An extremely thin tape neutral layer offset method and system

Abstract

This invention discloses a method and system for determining the neutral layer offset of an ultrathin strip, relating to the field of measurement. The method includes: surface cleaning of an ultrathin strip sample and constructing a base layer on the side of the ultrathin strip; dispersing nanoparticles in a volatile solvent to form a speckled suspension, and spraying the speckled suspension onto the surface of the base layer to form a discrete speckled layer on the side of the ultrathin strip; applying a load to the ultrathin strip and acquiring a sequence of microscopic images of the deformation process on the side of the ultrathin strip; processing the microscopic image sequence using a digital image correlation algorithm to calculate the longitudinal strain distribution along the thickness direction of the ultrathin strip cross-section; fitting the longitudinal strain distribution and determining the position of the physical neutral layer based on the thickness coordinates corresponding to zero longitudinal strain; and determining the neutral layer offset based on the difference between the physical neutral layer position and the geometric center position of the ultrathin strip. Using this invention can reduce the influence of the observation layer on the original mechanical behavior of the ultrathin strip, which is beneficial to improving the stability of the neutral layer offset measurement results.

Description

Technical Field

[0001] This invention relates to the field of measurement, and more specifically to a method and system for measuring the neutral layer offset of an ultrathin strip. Background Technology

[0002] With the development of precision electronics and aerospace industries, the application of ultra-thin strips (thickness typically less than or equal to 0.2 mm, especially 0.1 mm) is becoming increasingly widespread. In the flattening, straightening, and forming processes of ultra-thin strips, the accurate position of the neutral layer is a core parameter determining processing accuracy and strip shape quality.

[0003] Traditional methods for measuring the neutral layer typically employ artificial spray painting to create speckle patterns, followed by digital image correlation (DIC) measurements. However, for extremely thin strips, the thickness of the conventional paint layer (typically 20-50 μm) is on the same order of magnitude as the substrate thickness. According to the principles of composite mechanics, a reasonably thick paint layer can produce significant "additional stiffness effects" and "bimetallic sheet effects," altering the original mechanical behavior and strain distribution of the extremely thin strip.

[0004] In existing technologies, interference introduced by the observation methods themselves causes deviations between the measured neutral layer offset and the true value. This measurement error neither meets the theoretical verification requirements of modern ultra-precision machining nor guides the optimization of actual production process parameters. Summary of the Invention

[0005] This invention discloses a method and system for measuring the neutral layer offset in ultrathin bands. It addresses at least the problem of measurement errors introduced by observation methods in existing neutral layer measurement techniques, thus ensuring the authenticity of neutral layer measurement data.

[0006] According to one aspect of this application, a method for neutral layer offset of an ultrathin strip is provided, comprising: surface cleaning of an ultrathin strip sample, constructing a base color layer on the side of the ultrathin strip; dispersing nanoscale particles in a volatile solvent, and forming a speckled suspension by shaking; atomizing and spraying the speckled suspension onto the surface of the base color layer using aerosol deposition, and forming a discrete speckled layer after the solvent evaporates; applying a load to the ultrathin strip that has passed quality review, and acquiring a sequence of microscopic images of the deformation process of the side of the ultrathin strip using a microscopic optical system; processing the microscopic image sequence using a digital image correlation algorithm, calculating the microscopic strain field along the thickness direction of the cross section, obtaining a strain distribution data sequence, determining the position where the longitudinal strain is zero by robust fitting, and calculating the neutral layer offset.

[0007] Furthermore, the ultrathin strip sample undergoes surface cleaning treatment, and a base color layer is constructed on the side of the ultrathin strip by using a micro-contact transfer process, which involves vertically dipping the side of the ultrathin strip into a pre-coated uniform primer film on the carrier.

[0008] Furthermore, the thickness of the base color layer is no greater than 3 μm; the particle size of the nanoparticles is 50 nm to 500 nm; and the volatile solvent is at least one of anhydrous ethanol and isopropanol.

[0009] Furthermore, the aerosol deposition is carried out using a pulse spraying method, with a spraying distance of 15cm to 25cm and a spraying air pressure of 0.1MPa to 0.3MPa.

[0010] Furthermore, the microscopic image sequence is processed using a digital image correlation algorithm to calculate the microscopic strain field along the thickness direction of the cross section, and the strain distribution data sequence is obtained by: establishing a coordinate system along the thickness direction of the ultrathin strip based on the reference image and the deformed image; extracting longitudinal strain sampling points at multiple thickness locations within the ultrathin strip solid region; and obtaining a data sequence of longitudinal strain variation with thickness coordinates.

[0011] Furthermore, the robust fitting includes: using a random sampling consensus algorithm to perform linear fitting on the longitudinal strain sampling points, so as to obtain the fitting relationship between longitudinal strain and thickness coordinates after removing outliers.

