Method for predicting full-field stress of concrete lining in cold and arid region

Abstract

The present application provides a kind of cold arid region concrete lining full-field stress prediction method, it is related to stress prediction technical field, the present application is first based on frost heaving cycle test calibration under different damage degree elastic modulus, and through three-point bending fracture test simultaneously obtains the horizontal strain field, vertical strain field and acoustic emission parameter of test piece surface, further constructs the single-point coupling relationship of acoustic emission parameter and tensile stress, and it is extended to two-dimensional plane to calculate the tensile stress of each pixel point;With composite aggregate replacement rate and acoustic emission parameter as input characteristics, with pixel point tensile stress as output target, multiple machine learning models are constructed and optimized to form a high-precision stress prediction model;Based on stress prediction model, the surface stress of concrete lining is predicted.The present application solves the technical problems that traditional single-point measurement cannot visualize full-field stress and acoustic emission and strain information lack correlation, improves the reliability and authenticity of stress prediction result.

Description

Technical Field

[0001] This invention relates to the field of stress prediction technology, specifically a method for predicting the stress of concrete lining in cold and arid regions. Background Technology

[0002] Lining structures in arid and cold regions are subjected to the coupled effects of salt erosion, freeze-thaw cycles, and repeated frost heave loads on the foundation soil, making them prone to internal microcrack initiation and propagation, fracture performance degradation, and structural frost heave failure. Pumice-coal gangue composite aggregate concrete is a lining material for arid and cold regions that combines thermal insulation and durability; however, its mechanical behavior under the coupled effects of multiple factors is complex, and the stress field evolution is closely related to the damage process. Existing technologies have the following drawbacks: Stress monitoring of concrete linings mainly relies on single-point measurement methods such as strain gauges and vibrating wire stress gauges, which can only obtain stress data at discrete points and cannot achieve continuous visualization of the stress across the entire lining surface, making it difficult to capture stress concentration under frost heave. Acoustic emission technology can capture the evolution of micro-damage inside concrete, but current applications are mostly limited to damage location and single-point stress prediction, without establishing a correlation with the full-field strain information of digital imaging technology. This makes it impossible to achieve coupled analysis of "internal damage-surface strain-full-field stress", and it is difficult to support accurate evaluation of the overall mechanical state of the lining structure.

[0003] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to provide a method for predicting the stress of concrete lining in cold and arid regions, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for predicting the overall stress of concrete lining in cold and arid regions, comprising the following steps: Step 1: Prepare concrete specimens with different pumice-coal gangue composite aggregate replacement rates, and conduct multiple freeze-thaw cycles on each concrete specimen to obtain the dynamic elastic modulus of the concrete specimen after each freeze-thaw cycle. Based on this, determine the elastic modulus of the concrete specimen under different damage levels. Step 2: Conduct fracture tests on concrete specimens that have undergone different numbers of frost heave cycles. Construct the horizontal and vertical strain fields of each pixel on the surface of each concrete specimen through non-contact visual measurement. At the same time, collect the acoustic emission characteristic parameters of the concrete specimens throughout the fracture test. Step 3: Establish a single-point coupling relationship between acoustic emission parameters and tensile stress of the specimen at the test measurement point, and extend this single-point relationship to a two-dimensional plane by combining the strain field data of the specimen, and calculate the true tensile stress of all pixels on the surface of the specimen; Step 4: Using the composite aggregate replacement rate and acoustic emission characteristic parameters as input features and the tensile stress of each pixel as output, construct multiple machine learning prediction models. Use grid search combined with K-fold cross-validation algorithm to optimize the model hyperparameters and select the model with the highest prediction accuracy as the concrete lining stress prediction model. Step 5: Collect the horizontal and vertical strain fields of the concrete lining structure in the cold and arid region to be tested, reconstruct the full-field strain field of the structure, complete the pixel-level spatial registration of the full-field strain field data and the acoustic emission monitoring data, input the composite aggregate replacement rate of the lining to be tested and the acoustic emission parameters matched by each pixel into the trained stress prediction model, and output the tensile stress of each pixel in the full domain of the concrete lining.

