Knitted fabric photochromic performance rapid characterization test method and device

Abstract

The invention provides a method and a device for rapidly characterizing and testing photochromic performance of knitted fabric. The method comprises the following steps of: performing closed-loop light intensity control on an ultraviolet LED (Light Emitting Diode) array, performing tension coupling clamping, performing multi-angle composite photoelectric sensing array, performing dynamic modulation illumination and phase locking detection, performing texture compensation and performing soft measurement on a prediction model. Rapid, quantitative and repeatable characterization of the photochromic performance of the knitted fabric is achieved, key indexes such as steady-state color change and color difference, color change / fading rate and fatigue coefficient are output, and the problems that in the prior art, testing time is long, knitted structure characteristics are not considered, and tension coupling is lacked are solved.

Description

Technical Field

[0001] This application relates to the field of intelligent detection and photoelectric measurement technology for textiles, and in particular to a rapid characterization test method and apparatus for the photochromic properties of knitted fabrics. Background Technology

[0002] Photochromic textiles are a class of functional materials that undergo reversible color changes under ultraviolet light or specific wavelengths of light, and are widely used in anti-counterfeiting, smart wearables, and ultraviolet indicator applications. With the rapid development of smart textile technology, the application of photochromic textiles in fashion apparel, functional protective equipment, and outdoor sports gear is increasing. Knitted fabrics, in particular, have become an important carrier for photochromic functional textiles due to their excellent elasticity and comfort. Knitted fabrics exhibit a distinct loop structure, unique pore distribution, and thickness variation characteristics, resulting in varying degrees of stretching and deformation during actual wear. These structural characteristics significantly influence their photochromic performance.

[0003] Currently, the performance testing of photochromic textiles mainly employs the following technical approaches. The first is the standard light source excitation method, using standard sunlight or a standard D65 light source as the excitation source. The photochromic performance is evaluated by measuring the color difference before and after color change and the color difference during the recovery process. Typical standard long-term tests can refer to existing standards such as AATCC Test Method 30 or ISO 105-B02, involving continuous illumination and dark recovery. The color difference curve over time is calculated through spectrophotometry. The second approach uses a color-changing performance testing device, equipped with a standard light source, ultraviolet lamp, constant temperature heating and cooling stage, and a camera. Image processing is used to obtain the fabric's color change curve over time, which is used to analyze the color-changing response speed of heat-sensitive or photosensitive fabrics. The third approach is the dedicated spectrophotometer method, which focuses on spectral accuracy and theoretical model fitting, establishing a color hysteresis curve for performance evaluation.

[0004] However, this existing method has the following drawbacks: (1) The standard light source excitation method is greatly affected by the environment and cannot perform fine closed-loop control of light intensity, and is not adaptable to different fabric structures; (2) Existing color change performance testing devices focus on temperature control and image acquisition, and do not compensate for the three-dimensional light field distribution of knitted structures under stretching and curling states, lack dynamic modulation light and electrical phase-locked measurement methods, and the test time is relatively long; (3) Dedicated spectrophotometer methods usually require a long irradiation and recovery time, the equipment is complex and the cost is high, and they are mostly for fibers and thin fabrics. (3) Small-scale testing of fabrics is not user-friendly for online or rapid evaluation of wide-width knitted fabrics; (4) Most existing detection methods are based on flatness, static irradiation and single-angle observation, ignoring the real photochromic performance of knitted fabrics under wearing tension; (5) Most existing instruments use continuous irradiation plus static color measurement, without utilizing dynamic modulation light source and electrical synchronous detection to improve the signal-to-noise ratio and shorten the test time; (6) Most algorithms are traditional color difference calculations, lacking machine learning prediction models based on short-time dynamic signals, making it difficult to reliably infer long-term color fatigue performance and steady-state color difference index within tens of seconds.

[0005] In summary, existing technologies suffer from technical problems such as long testing time for photochromic properties of knitted fabrics, failure to consider the structural characteristics of knitted fabrics, lack of tension coupling control, and inability to achieve rapid and accurate prediction. Summary of the Invention

[0006] This invention provides a rapid characterization test method and apparatus for the photochromic properties of knitted fabrics, which can achieve rapid, quantitative and repeatable characterization of the photochromic properties of knitted fabrics, and solves the technical problems of long test time, lack of consideration for knitted structural characteristics and lack of tension coupling in the prior art.

[0007] Firstly, this invention provides a rapid characterization and testing method for the photochromic properties of knitted fabrics. This method establishes test conditions for the knitted fabric under controllable tension through sample preparation and tension setting steps. The knitted fabric sample is mounted on a tension control clamping mechanism, and the fabric is driven by a motor to achieve a preset elongation rate in both the transverse and longitudinal directions. The set tension is stably maintained through closed-loop control feedback from a tension sensor. Next, a planar light intensity closed-loop calibration step is performed. The ultraviolet LED array is activated with an initial driving current. The output signal of the reference ultraviolet light intensity sensor is read and converted by an ADC. The processor executes a closed-loop control algorithm to adjust the LED driving current, causing the measured light intensity to converge to the target light intensity and maintain stability. Then, a dynamic modulation illumination and synchronous sampling step is performed. The ultraviolet LED array is switched to a square wave or step-modulation mode with a modulation frequency of 0.2-1Hz and a duty cycle of 50%. Simultaneously, a multi-channel photoelectric sensor array is activated to synchronously acquire the output of each channel at the sampling frequency. Further steps include phase-locked detection and multi-dimensional feature extraction. The average value of each channel signal is calculated, and in-phase and quadrature components are obtained by multiplying and integrating with the modulation reference signal to achieve digital phase-locked detection. Multi-dimensional features related to the dynamic response of photochromic colors are extracted. Subsequently, a soft-measurement model calculation step for photochromic performance is performed. The extracted multi-dimensional feature vectors are preprocessed using Z-score normalization and input into a pre-trained machine learning model, outputting estimated values ​​for steady-state color difference, color change rate constant, fading rate constant, and fatigue coefficient. Finally, a rapid evaluation index calculation step is performed. Based on the estimated performance indicators, a rapid evaluation index for the photochromic properties of knitted fabrics is calculated, and quality grades are classified. In summary, this invention establishes a tension coupling testing system for knitted fabrics, achieving photochromic performance characterization under actual wearing stretching conditions. This overcomes the limitations of traditional flat testing. Furthermore, by combining closed-loop light intensity control, dynamic modulation illumination, phase-locked detection, and a soft-measurement prediction model, it improves testing accuracy and repeatability, providing an innovative technical solution for rapid quality control of the photochromic performance of knitted fabrics.