[0012] Furthermore, it also includes: before acquiring the microscopic image sequence, performing a quality evaluation on the local microscopic images of the discrete speckle layer; when the quality evaluation result does not meet the preset conditions, controlling the re-execution of the speckle layer preparation step.

[0013] Furthermore, the quality evaluation of the local microscopic image of the discrete speckle layer includes: inputting the local microscopic image into a pre-trained convolutional neural network model; and having the convolutional neural network model output a quality score or pass label characterizing speckle particle density, distribution uniformity, grayscale contrast, and background color coverage.

[0014] According to another aspect of this application, a neutral layer offset measurement system for ultrathin strips is also provided, comprising: a sample preparation module for surface cleaning of an ultrathin strip sample, constructing a base color layer on the side of the ultrathin strip, and forming a discrete speckle layer on the surface of the base color layer; a loading and image acquisition module for applying a load to the ultrathin strip after sample preparation and acquiring a sequence of microscopic images of the deformation process of the side of the ultrathin strip; a strain field solving module for processing the microscopic image sequence using a digital image correlation algorithm to obtain the longitudinal strain distribution of the ultrathin strip cross section along the thickness direction; a neutral layer positioning module for fitting the longitudinal strain distribution, determining the physical neutral layer position based on the position corresponding to zero longitudinal strain, and determining the neutral layer offset based on the difference between the physical neutral layer position and the geometric center position; and a quality evaluation module for performing a quality evaluation on the local microscopic images of the discrete speckle layer before acquiring the microscopic image sequence, and outputting a re-sample preparation command when the quality evaluation result does not meet preset conditions.

[0015] In this embodiment, the surface of an ultrathin strip sample is cleaned to create a base layer on the side of the strip. Nanoparticles are dispersed in a volatile solvent and shaken to form a speckled suspension. The speckled suspension is atomized and sprayed onto the surface of the base layer using aerosol deposition, and after solvent evaporation, a discrete speckled layer is formed. A load is applied to the ultrathin strip that has passed quality control, and a microscopic image sequence of the deformation process on the side of the strip is acquired using a microscopic optical system. The microscopic image sequence is processed using a digital image correlation algorithm to calculate the microscopic strain field along the thickness direction of the cross-section, obtaining a strain distribution data sequence. Robust fitting is used to determine the position where the longitudinal strain is zero, and the neutral layer offset is calculated. This invention solves the problem of measurement errors introduced by the observation methods in existing neutral layer measurement methods, ensuring the authenticity of the neutral layer measurement data. Attached Figure Description

[0016] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are illustrative, and the originals and elements are not necessarily drawn to scale. The drawings, which form a part of this application, are used to provide a further understanding of the application, and the illustrative embodiments of the application and their descriptions are used to explain the application and do not constitute an undue limitation of the application. In the drawings:

[0017] Figure 1 This is a flowchart of the method for measuring the neutral layer shift of an ultrathin band based on micron-level undisturbed speckle according to an embodiment of this application.

[0018] Figure 2 These are microscopic morphology images under a micrometer-scale speckle microscope according to embodiments of this application;

[0019] Figure 3 This is a schematic diagram illustrating the calculation principle of the neutral layer offset according to an embodiment of this application. Detailed Implementation

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. It should also be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0021] The technical terms used in the following embodiments will be explained first.

[0022] Robust fitting analysis is a data analysis method that can stably and accurately estimate model parameters even when the data contains noise, outliers (exterior points), or interference. Its core objective is to reduce the negative impact of outliers on the fitting results, thereby improving the reliability and generalization ability of the model.

[0023] The following embodiment provides a method for offsetting the neutral layer of an ultrathin strip, which includes the following steps:

[0024] The ultrathin strip sample undergoes surface cleaning, and a base coat is constructed on the side of the ultrathin strip. In this step, a micro-contact transfer process can be used, which involves vertically adhering the side of the ultrathin strip to pick up a pre-coated uniform primer film from the carrier.

[0025] Nanoparticles are dispersed in a volatile solvent and then shaken to form a speckled suspension.

[0026] Aerosol deposition is used to atomize and spray the speckled suspension onto the surface of the base color layer. After the solvent evaporates, a discrete speckled layer is formed.

[0027] The process involves reviewing local microscopic images of the speckle layer. If the review is successful, the process proceeds to the next step; otherwise, a prompt to re-prepare the sample is generated. Specifically, the local microscopic images are input into a pre-trained neural network model, which outputs the review result. In this step, the acquired local microscopic images of the speckle layer are input into a pre-trained convolutional neural network model. This model extracts and classifies the speckle particle density, distribution uniformity, and background color coverage features in the image, outputting a quality score. If the score is below a preset threshold, a prompt to re-prepare the sample is generated; alternatively, the model outputs a pass or fail label, with a prompt to re-prepare the sample for a fail result. It should be noted that using a neural network model for review is an optional process; other review methods can also be used. There are many ways to use neural network models for review. For example, reviewing local microscopic images of speckle layers: if the review is successful, proceed to the next step; if the review fails, prompt for sample re-preparation. This includes: inputting the acquired local microscopic images of speckle layers into a pre-trained convolutional neural network model, which extracts and classifies the speckle particle density, distribution uniformity, and background color coverage features in the image, and outputs a quality score. If the score is lower than a preset threshold, prompt for sample re-preparation; or, the model outputs a pass or fail label, prompting for sample re-preparation if the score is fail.