[0006] Furthermore, the dynamic elastic modulus of the concrete specimens after each freeze-swell cycle is obtained, and the elastic modulus of the concrete specimens under different degrees of damage is calibrated accordingly. The specific principle is as follows: The static elastic modulus of each concrete specimen without frost heave cycle test is used as the reference value of elastic modulus. After each frost heave cycle, the dynamic elastic modulus of the concrete specimen under the current damage level is measured by resonance method. The quotient of the dynamic elastic modulus and the elastic modulus reference value is calculated by subtracting 1 from the dynamic elastic modulus and the elastic modulus reference value to obtain the damage quantification value. The mapping relationship between dynamic elastic modulus, damage quantification value and equivalent static elastic modulus is established in advance. The equivalent static elastic modulus under the corresponding damage level is obtained by interpolation based on the measured dynamic elastic modulus. The equivalent static elastic modulus is used as the elastic modulus of the concrete specimen.

[0007] Furthermore, the principle of constructing the horizontal and vertical strain fields of each pixel on the surface of each concrete specimen through non-contact visual measurement is as follows: The fracture test was a three-point bending fracture test. Before the fracture test, surface images of each concrete specimen were acquired as reference images. Random speckle patterns were sprayed onto the surface of the concrete specimens as feature markers. The three-point bending fracture test was then performed on the concrete specimens. During the test, sequential digital images of the concrete specimen surface were acquired according to a preset sampling frequency. The reference image was divided into several rectangular subsets. A zero-mean normalized cross-correlation function was used to match the subsets of the sequential digital images with the reference image. The displacement vectors of the center points of each matrix subset in the reference image were calculated. Spatial interpolation was performed on the displacement vectors of all subset center points to expand the horizontal and vertical displacement fields of the entire pixel area of ​​the specimen. Spatial differentiation was performed on the horizontal and vertical displacement fields to obtain the horizontal and vertical normal strains of each pixel. The horizontal and vertical normal strains of all pixels were integrated to construct the horizontal and vertical strain fields of each pixel on the surface of the concrete specimen. Acoustic emission characteristic parameters are collected at preset measuring points on concrete specimens. These acoustic emission characteristic parameters include cumulative ring count, energy rate, amplitude, average frequency, b-value, and entropy value.

[0008] Furthermore, the principle for establishing the single-point coupling relationship between acoustic emission parameters at the test measurement point and the tensile stress of the specimen is as follows: In the three-point bending fracture test, several acquisition times were set at equal intervals. The cumulative ring count and applied load value at each acoustic emission measuring point were collected at each acquisition time, and the final value of the cumulative ring count when the concrete specimen completely fractured was recorded. The cross-sectional width, cross-sectional height, and support span of the rectangular concrete specimen are measured. At any sampling time, the load value is multiplied by the support span and then divided by 4 to obtain the mid-span bending moment of the concrete specimen. The cube of the cross-sectional height is multiplied by the cross-sectional width and then divided by 12 to obtain the rectangular cross-sectional moment of inertia of the concrete specimen. The mid-span bending moment is multiplied by half of the cross-sectional height and then divided by the rectangular cross-sectional moment of inertia to obtain the nominal tensile stress of the concrete specimen at each sampling time. Damage evolution is characterized by cumulative ring count: the cumulative ring count at each acquisition time is divided by the final value of the cumulative ring count to obtain the real-time damage variable at each acquisition time; the nominal tensile stress at each acquisition time is divided by 1 and the difference between the real-time damage variable to obtain the corrected tensile stress at each acquisition time. Using the cumulative ring count as the independent variable and the modified tensile stress as the dependent variable, multiple sets of discrete data sample points with one-to-one correspondence are formed. By using the Boltzmann function to perform nonlinear fitting on discrete sample points, a single-point coupling relationship between the cumulative ringing count and the corrected tensile stress at a single acoustic emission test point is obtained.