[0008] Preferably, the phase-locked detection and multi-dimensional feature extraction steps include calculating the signal I of each channel. i (t) Average value I over one or more modulation periods i,DC The in-phase component X of the same frequency component is obtained by multiplying and integrating with the modulation reference signal. i With orthogonal component Y i Calculate the rise time t of the modulated light at the instant it is turned on. rise Calculate the fall time t after the photoelectric signal is extinguished, given the time required for the photoelectric signal to reach 90% of its steady state. fall Calculate the hysteresis area A within the modulation period, which is the time required for the signal to attenuate to 10% of its initial value. hys The fatigue characteristic is represented by the area enclosed by the rising and falling curves, and the response peak A of multiple consecutive modulation cycles.N Perform exponential fitting A N = A0 × exp(-λ short The short-time attenuation constant λ is obtained by multiplying N by (× N). short The input feature used for the soft measurement model is the short-time decay constant and the long-term fatigue coefficient F. fatigue A mapping relationship exists. As described above, this invention improves the signal-to-noise ratio and anti-interference capability of the signal through digital phase-locked loop detection technology and multi-dimensional dynamic feature extraction, achieving quantitative characterization of the dynamic response process of photochromism. In particular, the quantitative extraction of short-term fatigue characteristics provides crucial data support for long-term performance prediction.

[0009] Preferably, the phase-locked detection and multi-dimensional feature extraction steps further include a texture compensation mechanism. Before formal testing, spatial uniformity data at each location is collected by a multi-channel array through a short-duration low-intensity scan. The standard planar board uses a barium sulfate standard white board with a reflectivity of 99%. The formula for calculating the texture compensation coefficient is C. tex,i = R std / R i,avg , where R std R represents the average response value of a standard whiteboard. i,avg The average response value of the i-th channel on the surface of the knitted fabric is multiplied by C during feature calculation. tex,i Normalization is performed to reduce the difference between local dark and bright areas caused by the knitted coil structure. In summary, this invention solves the problem of uneven light field caused by the coil structure of knitted fabrics through a texture compensation mechanism, achieving adaptive compensation for the special structural characteristics of knitted fabrics and improving the accuracy and consistency of knitted fabric testing.

[0010] Preferably, the closed-loop control algorithm in the light source plane intensity closed-loop calibration step is as follows: Among them I n K is the digitally processed LED drive current control quantity. p The proportionality coefficient is 0.1-0.5, K i E is the integral coefficient between 0.01 and 0.1. n,uv,set E is the set value for the target light intensity after digital processing. n,uv,meas The actual measured illuminance is digitized, with a control cycle of 50ms. The control algorithm is executed in an embedded digital processor, and all physical quantities are converted into dimensionless digital quantities for calculation after ADC conversion. In summary, this invention achieves closed-loop control of the ultraviolet LED array light intensity through a digital PI control algorithm, eliminating the impact of environmental changes and device aging on the stability of the light source, ensuring high consistency of test conditions and data comparability, and avoiding the problem of dimension mismatch in physical quantities by executing the algorithm in the digital domain.

[0011] Preferably, in the sample preparation and tension setting steps, the knitted fabric sample is conditioned for at least 4 hours under standard temperature and humidity conditions of 20°C and 65%RH. The tension control clamping mechanism adopts a four-sided adjustable tension clamp and is equipped with a fine-tuning mechanism driven by a motor-screw or stepper motor. The preset elongation rate is any value among 5%, 10%, or 15%, corresponding to simulated light, moderate, and heavy stretching states. In summary, this invention establishes a test environment simulating actual wearing conditions through standardized sample pretreatment and tension control, ensuring that the test results effectively characterize the photochromic properties of knitted fabrics under real-world usage conditions.

[0012] Preferably, in the dynamic modulation illumination and synchronous sampling step, the peak wavelength of the ultraviolet LED array is 365nm or 395nm, the sampling frequency is 100Hz, and the multi-angle composite photoelectric sensor array includes four integrated RGB three-channel color sensors arranged in a ring 100mm above the sample. Each sensing unit is located at azimuth angles of 0°, 90°, 180°, and 270° relative to the sample center, with a zenith angle of 45°. The multi-channel electrical signal acquisition is performed by observing the reflected light forming a three-dimensional reflection field on the surface of the knitted fabric at different incident angles and azimuths. In summary, this invention, through the geometric arrangement of the multi-angle composite photoelectric sensor array, achieves comprehensive acquisition of the three-dimensional reflection field of the knitted fabric, overcoming the limitations of single-angle observation.