[0028] A load was applied to the ultrathin strip that passed the quality audit, and a sequence of microscopic images of the deformation process on the side of the ultrathin strip was acquired using a microscopic optical system.

[0029] The microscopic image sequence is processed using a digital image correlation algorithm to calculate the microscopic strain field along the thickness direction of the cross section, obtain the strain distribution data sequence, and determine the position where the longitudinal strain is zero through robust fitting, and calculate the neutral layer offset.

[0030] In the above steps, a micro-contact transfer process can be used, which involves vertically dipping an extremely thin strip of a uniformly coated primer film onto the carrier.

[0031] There are many ways to perform calculations in this step, such as extracting longitudinal strain data along the thickness direction of the cross-section. The strain distribution line is obtained by least-squares linear fitting, and the coordinates of the intersection point of this line and the zero-strain axis are solved. Calculate the coordinates of the intersection point and the coordinates of the geometric center. The absolute value of the difference, where the absolute value of the difference is the neutral layer offset; extract the longitudinal strain data points along the thickness direction of the cross section. The data points were fitted using a random sampling consensus algorithm via linear regression. Outliers were then iteratively filtered and removed to obtain the optimal strain distribution linear equation. ;make The coordinates of the physical neutral layer can be obtained by solving the problem. The process iterates multiple times until the optimal model parameters containing the most interior points are found: curvature. and intercept .

[0032] Through the above steps, the problem of measurement errors introduced by the observation methods in the existing neutral layer measurement methods is eliminated, thus ensuring the authenticity of the neutral layer measurement data.

[0033] The following description is provided in conjunction with the accompanying drawings and optional embodiments.

[0034] This embodiment provides a method and system for measuring the neutral layer offset of ultrathin strips based on micron-level undisturbed speckle patterns, belonging to the fields of experimental mechanics and precision strip processing and testing. This embodiment addresses the problem of measurement distortion caused by the "added stiffness" and "bimetallic sheet effect" of traditional speckle coatings in ultrathin strips, proposing a undisturbed measurement scheme. The method first constructs a micron-level ultrathin base layer on the side of the ultrathin strip using a micro-contact transfer process, and then atomizes and sprays a suspension of nanoparticles using aerosol deposition, utilizing the characteristics of highly volatile solvents to form a discrete, unbonded speckle layer, achieving "zero stiffness" adhesion of the speckle layer. In the measurement process, a neural network model is introduced to perform real-time verification of the micro-speckle quality, and digital image correlation (DIC) technology is used to calculate the cross-sectional micro-strain field. Combined with the RANSAC robust fitting algorithm and an autoencoder anomaly detection mechanism, the longitudinal zero-strain position is accurately located to obtain the neutral layer offset. This embodiment solves the problems of difficult micro-section sample preparation and large observation interference in ultrathin strips, improving the physical authenticity and accuracy of the measurement.

[0035] The following describes one method provided in this embodiment. Because this method uses a neural network for measurement, it is called a method for measuring the neutral layer migration of an extremely thin band based on micrometer-level perturbationless speckle and a neural network. The steps included in this method are described below. It should be noted that some steps in this method are optional.

[0036] Step S1 involves surface pretreatment of the ultrathin strip sample, including cleaning to remove oil and impurities, ensuring side cleanliness and establishing a clean side observation benchmark. This benchmark can be flexibly tailored to different needs and will not be elaborated upon here. The cleaning method can also be flexibly selected according to actual needs, as long as the pre-defined standard is met after cleaning.

[0037] The definition of "ultra-thin strip" may vary across different fields. For example, the national standard GB / T 15574—2016, "Classification of Steel Products," broadly classifies plate and strip steel into thin plates (less than 3 mm) and thick plates (not less than 3 mm). In scientific research and production, plate and strip steel is classified into five categories based on thickness: ultra-thin strip, thin plate, medium plate, thick plate, and extra-thick plate. In this field, steel plates with a thickness of less than 0.2 mm are referred to as ultra-thin strip (this embodiment is particularly applicable to strip materials with a thickness of 0.1 mm and less than 0.1 mm; unless otherwise specified in the following embodiments, ultra-thin strip refers to strip materials with a thickness of less than or equal to 0.1 mm), steel plates with a thickness of 0.2-3 mm are referred to as thin plates, steel plates with a thickness of 3-20 mm are referred to as medium plates, steel plates with a thickness of 20-60 mm are referred to as thick plates, and steel plates with a thickness of more than 60 mm are referred to as extra-thick plates.