[0009] Furthermore, by combining the strain field data of the specimen, this single-point relationship is extended to a two-dimensional plane to calculate the true tensile stress of all pixels on the specimen surface. The specific principle is as follows: The equivalent elastic modulus corresponding to the current degree of frost heave damage is retrieved. Combined with the horizontal and vertical normal strains of each pixel, the elastic stress of the pixel is solved according to the plane Hooke's law. The plane shear strain is solved from the horizontal and vertical normal strains. The Poisson's ratio of the concrete is measured in advance. The horizontal and vertical normal stress components are solved by combining the equivalent elastic modulus, Poisson's ratio, and the two types of normal strains. The shear stress components are solved by combining the elastic modulus, Poisson's ratio, and shear strain. Substituted into the plane principal stress calculation formula, the elastic principal stress of each pixel without considering local damage is obtained. For each acoustic emission acquisition point, the damage correction factor specific to that point is obtained by dividing the elastic principal stress of that point by the obtained corrected tensile stress. Radial basis function interpolation is used, with the damage correction factor of each acquisition point as the interpolation constraint, to solve for the damage correction factor corresponding to all pixels on the surface of the specimen. The damage correction factor obtained by interpolation of each pixel is multiplied by the elastic principal stress of that pixel to obtain the true tensile stress of each pixel considering local damage.

[0010] Furthermore, a grid search combined with K-fold cross-validation algorithm is used to optimize the model's hyperparameters and select the model with the highest prediction accuracy. The specific principle is as follows: Using composite aggregate replacement rate, cumulative ring count, energy rate, amplitude, average frequency, b-value, and entropy as model input features, and the actual tensile stress of each pixel as the output label, a pixel-level supervised regression dataset is constructed. Six machine learning models were constructed, including logistic regression, support vector machine, artificial neural network, random forest, extreme gradient boosting tree and adaptive boosting tree, and multiple sets of hyperparameter combinations to be traversed were preset for each model. A grid search is used to traverse all hyperparameter combinations, and K-fold cross-validation is used to evaluate the performance of each set of parameters: the dataset is divided into K non-overlapping subsets, K-1 subsets are selected in turn to train the model, and the remaining 1 subset is used as the validation set. After K complete cycles, the average of the K rounds of validation metrics is taken as the performance score of the hyperparameter set; after traversing all combinations, the optimal hyperparameter combination for each model is determined. Each model was trained using optimal hyperparameters. The coefficient of determination and root mean square error between the model's predicted tensile stress and the actual tensile stress were calculated. The two evaluation indicators were normalized and weighted by setting weighting coefficients. A unified accuracy judgment index was obtained by calculating the weighted difference. Compare the accuracy judgment index of all models and select the model with the highest accuracy judgment index value.

[0011] Furthermore, the full-field strain field of the structure is reconstructed, and the full-field strain field data is spatially registered with the acoustic emission monitoring data at the pixel level. The specific principle is as follows: A non-contact visual measurement method, identical to that used for constructing the strain field on the surface of the specimen, was employed to acquire images of the concrete lining surface to be measured. The horizontal and vertical normal strains of each pixel in the lining were then calculated. The horizontal and vertical normal strains of a single pixel were integrated into a two-dimensional strain vector, and the set of two-dimensional strain vectors corresponding to all pixels constituted the full-field strain field of the lining. Several acoustic emission monitoring points were set up on the surface of the lining to be tested, and acoustic emission characteristic parameters of each point were collected. The inverse distance weighted interpolation algorithm was used, with the parameters of each monitoring point as the interpolation constraint, to solve the acoustic emission characteristic parameters corresponding to each pixel in the entire lining area. A unified image pixel coordinate system for the full-field strain field and interpolated pixel-level acoustic emission characteristic parameters is established to achieve one-to-one correspondence between pixels in the same space for two types of multi-source monitoring data, thus completing pixel-level spatial registration.