[0013] Preferably, in the calculation step of the soft measurement model for photochromic performance, the machine learning model adopts a 3-layer feedforward neural network. The 18 nodes of the input layer correspond to the preprocessed feature vector, the 12 nodes of the hidden layer use the ReLU activation function, and the 4 nodes of the output layer correspond to ΔE. ∞,est k color,est k fade,est F fatigue,est The training uses the mean squared error loss function MSE = (1 / n)∑(y true - y pred ) 2 With the Adam optimizer, feature preprocessing employs the Z-score normalization method F. norm = (F - μ) / σ, where the training dataset contains 150 samples of knitted photochromic fabrics with different formulations and their corresponding standard long-term test data. As described above, this invention establishes an accurate mapping relationship from short-term dynamic features to long-term performance indicators through a clearly defined neural network structure and training details, achieving fast and accurate performance prediction.

[0014] Preferably, in the rapid evaluation index calculation step, the rapid evaluation index Q for photochromic changes in knitted fabrics is... knit The calculation formula is Q knit = α × (ΔE ∞,est / ΔE ref) + β × (k color,est / k color,ref ) + γ ×(F fatigue,ref / F fatigue ,est), where ΔE ref =20 is the reference steady-state color difference value, k color,ref =0.5 s^-1 is the reference color change rate constant, F fatigue,ref =0.2 is the reference fatigue coefficient value, α=0.4 is the steady-state color change weighting coefficient, β=0.4 is the color change rate weighting coefficient, and γ=0.2 is the fatigue resistance performance weighting coefficient, according to Q knit The values ​​are divided into three quality levels: A, B, and C. The fatigue item uses a reciprocal form to ensure that a smaller fatigue coefficient results in a higher evaluation index. The aforementioned thresholds can be adjusted based on statistical experience from historical samples. In summary, this invention establishes a comprehensive evaluation system for the photochromic performance of knitted fabrics, achieving weighted fusion and hierarchical classification of multiple indicators. The reciprocal design of the fatigue item ensures that better fatigue resistance results in a higher evaluation index, demonstrating good logical consistency. The grading thresholds are determined based on the statistical distribution of the samples.

[0015] Secondly, this invention provides a rapid characterization and testing device for the photochromic properties of knitted fabrics, including an ultraviolet LED array light source module for providing ultraviolet light irradiation with a peak wavelength of 365nm or 395nm; a closed-loop light intensity control module including a reference ultraviolet light intensity sensor, a signal amplification circuit, an ADC converter, and a control processor for achieving closed-loop control of light intensity and maintaining stability; a tension control clamping mechanism including a four-sided adjustable tension clamp, a motor drive device, and a tension sensor for applying controllable tension to the knitted fabric and maintaining stability; a multi-angle composite photoelectric sensor array including multiple integrated RGB three-channel color sensors arranged in a ring for acquiring the three-dimensional reflection field of the knitted fabric surface and converting it into multi-channel electrical signals; a data acquisition and signal processing module including a multi-channel low-noise amplifier, a multi-channel high-speed ADC, and an embedded processor for realizing phase-locked amplification, feature extraction, and soft measurement model calculation; and host computer software and a model library for setting test parameters, displaying test results, calling pre-trained machine learning models, and outputting photochromic performance evaluation indicators.

[0016] Thirdly, the present invention provides an electronic device, the device including a processor and a memory; the memory is used to store a computer program and transfer the computer program to the processor; the processor is used to execute the method as described in the first aspect according to the instructions in the computer program.

[0017] Fourthly, the present invention provides a storage medium for storing computer-executable instructions, which, when executed by a computer processor, are used to perform the method described in the first aspect.

[0018] The above-mentioned beneficial effects of the present invention are as follows: By establishing a tension coupling test system for knitted fabrics, closed-loop light intensity control and dynamic modulation phase-locked detection, multi-angle composite photoelectric sensor array and texture compensation, and a soft measurement prediction model driven by short-time dynamic features, the present invention achieves rapid and accurate characterization of the photochromic performance of knitted fabrics. Compared with the prior art, the core inventiveness of this application lies in constructing a new testing method based on short-time dynamic response prediction of long-term performance. Its technical solution introduces dynamic modulation illumination (0.2-1Hz square wave ultraviolet light) and digital phase-locked detection to acquire the transient response curve of the photochromic color change of knitted fabrics with high signal-to-noise ratio within tens of seconds, and extracts dynamic features such as rise time, hysteresis area, and short-time decay constant from it; furthermore, it is the first to apply a soft measurement model for textile materials to this field, and through a pre-trained neural network model, establishes the above-mentioned short-time dynamic features and the steady-state color difference (ΔE) obtained by traditional long-cycle testing. ∞ This method establishes a quantitative mapping relationship between the color change / fading rate constant and the fatigue coefficient, enabling rapid prediction of key performance indicators without prolonged light exposure. The effectiveness of this method is achieved through a tension coupling clamping mechanism with adjustable elongation, ensuring the test conditions closely match actual wearing conditions; and a digital texture compensation algorithm for knitted coil structures, eliminating the interference of surface non-uniformity on optical measurements. These solutions collectively address the technical challenge of rapidly and quantitatively characterizing the photochromic properties of knitted fabrics. Attached Figure Description

[0019] Figure 1 A schematic flowchart of the rapid characterization test method for photochromic properties of knitted fabrics provided by the present invention.