[0038] Step S2 involves constructing a base color layer on the side of an ultra-thin strip using a micro-contact transfer process. The micro-contact transfer process includes vertically dipping the side of the ultra-thin strip into a pre-coated uniform primer film on the carrier.

[0039] There are many methods for constructing a base layer on the side of an ultrathin strip. In this embodiment, a micro-contact process is used, which is briefly described below. Micro-contact transfer is a microfabrication method based on a flexible stamp combined with self-assembled monolayer technology. It falls under the category of soft lithography and can manufacture complex three-dimensional structures and irregular curved surfaces at low cost. It is suitable for pattern transfer on various material substrates. Its core process involves transferring molecules to the substrate surface through contact with a flexible stamp, combined with a washing step to form a self-assembled monolayer pattern. This transfer process is existing technology and will not be described in detail here.

[0040] In this step, optionally, a micro-contact transfer base color is used to prepare the base color layer using a "planar transfer method": the primer is uniformly coated on a flat carrier surface, and the coating thickness is controlled to be uniform; then the side of the ultra-thin strip sample is perpendicularly contacted and dipped into the primer on the carrier, and allowed to dry naturally to form a base color layer with a thickness of micrometers on the side of the ultra-thin strip.

[0041] For example, the primer is a matte white paint with high hiding power, and the carrier is a flat glass plate or polished metal plate with a surface roughness Ra of less than 0.1 μm; the wet film thickness of the primer is controlled at 5-10 μm, and after contact transfer and drying, the dry film thickness formed on the side of the ultrathin strip is less than 3 μm.

[0042] Step S3: Disperse the nano-sized particles in a highly volatile solvent and form a speckled suspension by ultrasonic vibration.

[0043] In this step, a speckle spraying solution can be prepared using a nano speckle suspension: carbon powder particles with a particle size of 100 nm are added to anhydrous ethanol solvent; the mixture is dispersed for no less than 20 minutes using an ultrasonic vibration device to make the carbon powder uniformly suspended in ethanol, forming a nano speckle suspension.

[0044] For example, the mixing ratio of the toner particles to anhydrous ethanol is: 0.05g-0.2g of toner added to every 50mL of anhydrous ethanol; the ultrasonic oscillation frequency is above 40kHz to ensure that the toner agglomerates are completely broken up.

[0045] Step S4: Using aerosol deposition, the speckled suspension is atomized and sprayed onto the surface of the base color layer. After the solvent evaporates, a discrete micron-sized speckled layer is formed. The local microscopic image of the speckled layer is reviewed. If the review is passed, proceed to the next step. If the review is not passed, prompt for re-sample preparation.

[0046] Figure 2These are microscopic morphology images under a micrometer-scale speckle microscope according to embodiments of this application. Figure 2 The microstructure is shown in the figure. Figure 2 The average offset was 0.0135mm, and the maximum offset was 0.0168mm.

[0047] In this step, various parameters can be used to prepare the micron-scale speckle layer. For example, the preparation parameters of the micron-scale speckle layer are: the dry film thickness of the base color layer is less than 3 μm; the nano-sized particles are carbon powder with a particle size of 100 nm; the highly volatile solvent is anhydrous ethanol; the gas pressure for aerosol deposition is set to 0.1 MPa-0.3 MPa; the spraying distance is 15 cm-25 cm; and pulse spraying is used.

[0048] For example, aerosol discrete deposition can be achieved by using a precision spray gun in conjunction with an air pump system, setting the air pressure parameters and spraying distance; the nano-speckled suspension is atomized and sprayed onto an extremely thin base layer on the side; since anhydrous ethanol evaporates rapidly during flight or contact, only carbon powder particles remain attached to the surface of the base layer, forming discrete micron-sized speckles.

[0049] For example, the nozzle diameter of the precision spray gun is 0.2mm-0.5mm, the air pump output pressure is set to 1MPa-3MPa, and the vertical distance between the nozzle and the side of the ultra-thin strip is 15cm-25cm; a pulse spraying method of "small amount, multiple times" is adopted, with an interval of 3-5 seconds between each spraying to ensure that the ethanol evaporates completely.

[0050] There are many ways to conduct a review. In this embodiment, a neural network model can be used for quality review. This step can be called S4a.

[0051] For example, the acquired microscopic images of the speckle layer are input into a pre-trained convolutional neural network (CNN) model. This model extracts and classifies the speckle particle density, distribution uniformity, and background color coverage features in the image, and outputs a quality score. If the score is lower than a preset threshold, it prompts for re-sample preparation. Alternatively, it can output a pass / fail label; if the label is unqualified, it prompts for re-sample preparation. The CNN model employs a lightweight network architecture to meet real-time inference requirements.

[0052] The model uses multiple sets of training data, each set including an image and a label. The label indicates whether the quality score is acceptable or is a quality score. After training, the image to be reviewed is input into the model, and the model outputs a label indicating whether the image score is acceptable or unacceptable or is a quality score.