[0012] Compared with the prior art, the beneficial effects of the present invention are: This invention prepares concrete specimens with different composite aggregate replacement rates and conducts frost heave cycle tests, measuring the dynamic elastic modulus round by round. The elastic modulus under different damage levels is calibrated, and the degree of frost heave damage is quantitatively correlated with the degradation of the elastic modulus. This provides a material constitutive basis that dynamically adjusts with the damage state for subsequent stress calculations, avoiding prediction bias caused by a fixed elastic modulus. Fracture tests are then performed on the cyclic specimens, and the horizontal strain field, vertical strain field, and acoustic emission parameters during the fracture process are simultaneously acquired at each pixel point on the surface. This achieves a monitoring dimension improvement from point measurement to surface coverage, enhancing the information richness and initial accuracy of stress inversion.

[0013] This invention also solves the problem that acoustic emission information cannot guide the stress distribution across the entire field in traditional methods by constructing a single-point coupling relationship between acoustic emission parameters and tensile stress, and by using horizontal strain field data and damage elastic modulus to extend this relationship to a two-dimensional plane and calculate the tensile stress at each pixel. This makes the tensile stress calculation results closer to the real damage state. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This invention corrects the fitting curve of tensile stress as a function of cumulative ring count. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0016] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0017] Example: Please see Figures 1 to 2 The present invention provides a technical solution: A method for predicting the overall stress of concrete lining in cold and arid regions, comprising the following steps: Step 1: Prepare concrete specimens with different pumice-coal gangue composite aggregate replacement rates, and conduct multiple freeze-thaw cycles on each concrete specimen to obtain the dynamic elastic modulus of the concrete specimen after each freeze-thaw cycle. Based on this, determine the elastic modulus of the concrete specimen under different damage levels. In this embodiment, the pumice-coal gangue composite aggregate replacement rate represents the proportion of pumice-coal gangue composite aggregate to the total mass of all coarse aggregate. The pumice and coal gangue are crushed, sieved, washed and dried respectively. The water-cement ratio, sand ratio and cement dosage are fixed, and the composite aggregate replacement rate is set to 0%, 20%, 40%, 60% and 80%. The coarse aggregate is replaced with a mixture of pumice and coal gangue according to the composite aggregate replacement rate, and concrete specimens are prepared.

[0018] In this embodiment, the dynamic elastic modulus of the concrete specimen after each freeze-swell cycle is obtained, and the elastic modulus of the concrete specimen under different degrees of damage is calibrated accordingly. The specific principle is as follows: The static elastic modulus of each concrete specimen without frost heave cycle test is used as the reference value of elastic modulus. After each frost heave cycle, the dynamic elastic modulus of the concrete specimen under the current damage level is measured by resonance method. The quotient of the dynamic elastic modulus and the elastic modulus reference value is calculated by subtracting 1 from the dynamic elastic modulus and the elastic modulus reference value to obtain the damage quantification value. The mapping relationship between dynamic elastic modulus, damage quantification value and equivalent static elastic modulus is established in advance. The equivalent static elastic modulus under the corresponding damage level is obtained by interpolation based on the measured dynamic elastic modulus. The equivalent static elastic modulus is used as the elastic modulus of the concrete specimen.

[0019] Before undergoing frost heave cyclic testing, uniaxial compression tests were performed on concrete specimens. Resistance strain gauges were installed on the concrete specimens to measure the longitudinal strain of the concrete specimens in the elastic stage. The applied load values ​​were recorded simultaneously. The increment of the load value and the corresponding increment of the longitudinal strain were extracted. The stress increment was obtained by dividing the increment of the load value by the transverse cross-sectional area of ​​the concrete specimen. The elastic modulus reference value of the concrete specimen was obtained by dividing the stress increment by the corresponding increment of the longitudinal strain.