[0020] Figure 2 This is a schematic diagram of the system architecture of the testing device provided by the present invention. Detailed Implementation

[0021] The following description and accompanying drawings fully illustrate specific embodiments of this application to enable those skilled in the art to practice them. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operations may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. The scope of embodiments of this application includes the entire scope of the technical solutions and all available equivalents of the technical solutions. In this document, each embodiment may be referred to individually or collectively with the term "invention," which is merely for convenience and, if more than one invention is disclosed, is not intended to automatically limit the scope of the application to any single invention or inventive concept. Relational terms such as "first" and "second" are used herein only to distinguish one entity or operation from another, without requiring or implying any actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed. The various embodiments in this document are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the structures, products, etc., disclosed in the embodiments, since they correspond to the disclosed parts, the descriptions are relatively simple; relevant details can be found in the method section.

[0022] Before describing the technical solution in detail, the terms used in this application will be explained first:

[0023] Photochromic properties: refers to the ability of textiles to undergo reversible color changes under ultraviolet light irradiation, and is mainly characterized by the following quantitative parameters:

[0024] Steady-state color change ΔE ∞ (The maximum color difference value achieved under continuous illumination, expressed as ΔE) ∞ = sqrt((L* final - L* initial ) 2 + (a* final - a* initial ) 2 + (b* final - b* initial ) 2 )Calculate, where L*, a*, and b* are the luminance, red-green hue, and blue-yellow hue values ​​in the CIE LAB color space, respectively; the color change rate constant k color (Characterizing the rate of change from the initial state to the color-changing state, by analyzing the color difference time curve ΔE(t) = ΔE) ∞× (1 - exp(-k color The fading rate constant k is obtained by exponential fitting of (× t) (unit: s^-1, where t is the time variable); fade (Characterizing the rate of recovery from the discolored state to the initial state, by analyzing the recovery curve ΔE(t) = ΔE0 × exp(-k) fade The result is obtained by fitting (× t), with units of s^-1, where ΔE0 is the initial color difference value at the fading stage; fatigue coefficient F fatigue (A parameter characterizing the degree of performance degradation after long-term, repeated color-changing cycles; the smaller the value, the better the fatigue resistance.)

[0025] Phase-locked detection (PLD): A signal processing technique that extracts the in-phase components of a specific frequency range by multiplying and integrating the signal under test with a reference modulation signal. and orthogonal components Where T is the integration period, ω is the modulation angular frequency, and I i (t) represents the time-series signal of the i-th channel, where cos(ωt) and sin(ωt) are the orthogonal components of the reference signal. Phase-locked detection can significantly improve the signal-to-noise ratio and effectively suppress ambient light interference and electronic noise.

[0026] Tension coupling test: This refers to applying controllable transverse and longitudinal tension to knitted fabrics during photochromic performance testing to simulate stretching deformation under actual wearing conditions. It is achieved through a four-sided frame clamping mechanism, with the transverse tension T... warp and longitudinal tension T weft It can be adjusted independently, and the tension is indirectly controlled by the elongation, elongation ε = (L stretched - L original ) / L original × 100%, where L original L is the original length. stretched This is the length after stretching.

[0027] Soft measurement model: A machine learning-based indirect measurement technique that achieves rapid and accurate prediction of target variables by establishing a mathematical mapping model between easily and quickly measurable input features and target variables that are difficult to measure directly or require long-term measurement. The neural network model used in this application is a 3-layer feedforward network. The network weights are updated using a backpropagation algorithm, and the loss function is the mean squared error (MSE) = (1 / n)∑(y true - y pred ) 2 Where n is the number of samples, y true For the true value, y pred These are predicted values.

[0028] Short-time fatigue characteristic quantification: Short-time characteristic parameters are obtained by analyzing the attenuation of response peaks over multiple consecutive modulation cycles. The specific steps are as follows: (1) Extract the response peak A of each modulation cycle. N (2) Establish the exponential decay model A, where N is the period number; N = A0 × exp(-λ short × N), where A0 is the initial peak value, λ short (3) λ is the short-time decay constant, and N is the period number; λ is obtained by least squares fitting. short Value, goodness-of-fit requirement R 2 >0.85; (4) This short-term decay constant is used as an input feature of the soft measurement model, and is related to the long-term fatigue coefficient F fatigue A mapping relationship is established using a neural network.

[0029] Texture compensation: Digital correction processing for localized light field inhomogeneities caused by the loop structure of knitted fabrics. The standard planar plate uses a barium sulfate standard white plate with a reflectivity of 99%±1%, exhibiting Lambertian scattering characteristics. The compensation coefficient is calculated using the formula C. tex,i = R std / R i,avg , where R std R represents the average response value of a standard whiteboard under the same lighting conditions. i,avg This represents the average response value of the i-th channel during pre-scanning on the knitted fabric. The compensated feature value is the original feature value multiplied by the corresponding compensation coefficient.

[0030] Dynamically modulated illumination: A square-wave modulated ultraviolet LED array light source is used, with the modulation frequency determined based on the photochromic reaction kinetics. Photochromic reactions typically follow first-order kinetics, with a reaction time constant τ ranging from 1 to 10 seconds. To fully excite the color-changing response while ensuring testing efficiency, a modulation period of 2-5 times the time constant is selected, corresponding to a modulation frequency of 0.2-1 Hz. The duty cycle is set to 50% to ensure that the lighting and extinguishing times are equal.

[0031] Multi-angle composite photoelectric sensor array: A ring array composed of four photoelectric sensors is used to collect the three-dimensional reflected light field of the knitted fabric surface. The sensor arrangement geometry is as follows: 100mm height from the sample surface, azimuth angles relative to the sample center of 0°, 90°, 180°, and 270°, and a zenith angle of 45° for all sensors. Each sensor integrates three RGB channels, totaling 12 measurement channels, with a sampling frequency of 100Hz and a dynamic range of 16 bits, enabling it to capture the anisotropic reflection characteristics of the knitted fabric caused by the coil structure from all directions.

