cloud distributed architecture includes: A fault-tolerant detection model is deployed at the edge nodes of the weaving process, and a bit-flip fault-tolerant mechanism is used to process the real-time collected warp and weft yarn breakage data. After receiving the abnormal data, the cloud platform simulates the tearing life in the actual scenario through the digital twin model, and reversely infers the process parameter combination that meets the target strength.

9. The method for testing the tearing strength of textile fabrics according to claim 1, wherein: The method further comprises an environment adaptive calibration step: A mapping relationship table between temperature and humidity and material strength attenuation was established, and the test results were dynamically corrected through Gaussian process regression. When the ambient humidity exceeded the preset humidity sensitivity threshold, a random correction coefficient was applied to the predicted tear strength value of cotton fabrics. The illumination invariance feature extraction algorithm is used to separate the brightness component in the HSV color space, and the influence of color temperature fluctuation on fiber texture recognition is eliminated through histogram normalization.

10. A textile fabric tearing strength testing system, characterized in that: include: Detection mode switching module, used to synchronously collect fabric microstructure, macromechanics and composition data through a multimodal sensing system, and dynamically switch detection modes based on real-time recognition results; A tear strength prediction module, which is used to input spatiotemporally correlated defect data into a prediction model enhanced by physical constraints to generate crack growth trends and tear strength predictions. This prediction model integrates Griffith's crack growth theory and a hidden Markov model to achieve dual constraints based on material mechanics and dynamic defect evolution. The process optimization module is used to use the non-contact detection module and the flexible contact detection module to work together, combining the edge and cloud distributed architecture to output process optimization solutions.

Images