[0020] The damage quantification value reflects the cumulative deterioration of concrete materials under the current frost heave cycle, specifically manifested as the relative degradation rate of elastic stiffness, and its value range is as follows: The closer the damage quantification value is to 0, the more intact the concrete specimen is; the closer it is to 1, the more severe the stiffness degradation and the higher the cumulative damage. Under the action of freeze-thaw cycles, the concrete develops more internal microcracks and increased porosity, leading to a decrease in its resistance to deformation, which is macroscopically manifested as a decrease in the elastic modulus. Therefore, the relative decrease in the elastic modulus is used as a quantitative indicator of the degree of damage.

[0021] For each concrete specimen, its dynamic elastic modulus and damage quantification value after each freeze-thaw cycle are obtained, and combined with the equivalent static elastic modulus, several discrete data points are obtained: ,in, Indicates the first Dynamic elastic modulus after one frost heave cycle An index representing the number of frost heave cycles. , The total number of freeze-thaw cycles. Indicates the first Quantitative values ​​of damage after one frost heave cycle Indicates the first The equivalent static elastic modulus after one freeze-heave cycle; for any measured dynamic elastic modulus If the following conditions are met: Then, the result is calculated using linear interpolation. Corresponding equivalent static elastic modulus for: ;in, Indicates the first Dynamic elastic modulus after one frost heave cycle Indicates the first Equivalent static elastic modulus after one freeze-swell cycle.

[0022] Step 2: Conduct fracture tests on concrete specimens that have undergone different numbers of frost heave cycles. Construct the horizontal and vertical strain fields of each pixel on the surface of each concrete specimen through non-contact visual measurement. At the same time, collect the acoustic emission characteristic parameters of the concrete specimens throughout the fracture test. In this embodiment, the principle of constructing the horizontal and vertical strain fields of each pixel on the surface of each concrete specimen through non-contact visual measurement is as follows: The fracture test was a three-point bending fracture test. Before the fracture test, surface images of each concrete specimen were acquired as reference images. Random speckle patterns were sprayed onto the surface of the concrete specimens as feature markers. The three-point bending fracture test was then performed on the concrete specimens. During the test, sequential digital images of the concrete specimen surface were acquired according to a preset sampling frequency. The reference image was divided into several rectangular subsets. A zero-mean normalized cross-correlation function was used to match the subsets of the sequential digital images with the reference image. The displacement vectors of the center points of each matrix subset in the reference image were calculated. Spatial interpolation was performed on the displacement vectors of all subset center points to expand the horizontal and vertical displacement fields of the entire pixel area of ​​the specimen. Spatial differentiation was performed on the horizontal and vertical displacement fields to obtain the horizontal and vertical normal strains of each pixel. The horizontal and vertical normal strains of all pixels were integrated to construct the horizontal and vertical strain fields of each pixel on the surface of the concrete specimen. The reference image is divided into several rectangular subsets, and the sequence of digital images is divided into several candidate regions. For each candidate region in the sequence of digital images, the zero-mean normalized cross-correlation function (ZNCC) is used to search for the rectangular subset in the reference image that best matches each candidate region. The principle is as follows: using the zero-mean normalized cross-correlation function as a similarity metric, the same number of pixels are selected in both the matrix subsets and the candidate regions, and the gray values ​​of these pixels are obtained. The mean gray values ​​of the matrix subsets and the candidate regions are calculated. Based on the two gray-level matrices, the ZNCC value is calculated using the following formula: in, This represents the ZNCC value between the candidate region and the rectangular subset. Represents the index of the pixel, and , Indicates the total number of pixels. Represents the first element in the rectangular subset. The grayscale value of each pixel This represents the average grayscale value of the pixels within the rectangular subset. Indicates the first candidate region The grayscale value of each pixel This represents the average grayscale value of the pixels in the candidate region.