[0053] Step S5 involves setting up a loading device to apply a load to the ultra-thin strip that has passed quality inspection, and using a microscopic optical system to acquire a sequence of microscopic images of the deformation process on the side of the ultra-thin strip. In this step, the loading device is set up to apply a load to the ultra-thin strip, and a high-resolution sequence of images of the deformation process on the side of the ultra-thin strip is captured in real time using an industrial microscope and image acquisition equipment.

[0054] For example, the industrial microscope is equipped with a coaxial light source and a magnification set to 20x to 50x to ensure that the field of view covers the entire thickness range of the ultrathin band; the resolution of the acquired images is better than 2448×2048 pixels.

[0055] Step S6: The microscopic image sequence is processed using the digital image correlation (DIC) algorithm to solve the cross-sectional micro-strain field (i.e., to calculate the micro-strain field along the thickness direction of the cross-section), obtain the strain distribution data sequence, and determine the position where the longitudinal strain is zero through robust fitting analysis, and calculate the neutral layer offset.

[0056] Digital Image Correlation (DIC), a typical application of this technology, is a non-contact, full-field deformation measurement technique. It achieves displacement and strain measurement by analyzing the similarity of multiple images, and features convenient operation and strong environmental adaptability. DIC technology is widely used in the aerospace field, and its development needs to address the problems of thermal radiation interference and speckle quality degradation under high-temperature environments.

[0057] There are many ways to calculate the neutral layer offset. One such method involves extracting longitudinal strain data along the thickness direction of the cross-section. The strain distribution line is obtained by least-squares linear fitting, and the coordinates of the intersection point of this line and the zero-strain axis are solved. Calculate the coordinates of the intersection point and the coordinates of the geometric center. The absolute value of the difference, where the absolute value of the difference is the neutral layer offset; for example, extracting longitudinal strain data points along the thickness direction of the cross section. The Random Sampling Consensus (RANSAC) algorithm was used to perform linear regression fitting on the data points. Outliers were iteratively filtered out to obtain the optimal strain distribution line equation. ;make The coordinates of the physical neutral layer can be obtained by solving the problem. .

[0058] There are many methods for deleting outliers. For example, a neural network model can be used, which involves iteratively filtering out outliers. This includes: inputting the calculated strain distribution data sequence into a trained autoencoder model; calculating the reconstruction error between the input data and the model's reconstructed data; and if the reconstruction error exceeds a preset threshold, determining that the current calculation result has a physical anomaly and triggering a second fitting or removing the outlier data points.

[0059] Figure 1 This is a flowchart of the method for measuring neutral layer shift in an ultrathin band based on micron-level perturbationless speckle and neural network according to an embodiment of this application. Figure 1 The process includes two main steps. The first step, physical sample preparation and collection (corresponding to steps S1-S5), involves selecting a 0.05mm thick 304 stainless steel precision strip as the sample. Surface treatment: The sample is ultrasonically cleaned with acetone for 3 minutes. Base coat preparation: A high-opacity matte white paint is applied to a glass slide with Ra < 0.1μm, and the wet film thickness is controlled at 8μm. The ultra-thin strip is pressed vertically against the glass slide, lifted, and allowed to air dry naturally, forming a white background layer approximately 2μm thick. Spot spraying: 0.1g of carbon powder with a particle size of 100nm is added to 100mL of anhydrous ethanol and ultrasonically dispersed at 40kHz for 20 minutes. Using a 0.3mm nozzle sprayer, with an air pressure set to 0.2MPa and a distance of 20cm, a "spot spray" method is performed. After the ethanol evaporates, discrete black micro-particle spots are formed. Image acquisition: A 50x industrial microscope equipped with coaxial light, with a resolution of 2448×2048, was used. A sequence of images was acquired at a frame rate of 5fps during the three-point bending loading process. The second major step is algorithm implementation and data analysis (corresponding to step S6).

[0060] The following describes an optional implementation of step S6 (neutral layer intelligent analysis). This optional implementation employs a noise-resistant robust analysis algorithm, and the specific process includes the following sub-steps:

[0061] S600, intelligent auditing of the entire experimental process based on neural network models.

[0062] Pre-audit: As mentioned earlier, a neural network model is used to score the speckle quality. Post-audit: A one-dimensional convolutional autoencoder (1D-CNN Autoencoder) is used to process the calculated neutral layer offset curve. Perform physical plausibility verification. If the reconstruction error exceeds the threshold, the system marks the data points with reconstruction errors exceeding the threshold as "abnormal" and triggers a second fitting or removal.

[0063] S601, Image preprocessing and geometric benchmark establishment.

[0064] Read the reference image before loading and the deformed image after loading. If the two are not the same size, first resample the deformed image based on feature point matching to align its pixels with the reference image.