[0032] Standard long-term testing: Performance testing of photochromic textiles is conducted in accordance with AATCC Test Method 30-2017 or similar standards. This includes continuous ultraviolet irradiation of the sample under controlled conditions until a steady state is reached, followed by recovery observation in the dark. The entire process lasts 24-72 hours. The Lab value at each time point is measured using a spectrophotometer, and the color difference ΔE is calculated. A color difference-time curve is plotted, and the true values ​​of each performance parameter are obtained by fitting an exponential function.

[0033] Digital PI control algorithm: Proportional-Integral controller, control equation is , where I n For the digital control output converted by ADC, K p K is the proportionality coefficient. i e is the integral coefficient. n (k) represents the digitization error at the current time k, ∑e n (j) represents the accumulation of historical errors. All physical quantities are converted into dimensionless digital quantities by the ADC and then processed by the digital processor, avoiding the problem of inconsistent dimensions of physical quantities. The proportional term provides fast response, and the integral term eliminates steady-state error.

[0034] Z-score standardization: A data standardization method, the formula is F. norm = (F - μ) / σ, where F is the original eigenvalue, μ is the eigenmean, and σ is the eigenstandard deviation. norm These are the standardized feature values. The standardized data have a mean of 0 and a variance of 1.

[0035] Rapid Evaluation Index: A unitless index for comprehensively evaluating the photochromic properties of knitted fabrics, calculated using the formula Q. knit = α × (ΔE ∞,est / ΔE ref ) + β × (k color,est / k color,ref ) + γ × (F fatigue,ref / F fatigue,est ), where α, β, and γ are weighting coefficients, satisfying α + β + γ = 1; ΔE ref k color,ref F fatigue,ref The fatigue coefficient is used as a reference benchmark; the fatigue coefficient is used in reciprocal form to ensure that the smaller the fatigue coefficient, the higher the evaluation index.

[0036] Digital processing: refers to the process of converting physical quantities into digital quantities through an analog-to-digital converter (ADC). The converted digital quantities are dimensionless values, which facilitates mathematical operations in digital processors and avoids the problem of inconsistent dimensions when different physical quantities are directly calculated.

[0037] Soft measurement model data training methods:

[0038] To ensure the prediction accuracy of the soft measurement model, this invention adopts the following specific model construction and training process:

[0039] Training dataset construction: 150 samples of knitted photochromic fabrics with different formulations and weave structures were selected, including different types of photochromic materials such as spirulina, azobenzene, and spiropyran, and covering different knit structures such as single-sided knit, double-sided knit, and rib. Each sample underwent a standard 72-hour long-cycle test to obtain the true ΔE. ∞ k color k fade F fatigue The values ​​are used as training labels, and the feature vectors obtained by the 30-60 second short-term test of this method are used as input data.

[0040] Feature vector composition: The feature vector F of each sample contains 18 elements: 4 time features (t... rise t fall A hys , λ short ), 12 phase-locked loop features (4 channels × 3 RGB sub-channels X i and Y i Components, 12 in total), 1 texture compensation feature (average C tex ), 1 tension status indicator.

[0041] Data preprocessing: Z-score normalization method F norm = (F - μ) / σ, where μ and σ are the mean and standard deviation of each feature in the training set, respectively. The standardized feature distribution has a mean of 0 and a variance of 1, which is used to achieve neural network convergence.

[0042] Neural network structure: 18 nodes in the input layer, 12 nodes in the hidden layer using the ReLU activation function f(x) = max(0,x), and 4 nodes in the output layer corresponding to ΔE. ∞,est k color,est k fade,est F fatigue,est Without an activation function, it produces a linear output.

[0043] Training process: The Adam optimizer was used with an initial learning rate of 0.001, β1=0.9, and β2=0.999. The batch size was set to 16, and training lasted for 300 epochs. The loss function used was the mean squared error (MSE) = (1 / n)∑(y true - y pred ) 2Where n is the number of samples. The training set, validation set, and test set are divided in a ratio of 7:1.5:1.5. An early stopping mechanism is used, and training stops when the loss on the validation set does not decrease for 20 consecutive epochs.

[0044] The following is the technical solution of this application:

[0045] This invention provides a rapid characterization and testing method for the photochromic properties of knitted fabrics. For example... Figure 1 As shown, the method includes the following detailed steps:

[0046] Step 101: Sample Preparation and Tension Setting

[0047] A 100mm × 100mm knitted fabric sample was selected and conditioned in a standard environmental chamber at 20℃±2℃ and 65%±5%RH for at least 4 hours to ensure moisture balance. The sample was then mounted on a four-frame tension control clamping mechanism, which includes four adjustable clamping frames, each equipped with a force sensor to monitor the clamping force. Independent tension control in the transverse and longitudinal directions was achieved via a stepper motor-driven lead screw mechanism, with a position control accuracy of ±0.1mm. The elongation rate was set to 5%, 10%, or 15% according to testing requirements, corresponding to light, medium, and heavy tension states. Tension control employed a digital PI closed-loop algorithm, with the error between the set value and the actual value less than ±2%, and a holding time of at least 1 minute to ensure stress relaxation reached a stable state. The aforementioned four-frame tension control clamping mechanism is an existing structural design and will not be described in detail here.

[0048] Step 102: Closed-loop calibration of light intensity in the light source plane

[0049] An ultraviolet LED array consisting of 36 high-power LED chips, with a peak wavelength of 365nm ± 5nm and a power of 3W per chip, was activated. A UVA-responsive photodiode was used as the reference ultraviolet intensity sensor, positioned 5mm from the sample surface. The sensor output was converted into a digital signal by an adjustable-gain amplifier and a 16-bit ADC.