[0023] The ZNCC value is used to determine whether a candidate region in a sequence of digital images and a rectangular subset in a reference image are at the same location, and its value range is [value range missing]. The closer the value is to 1, the more consistent the grayscale distribution pattern of the candidate region and the rectangular subset is, and the higher the degree of matching between them. Using the maximization of the ZNCC value as the criterion, all candidate regions are traversed within a preset search range. The candidate region with the largest ZNCC value is determined as the region that best matches the current rectangular subset, and the position offset of its center point is used as the displacement vector of the subset's center point, completing a one-to-one correspondence matching from the reference image to the sequence of digital images. After obtaining the displacement vector of the subset's center point, interpolation is performed to obtain the displacement vector at any position of the concrete specimen, constructing the horizontal and vertical displacement fields.

[0024] Sensors are arranged at preset measuring points on the concrete specimen to collect acoustic emission characteristic parameters, including cumulative ring count, energy rate, amplitude, average frequency, b-value, and entropy value; where b-value is the slope of the fitting curve of the Gutenberg-Richard damage law, which is used to characterize the proportion distribution of microcracks and macrocracks inside the concrete specimen.

[0025] The methods for collecting cumulative ring count, energy rate, amplitude, average frequency, and entropy value are as follows: The cumulative ring count represents the number of times the signal exceeds the threshold voltage during acoustic emission; each time the acoustic emission signal exceeds the threshold voltage, it is counted as one ring, and the cumulative ring count is generated by counting all the rings. Energy rate represents the relative energy released by the acoustic emission signal waveform per unit time, reflecting the energy release rate of the acoustic emission event in the time dimension; the relative energy value of the signal is obtained by integrating the square of the acoustic emission signal waveform, and the energy rate is obtained by dividing the relative energy value by the impact time of the acoustic emission signal. The amplitude represents the difference between the maximum peak voltage and the minimum trough voltage in the acoustic emission signal waveform. The amplitude is obtained by subtracting the minimum trough voltage from the maximum peak voltage in the acoustic emission signal waveform. The average frequency represents the ratio of the cumulative ring count to the impact duration in an acoustic emission signal impact event; the average frequency is obtained by dividing the cumulative ring count by the impact time of the acoustic emission signal. The b-value is used to quantitatively characterize the relative proportion of microcracks to macrocracks within concrete materials, thereby determining the damage stage and instability risk of the structure. Several amplitude thresholds are set at fixed intervals for the amplitude of acoustic emission events. The number of acoustic emission events exceeding each threshold is counted, and each amplitude threshold is mapped one-to-one with the number of acoustic emission events exceeding that threshold, forming several experimental data points. A linear regression equation is then constructed based on the Gutenberg-Richard damage law. in, Indicates the amplitude threshold. This represents the number of acoustic emission events with an amplitude greater than the amplitude threshold A. Represents a constant term. Indicates the value of b; Substitute each experimental data point into the linear regression equation and use the least squares method to fit the solution. The optimal fitting criterion is to minimize the sum of squared residuals of all experimental data points. The optimal coefficients a and b of the linear regression equation are obtained by iterative solution. At this time, b is the b value of the concrete specimen under real-time damage state.

[0026] The b-value reflects the ratio of microcracks to macrocracks within a concrete specimen, used to distinguish damage modes. A larger b-value indicates more microcracks than macrocracks in the concrete specimen, suggesting the structure is in a relatively safe state. A smaller b-value indicates more macrocracks, suggesting the structure is close to instability. The fixed interval for setting the amplitude threshold uses the 1 dB interval commonly used in acoustic emission analysis, which allows for the capture of subtle changes in amplitude distribution.

[0027] Entropy value represents the degree of disorder in the spectrum of acoustic emission signal. The larger the entropy value, the more dispersed the signal energy is in the frequency domain, the more complex the components are, and the higher the randomness. The smaller the entropy value, the more concentrated the signal energy is in a few frequency bands, the simpler the spectrum structure, the stronger the regularity, and the more dominant the frequency characteristics. The proportion of energy in each frequency band of the acoustic emission signal to the total energy is obtained, and the entropy value is calculated based on Shannon's formula.