[0065] Establish an image coordinate system: with the column direction of the image as the X-axis (length direction of the ultrathin strip) and the row direction as the Y-axis (thickness direction of the ultrathin strip).

[0066] The initial position of the ultrathin strip is determined by using the grayscale projection method: the reference grayscale image is averaged in the row direction, and the longest continuous segment with grayscale values ​​higher than a preset threshold (such as 20%-50% of the maximum grayscale) is extracted to obtain the initial upper and lower boundary row indices.

[0067] Boundary refinement using gradient operators: Within the initial boundary neighborhood, the Scharr operator combined with Gaussian filtering is applied to calculate the local gradient peak, accurately locating the upper boundary curve of the ultrathin band. and lower boundary curve .

[0068] Computational geometric centerline :

[0069]

[0070] Simultaneously, calibration coefficients are calculated based on the ratio of the input ultrathin strip physical thickness to the image pixel thickness. .

[0071] S602 Strain Field Construction and Region Masking

[0072] The full-field displacement and strain were calculated using the Digital Image Correlation (DIC) algorithm. A binary mask was constructed based on the upper and lower boundaries determined in S6.1, retaining only the strain data within the extremely thin solid region.

[0073] Extracting the longitudinal strain component within the mask A two-dimensional Gaussian smoothing filter (Sigma parameter adjustable, preferably in the range of 0.5-2.0) is applied to remove high-frequency noise caused by the discrete distribution of micro-speckled particles, resulting in a smooth distribution along each column of the X-axis. data.

[0074] S603 Robust Linear Fitting Based on RANSAC

[0075] For each column (X coordinate) of the ultrathin band, the effective data points Establish a first-order linear deformation model: Shutterstock

[0076]

[0077] Considering the potential for local outliers in microscopic measurements, this embodiment preferably uses the Random Sampling Consensus Algorithm (RANSAC) combined with linear regression for fitting.

[0078] The specific process is as follows: randomly select a subset of data to estimate the model parameters, calculate the residuals from the remaining data points to the model, and mark points with residuals less than a threshold as inliers; iterate multiple times until the optimal model parameters containing the most inliers are found. (curvature) and (Intercept). If the system detects that computing resources are limited or the data quality is extremely high, it can automatically downgrade to the standard least squares method for fitting.

[0079] S604, Neutral layer positioning and smoothing.

[0080] According to the definition of mechanics of materials, let the fitted model contain... The physical neutral layer pixel coordinates corresponding to this column are calculated. :

[0081]

[0082] Discrete calculated for all columns The sequence is processed for continuity. Savitzky-Golay filters or spline interpolation algorithms are used to smooth the sequence, removing NaN values ​​and abrupt changes to generate a continuous neutral layer curve. .

[0083] S605, Offset Calculation and Statistics. Figure 3 This is a schematic diagram illustrating the calculation principle of the neutral layer offset according to an embodiment of this application. The following is in conjunction with... Figure 3 Please provide an explanation.

[0084] Mapping image pixel coordinates to physical space coordinates:

[0085]

[0086]

[0087] Calculate the neutral layer offset at each point along the length of the ultrathin strip. :

[0088]

[0089] The system automatically calculates the average and maximum offsets across the entire field and overlays the results onto the deformation map. It also exports a structured data file (CSV) containing coordinates, boundaries, neutral layer positions, and offsets.

[0090] This embodiment also provides a measurement system for the neutral layer migration of an extremely thin band based on micrometer-level undisturbed speckle. This measurement system is used to implement the above method, wherein the measurement system includes:

[0091] The sample preparation module includes a precision planar carrier and a micro-atomizing spraying device for performing the micro-contact transfer and aerosol deposition processes described above. The acquisition module includes a loading device and an industrial microscope, configured to acquire microscopic images of extremely thin strips with side views. The intelligent processing module (referred to as the processing module) is equipped with a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the following:

[0092] (1) After sample preparation, call the neural network model to perform quality review on the speckle image;

[0093] (2) Perform digital image correlation (DIC) operations;

[0094] (3) Perform neutral layer fitting and offset calculation based on RANSAC algorithm.

[0095] This embodiment also provides a system, referred to as an ultrathin band neutral layer offset system, comprising:

[0096] The preparation module is used to perform surface cleaning treatment on the ultrathin strip sample, construct a base color layer on the side of the ultrathin strip, disperse nano-sized particles in a volatile solvent, and form a speckled suspension by shaking.

[0097] The spraying module is used to atomize and spray the speckled suspension onto the surface of the base color layer using aerosol deposition. After the solvent evaporates, a discrete speckled layer is formed. The preparation module and the spraying module realize the function of the sample preparation module.