[0050] The closed-loop control algorithm uses a digital PI controller, and the control equation is: , where the error e n (k) = E n,uv,set - E n,uv,meas (k), I n K is the digital LED drive current control quantity converted by ADC. p =0.2 is the proportionality coefficient, K i =0.05 is the integral coefficient, E n,uv,set E is the digital target light intensity setpoint converted by the ADC. n,uv,measThe actual measured illuminance is obtained via ADC conversion, with a control period of 50ms. The target light intensity is set to 10mW / cm². 2 Closed-loop control is achieved by adjusting the LED drive current, and the light intensity fluctuation is less than ±1% after stabilization.

[0051] Step 103: Dynamically modulated illumination and synchronous sampling

[0052] After the light intensity stabilizes, switch to dynamic modulation mode. The modulation waveform is a square wave with a frequency f. mod The frequency was set to 0.5Hz with a duty cycle of 50%. A multi-angle composite photoelectric sensor array, consisting of four RGB color sensors arranged in a ring, was used. Sensor geometry parameters: 100mm from the sample surface, azimuth angles of 0°, 90°, 180°, and 270°, and zenith angle of 45°. Data acquisition lasted 30-60 seconds, including 15-60 complete modulation cycles.

[0053] Step 104: Phase-locked detection and multidimensional feature extraction

[0054] For 12-channel time series data I i (t) Perform phase-locked loop (PLL) detection. First, calculate the DC component of each channel. Then calculate the in-phase components. and orthogonal components Where T is the integration period, and ω = 2πf mod This is the modulation angular frequency.

[0055] Dynamic feature extraction includes: rise time t rise (Time required for the value to rise from 10% to 90% of its steady state), time of fall t fall (Time required to return from 90% to 10% of the baseline value), Hysteresis area A hys (Calculated using the trapezoidal integral method) ).

[0056] The algorithm for calculating short-time fatigue characteristics is as follows: extract the response peak value A for each modulation cycle. N Establish an exponential decay model Where A0 is the initial peak value, λ short Let λ be the short-time decay constant and N be the period index. The least squares method is used to fit the parameter λ. short As input features.

[0057] Texture compensation processing: using 1mW / cm 2 The knitted fabric was pre-scanned with low-intensity ultraviolet light for 5 seconds, while the response R of a barium sulfate standard white plate was measured. std Calculate the texture compensation coefficient C. tex,i = R std / R i,avg , where Ri,avg The average response value of the i-th channel during pre-scanning of the knitted fabric is normalized by multiplying all feature values ​​by the corresponding compensation coefficient.

[0058] Step 105: Calculation of soft measurement model for photochromic performance

[0059] The extracted feature vector F contains 18 elements, which are then normalized by the Z-score. norm = (F - μ) / σ is then input into a 3-layer feedforward neural network, where μ is the feature mean and σ is the feature standard deviation. Network structure: 18 nodes in the input layer, 12 nodes in the hidden layer (ReLU activation), and 4 nodes in the output layer. The model output ΔE ∞,est k color,est k fade,est F fatigue,est Performance metrics include inference time <50ms.

[0060] Step 106: Calculation of the rapid evaluation index

[0061] Calculate the rapid evaluation index of photochromic changes in knitted fabrics: Q knit = α × (ΔE ∞,est / ΔE ref ) + β ×(k color,est / k color,ref ) + γ × (F fatigue,ref / F fatigue,est ), where α=0.4 is the steady-state color change weighting coefficient, β=0.4 is the color change rate weighting coefficient, γ=0.2 is the fatigue resistance performance weighting coefficient, and ΔE ref =20 is the reference steady-state color difference value, k color,ref =0.5 s^-1 is the reference color change rate constant, F fatigue,ref =0.2 is the reference fatigue coefficient value. The fatigue term is in reciprocal form to ensure that the smaller the fatigue coefficient (the better the fatigue resistance), the higher the evaluation index.

[0062] Quality grade classification: Grade A (Q knit ≥1.2), Grade B (0.8≤Q) knit <1.2), Grade C (Q) knit <0.8), the above threshold is an empirical division determined based on the statistical distribution of 150 training samples.

[0063] Example 1: Rapid Detection and Verification

[0064] A textile company tested a new type of photochromic knitted T-shirt fabric. Test conditions: sample size 100×100mm, conditioned for 4 hours, tension 10% elongation (15N×12N), 365nm ultraviolet LED, light intensity 10mW / cm². 2±0.8%.

[0065] Dynamic modulation: 0.5Hz, 50% duty cycle, 60-second test.

[0066] Multi-angle array: 4 RGB sensors, 100Hz sampling, a total of 7200×12 data points.

[0067] Feature extraction results: t rise =2.3s, t fall =8.7s, A hys =0.45, main channel X1=0.82, Y1=0.15, short-time attenuation constant λ short =0.008 (unit: 1 / cycle), average texture compensation coefficient C tex =0.93.

[0068] Neural network prediction: ΔE ∞,est =24.5, k color,est =0.85s^-1, k fade,est =0.18s^-1, F fatigue,est =0.12.

[0069] Note: Where λ short =0.008 is the instantaneous attenuation constant obtained by fitting short-time modulation data, while the fatigue coefficient F obtained by fitting standard long-cycle test data is... fatigue,true =0.115, compared to the F-value predicted by the soft measurement. fatigue,est The error is 4.3% between 0.12 and 0.12.

[0070] Evaluation Index: Q knit = 0.4×(24.5 / 20) + 0.4×(0.85 / 0.5) + 0.2×(0.2 / 0.12) =0.49 + 0.68 + 0.33 = 1.50, rated as Grade A.