[0098] The processing module is used to review local microscopic images of the speckle layer. If the review is successful, the process proceeds to the next step; otherwise, a prompt for sample preparation is given. The local microscopic images are input into a pre-trained neural network model, which outputs the review result. A digital image correlation algorithm is used to process the microscopic image sequence, calculating the microscopic strain field along the thickness direction of the cross-section to obtain a strain distribution data sequence. Robust fitting is used to determine the location where the longitudinal strain is zero, and the neutral layer offset is calculated. The microscopic image sequence is obtained by applying a load to the ultrathin strip that has passed the quality review and using a microscopic optical system to acquire microscopic images of the deformation process on the side of the ultrathin strip.

[0099] Optionally, the preparation module is used for:

[0100] The micro-contact transfer process involves vertically dipping an extremely thin strip of a pre-coated, uniform primer film onto a carrier.

[0101] The processing module is used to: input the acquired speckle layer local microscopic image into a pre-trained convolutional neural network model, which extracts and classifies the speckle particle density, distribution uniformity and background color coverage features in the image, and outputs a quality score. If the score is lower than a preset threshold, it prompts for re-sample preparation; or, the model outputs a qualified or unqualified label, and if unqualified, it prompts for re-sample preparation.

[0102] Optionally, the processing module is used to:

[0103] Extract longitudinal strain data along the thickness direction of the cross section The strain distribution line is obtained by least-squares linear fitting, and the coordinates of the intersection point of this line and the zero-strain axis are solved. Calculate the coordinates of the intersection point and the coordinates of the geometric center. The absolute value of the difference, where the absolute value of the difference is the neutral layer offset; extract the longitudinal strain data points along the thickness direction of the cross section. The data points were fitted using a random sampling consensus algorithm via linear regression. Outliers were then iteratively filtered and removed to obtain the optimal strain distribution linear equation. ;make The coordinates of the physical neutral layer can be obtained by solving the problem. The process iterates multiple times until the optimal model parameters containing the most interior points are found: curvature. and intercept .

[0104] The above embodiments solve at least one of the following problems existing in the prior art:

[0105] 1. Coating stiffening effect: Extremely thin strips (especially metal strips with a thickness of less than 0.1 mm) have extremely low stiffness. The coating thickness formed by traditional spray painting with speckles is usually tens of micrometers, which will form a significant "bimetallic strip effect" and seriously change the original mechanical behavior of the extremely thin strip.

[0106] 2. Difficulty in preparing samples with small cross sections: The thickness of the ultra-thin strip is extremely narrow (on the order of micrometers), making it difficult to form a uniform base color only on the side using traditional spraying methods, and it is also very easy to contaminate the board surface.

[0107] 3. Insufficient observation resolution: Conventional speckle particles are relatively large, and the particles are sparse under the high magnification field of view of industrial microscopes, which leads to the loss of correlation in digital image correlation (DIC) calculation and makes it impossible to accurately measure neutral layer shift.

[0108] 4. Low experimental error tolerance: There is a lack of intelligent review mechanism for the preparation quality of micron-level speckle patterns, resulting in many invalid experiments.

[0109] The above embodiments can eliminate the influence of additional stiffness: by utilizing the high volatility of anhydrous ethanol, only nano-sized carbon powder particles are deposited on the sample surface, without forming a continuous cemented film layer, thus truly realizing the "zero stiffness" and "zero thickness" effect of the speckle layer and ensuring the authenticity of the neutral layer measurement data.

[0110] The above embodiments also solve the problem of micro-section sample preparation: the innovative use of "side-dip" contact transfer process avoids edge accumulation and board surface contamination caused by direct spraying of primer, and can form an ultra-thin background with uniform thickness and neat edges on a cross section only tens of micrometers wide.

[0111] The above embodiments achieve high-precision microscopic measurement: the speckle pattern formed by ultrasonic dispersion of 100nm-level carbon powder exhibits extremely high contrast and dense gray-scale gradient characteristics under an industrial microscope, which significantly improves the calculation accuracy and spatial resolution of the DIC algorithm in the microscopic field of view.

[0112] The above embodiments improve experimental efficiency: By introducing an intelligent review mechanism, the quality of speckle at the micron level is automatically determined before the experiment begins, avoiding invalid experiments due to misjudgment of speckle quality by the naked eye.

[0113] In the above embodiments, the ultrathin strip sample undergoes surface cleaning, and a base color layer is constructed on the side of the ultrathin strip. Nanoparticles are dispersed in a volatile solvent to form a speckle suspension, and the speckle suspension is sprayed onto the surface of the base color layer using aerosol deposition to form a discrete speckle layer on the side of the ultrathin strip. A load is applied to the ultrathin strip, and a sequence of microscopic images of the deformation process on the side of the ultrathin strip is acquired. The sequence of microscopic images is processed using a digital image correlation algorithm to calculate the longitudinal strain distribution along the thickness direction of the ultrathin strip cross-section. The longitudinal strain distribution is fitted, and the position of the physical neutral layer is determined based on the thickness coordinates corresponding to zero longitudinal strain. The neutral layer offset is determined based on the difference between the position of the physical neutral layer and the geometric center position of the ultrathin strip. Optionally, before acquiring the sequence of microscopic images, the quality of local microscopic images of the speckle layer is evaluated. The technical solution of the above embodiments can reduce the influence of the observation layer on the original mechanical behavior of the ultrathin strip, which is beneficial to improving the stability of the neutral layer offset measurement results.