[0071] Verification Result Comparison Table:

[0072]

[0073] All metrics were within acceptable error ranges, validating the method's accuracy. Testing time was reduced from the traditional 72 hours to 60 seconds, demonstrating a significant efficiency improvement.

[0074] Example 2: Digital Control System

[0075] To verify the effectiveness of the digital PI control algorithm, closed-loop control tests were conducted in a real-world system. The target light intensity was set to 10 mW / cm². 2 The corresponding ADC digital value is 2048 (full scale 4096 corresponds to 20mW / cm²). 2Control parameter K p =0.2, K i =0.05, control cycle 50ms.

[0076] Initial state: LED drive current is 0, corresponding to digital quantity I n =0. The light intensity sensor reading is 0, corresponding to the digital quantity E. n,uv,meas =0.

[0077] First control cycle: e n (0) = 2048 - 0 = 2048, I n (1) = 0 + 0.2×2048 + 0.05×2048 = 512.

[0078] Steady-state convergence process: After approximately 20 control cycles (1 second), the system converges to near the target value, with a steady-state error of less than ±1%, corresponding to an ADC digital error of less than ±20. The entire control process is executed in the digital domain.

[0079] Example 3: Production Line Application

[0080] This invention's system has been integrated into a knitted fabric production line. It employs a 30-second rapid detection mode, monitoring 50 batches of products per shift.

[0081] Weekly monitoring results statistics:

[0082]

[0083] System stability: After 168 hours of continuous operation, the downtime is <0.2%, the test repeatability CV is <2.5%, and the consistency with the standard test is 93.2%, meeting the requirements of industrial applications.

[0084] Real-time monitoring through the MES system triggers process adjustment alarms for products below grade B, achieving closed-loop quality control and increasing the product qualification rate by 8.5%.

[0085] like Figure 2 As shown, the testing device mainly includes:

[0086] Light source system: 36 365nm LED chips, 6×6 array, total power 108W, constant current drive (0-3A, 1mA resolution).

[0087] Control system: ARM Cortex-M4 main controller (168MHz, 512KB Flash), supports FreeRTOS real-time system, and all control algorithms are executed in the digital domain.

[0088] Sensing system: UVA reference sensor (315-400nm response), 16-bit ADC converter (0-65535 digital range), RGB array sensor (16-bit ADC, 2.5-76800 Lux).

[0089] Mechanical system: aluminum alloy four-sided frame clamp, stepper motor drive (1.8° step angle), force sensor (50N range, 0.1% nonlinearity).

[0090] Data processing: FPGA multi-channel synchronization (200kHz sampling), embedded processor (phase-locked detection + neural network inference <500ms).

[0091] Host computer software: Qt framework, cross-platform support, graphical interface, model management, real-time monitoring.

[0092] The present invention also provides an electronic device, including an Intel Core i7 processor and 16GB DDR4 memory, for performing test control, data acquisition, feature extraction, model inference, and result output functions. It also provides an SSD solid-state drive storage medium for storing a complete software package, including drivers, test software, machine learning model files, and calibration data.

Claims

1. A rapid characterization and testing method for the photochromic properties of knitted fabrics, characterized in that, include: The sample preparation and tension setting steps involve mounting the knitted fabric sample on the tension control clamping mechanism, driving the knitted fabric to achieve the preset elongation rate in both the transverse and longitudinal directions via a motor, and using a tension sensor feedback closed-loop control to stably maintain the set tension. The light intensity closed-loop calibration steps for the light source plane are as follows: the ultraviolet LED array is started and lit with the initial driving current; the output signal of the reference ultraviolet light intensity sensor is read and converted by the ADC; the processor executes the closed-loop control algorithm to adjust the LED driving current so that the measured light intensity converges to the target light intensity and maintains stability. The dynamic modulation illumination and synchronous sampling steps involve switching the ultraviolet LED array to a square wave or step mode of dynamic modulation, and simultaneously activating a multi-channel photoelectric sensor array to synchronously acquire the output of each channel at a sampling frequency. The phase-locked detection and multi-dimensional feature extraction steps involve calculating the average value of each channel signal, obtaining in-phase and quadrature components by multiplying and integrating with the modulation reference signal to achieve digital phase-locked detection, and extracting multi-dimensional features related to the dynamic response of photochromism. The calculation steps of the soft measurement model for photochromic performance are as follows: the extracted multidimensional feature vector is preprocessed by Z-score standardization and then input into the pre-trained machine learning model to output the estimated values ​​of steady-state color difference, color change rate constant, fading rate constant and fatigue coefficient. The rapid evaluation index calculation steps involve calculating the rapid evaluation index of photochromic properties of knitted fabrics based on estimated performance indicators and classifying the quality level.

2. The rapid characterization and testing method for photochromic properties of knitted fabrics according to claim 1, characterized in that, The phase-locked detection and multi-dimensional feature extraction steps include: Calculate the signal I of each channel. i (t) Average value I over one or more modulation periods i,DC ; By multiplying and integrating with the modulation reference signal, the in-phase component X of the same frequency component is obtained. i With orthogonal component Y i ; Calculate the rise time t of the modulated light switch at the instant it is turned on. rise This is the time required for the photoelectric signal to reach 90% of its steady state. Calculate the fall time t after extinction. fall , is the time required for the signal to attenuate to 10% of its initial value; Calculate the hysteresis area A within the modulation period. hys , which is the area enclosed by the rising and falling curves; The response peak A of multiple consecutive modulation cycles N Perform exponential fitting A N = A0 × exp(-λ short The short-time attenuation constant λ is obtained by multiplying N by (× N). short As input features for soft measurement models.