[0114] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for offsetting the neutral layer of an ultrathin strip, characterized in that, include: The ultrathin strip sample was surface-cleaned, and a base color layer was constructed on the side of the ultrathin strip. Nanoscale particles are dispersed in a volatile solvent and then shaken to form a speckled suspension. Aerosol deposition is used to atomize and spray a suspension of speckled nanoparticles onto the surface of a base layer. After the solvent evaporates, a discrete speckled layer is formed. A load is applied to the ultrathin strip that has passed quality review, and a sequence of microscopic images of the deformation process on the side of the ultrathin strip is acquired using a microscopic optical system. Before acquiring the microscopic image sequence, the quality of local microscopic images of the discrete speckle layer is evaluated. If the quality evaluation result does not meet preset conditions, the speckle layer preparation step is re-executed. The quality evaluation includes: inputting the microscopic images into a pre-trained convolutional neural network model; the convolutional neural network model outputs a quality score or pass / fail label characterizing speckle particle density, distribution uniformity, grayscale contrast, and background color coverage. The convolutional neural network model uses a one-dimensional convolutional autoencoder to verify the physical rationality of the calculated neutral layer offset curve. If the reconstruction error exceeds a threshold, data points with reconstruction errors exceeding the threshold are marked as anomalies, and a second fitting or removal of anomaly data points is triggered. The microscopic image sequence is processed using a digital image correlation algorithm to calculate the microscopic strain field along the thickness direction of the cross section, obtain the strain distribution data sequence, and determine the position where the longitudinal strain is zero through robust fitting, and calculate the neutral layer offset.

2. The method according to claim 1, characterized in that, The ultrathin strip sample underwent surface cleaning treatment, and a base color layer was constructed on the side of the ultrathin strip, including: The micro-contact transfer process involves vertically dipping an extremely thin strip of a pre-coated, uniform primer film onto a carrier.

3. The method according to claim 1, characterized in that: The thickness of the base color layer is no greater than 3μm; The particle size of the nanoparticles is 50 nm to 500 nm; The volatile solvent is at least one of anhydrous ethanol and isopropanol.

4. The method according to claim 1, characterized in that, The aerosol deposition is performed using a pulse spraying method with a spraying distance of 15cm to 25cm and a spraying air pressure of 0.1MPa to 0.3MPa.

5. The method according to claim 1, characterized in that, The microscopic image sequence is processed using a digital image correlation algorithm to calculate the microscopic strain field along the thickness direction of the cross-section, resulting in a strain distribution data sequence including: A coordinate system along the thickness direction of the ultrathin strip is established based on the reference image and the deformed image; Extract longitudinal strain sampling points at multiple thickness locations within the ultrathin strip solid region; Obtain the data sequence of longitudinal strain as a function of thickness coordinates.

6. The method according to claim 1, characterized in that, Robust fitting includes: The longitudinal strain sampling points were linearly fitted using a random sampling consensus algorithm to remove outliers and obtain the fitting relationship between longitudinal strain and thickness coordinates.

7. An ultrathin neutral layer offset system, characterized in that, include: The sample preparation module is used to perform surface cleaning treatment on ultra-thin strip samples, construct a base color layer on the side of the ultra-thin strip, and form a discrete speckle layer on the surface of the base color layer; The loading and image acquisition module is used to apply load to the ultrathin strip after sample preparation and to acquire a sequence of microscopic images of the deformation process on the side of the ultrathin strip. The strain field solving module is used to process the microscopic image sequence using a digital image correlation algorithm to obtain the longitudinal strain distribution along the thickness direction of the ultrathin strip cross section. The neutral layer positioning module is used to fit the longitudinal strain distribution, determine the physical neutral layer position based on the position corresponding to zero longitudinal strain, and determine the neutral layer offset based on the difference between the physical neutral layer position and the geometric center position. The quality evaluation module is used to evaluate the quality of local microscopic images of the discrete speckle layer before acquiring the microscopic image sequence, and output a re-sample instruction when the quality evaluation result does not meet the preset conditions. The quality evaluation includes: inputting the microscopic image into a pre-trained convolutional neural network model; the convolutional neural network model outputting a quality score or pass / fail label characterizing speckle particle density, distribution uniformity, grayscale contrast, and background color coverage; wherein the convolutional neural network model uses a one-dimensional convolutional autoencoder to verify the physical rationality of the calculated neutral layer offset curve; if the reconstruction error exceeds a threshold, data points with reconstruction errors exceeding the threshold are marked as anomalies, and a second fitting or removal of anomaly data points is triggered.

Images