3. The rapid characterization and testing method for the photochromic properties of knitted fabrics according to claim 2, characterized in that, The phase-locked detection and multi-dimensional feature extraction steps also include a texture compensation mechanism: before the formal test, a short-duration low-intensity scan is performed to collect spatial uniformity data at each location by the multi-channel array; the standard planar plate uses a barium sulfate standard white plate with a reflectivity of 99%; the formula for calculating the texture compensation coefficient is C. tex,i = R std / R i,avg , where R std R represents the average response value of a standard whiteboard. i,avg This represents the average response value of the i-th channel on the surface of the knitted fabric. During feature calculation, each channel feature is multiplied by the aforementioned C. tex,i Normalization is performed to reduce the difference between local dark and bright areas caused by the knitted loop structure.

4. The rapid characterization and testing method for the photochromic properties of knitted fabrics according to claim 1, characterized in that, The closed-loop control algorithm in the light source plane intensity closed-loop calibration step is as follows: ; where I n K is the digitally processed LED drive current control quantity. p A proportionality coefficient of 0.1-0.5, K i E is the integral coefficient between 0.01 and 0.

1. n,uv,set E is the set value for the target light intensity after digital processing. n,uv,meas The actual measured illuminance is digitally processed, with a control cycle of 50ms.

5. The rapid characterization and testing method for photochromic properties of knitted fabrics according to claim 1, characterized in that, In the sample preparation and tension setting steps: the knitted fabric sample is conditioned for at least 4 hours under standard temperature and humidity conditions of 20℃ and 65%RH; the tension control clamping mechanism adopts a four-sided adjustable tension clamping frame, equipped with a fine-tuning mechanism driven by a motor-screw or stepper motor; the preset elongation rate is any value of 5%, 10%, or 15%, corresponding to simulate light, medium, and heavy stretching states. In the dynamic modulation illumination and synchronous sampling steps: the peak wavelength of the ultraviolet LED array is 365nm or 395nm; the sampling frequency is 100Hz; the multi-channel photoelectric sensing array includes four integrated RGB three-channel color sensors arranged in a ring 100mm above the sample; each sensing unit is located at azimuth angles of 0°, 90°, 180°, and 270° relative to the center of the sample, with a zenith angle of 45°, to observe the multi-channel electrical signal acquisition of the three-dimensional reflection field formed by the reflected light on the surface of the knitted fabric at different incident angles and azimuths.

6. The rapid characterization and testing method for the photochromic properties of knitted fabrics according to claim 1, characterized in that, In the calculation steps of the soft measurement model for photochromic performance: the machine learning model adopts a 3-layer feedforward neural network, with 18 nodes in the input layer corresponding to the preprocessed feature vector, 12 nodes in the hidden layer using the ReLU activation function, and 4 nodes in the output layer corresponding to ΔE. ∞,est k color,est k fade,est F fatigue,est Training uses the mean squared error loss function MSE = (1 / n)∑(y true -y pred ) 2 And Adam optimizer; feature preprocessing uses Z-score normalization method F norm = (F - μ) / σ; The training dataset contains 150 samples of knitted photochromic fabrics with different formulations and their corresponding standard long-cycle test data.

7. The rapid characterization and testing method for the photochromic properties of knitted fabrics according to claim 1, characterized in that, In the rapid evaluation index calculation step: the rapid evaluation index Q of the photochromic knitted fabric. knit The calculation formula is Q knit =α × (ΔE ∞,est / ΔE ref ) + β × (k color,est / k color,ref ) + γ × (F fatigue,ref / F fatigue,est ); where ΔE ref =20、k color,ref =0.5 s^-1、F fatigue,ref =0.2 is the reference level sample index, α=0.4, β=0.4, γ=0.2 are weighting coefficients; according to the aforementioned Q knit The values ​​are divided into three quality levels: A, B, and C. The fatigue item uses a reciprocal form to ensure that the smaller the fatigue coefficient, the higher the evaluation index. The above thresholds can be adjusted based on statistical experience from historical samples.

8. A rapid characterization and testing device for the photochromic properties of knitted fabrics, employing the rapid characterization and testing method for the photochromic properties of knitted fabrics as described in any one of claims 1-7, characterized in that, include: Ultraviolet LED array light source module, used to provide ultraviolet light irradiation with a peak wavelength of 365nm or 395nm; The closed-loop light intensity control module includes a reference ultraviolet light intensity sensor, a signal amplification circuit, an ADC converter, and a control processor, which are used to realize closed-loop control of light intensity and maintain stability. The tension control clamping mechanism includes a four-sided adjustable tension clamping frame, a motor drive device, and a tension sensor, used to apply controllable tension to the knitted fabric and maintain stability. A multi-angle composite photoelectric sensor array, including multiple integrated RGB three-channel color sensors arranged in a ring, is used to collect the three-dimensional reflection field of the knitted fabric surface and convert it into multi-channel electrical signals; The data acquisition and signal processing module includes a multi-channel low-noise amplifier, a multi-channel high-speed ADC, and an embedded processor, used to realize lock-in amplification, feature extraction, and soft measurement model calculation; The host computer software and model library are used to set test parameters, display test results, call pre-trained machine learning models, and output photochromic performance evaluation indicators.

9. An electronic device, characterized in that, The device includes a processor and a memory; the memory is used to store a computer program and transfer the computer program to the processor; the processor is used to execute the method as described in any one of claims 1-7 according to the instructions in the computer program.

10. A storage medium, characterized in that, Used to store computer-executable instructions, which, when executed by a computer processor, are used to perform the method as described in any one of claims 1-7.

Images