[0028] Step 3: Establish a single-point coupling relationship between acoustic emission parameters and tensile stress of the specimen at the test measurement point, and extend this single-point relationship to a two-dimensional plane by combining the strain field data of the specimen, and calculate the true tensile stress of all pixels on the surface of the specimen; In this embodiment, the principle for establishing the single-point coupling relationship between the acoustic emission parameters at the test measurement point and the tensile stress of the specimen is as follows: In the three-point bending fracture test, several acquisition times were set at equal intervals. The cumulative ring count and applied load value at each acoustic emission measuring point were collected at each acquisition time, and the final value of the cumulative ring count when the concrete specimen completely fractured was recorded. The cross-sectional width, height, and support span of a rectangular concrete specimen are measured. At any sampling time, the load value is multiplied by the support span and then divided by 4 to obtain the mid-span bending moment of the concrete specimen. The cube of the cross-sectional height is multiplied by the cross-sectional width and then divided by 12 to obtain the rectangular cross-sectional moment of inertia of the concrete specimen. The mid-span bending moment is multiplied by half the cross-sectional height and then divided by the rectangular cross-sectional moment of inertia to obtain the nominal tensile stress of the concrete specimen at each sampling time. The specific formula is as follows: in, Indicates the bending moment at the mid-span section. Indicates the load value. Indicates the span of the support. Represents the moment of inertia of a rectangular cross section. Indicates the width of the cross section. Indicates the cross-sectional height. Indicates nominal tensile stress; Nominal tensile stress represents the maximum tensile stress at the mid-span section, calculated as a perfectly elastic body without considering internal material damage. The formula for calculating nominal tensile stress is based on the principle of bending normal stress calculation in elastic beam bending theory. Under this principle, the maximum bending normal stress at the mid-span section of the beam occurs at the upper and lower edges of the section furthest from the neutral axis. The formula logic is: the stress at any point on the section is equal to the bending moment of the section multiplied by the distance from that point to the neutral axis of the section, then divided by the moment of inertia of the section about the neutral axis. Therefore, the calculation process for nominal tensile stress is derived as follows: simplify the three-point bending specimen into a simply supported beam, subjected to a concentrated force at mid-span, with the support reactions at both ends of the simply supported beam being... And the distance from each of the two end supports to the middle is The bending moment at the mid-span section is equal to For a rectangular cross-section with width s and height h, its moment of inertia about the neutral axis is: ,in, This represents the perpendicular distance from any point on the cross-section to the neutral axis; the bending normal stress is linearly distributed along the height of the cross-section, being zero at the neutral axis and maximum at the upper and lower edges, with the distance from the upper and lower edges to the neutral axis being... Substitute the obtained bending moment at the mid-span section, the moment of inertia of the rectangular section, and the distance from the location of the maximum bending normal stress to the neutral axis into the bending normal stress calculation formula to calculate the bending normal stress, which is the nominal tensile stress.

[0029] Damage evolution is characterized by cumulative ring count: the cumulative ring count at each acquisition time is divided by the final value of the cumulative ring count to obtain the real-time damage variable at each acquisition time; the nominal tensile stress at each acquisition time is divided by 1 and the difference between the real-time damage variable to obtain the corrected tensile stress at each acquisition time. The corrected tensile stress reflects the effective value of the actual tensile stress borne by concrete under external loads. It eliminates the artificially inflated elastic stress caused by accumulated damage such as microcracks and pores, and is closer to the true mechanical state of the material's microstructure. Under frost heave and load, concrete will generate microcracks that continue to expand, resulting in a reduction in the effective cross-sectional area that actually bears the tensile force. The nominal tensile stress equals the macroscopic external load divided by the complete cross-sectional area, and the corrected tensile stress equals the macroscopic external load divided by the effective area. Due to the damage, the effective area that actually participates in the load-bearing is reduced relative to the complete cross-sectional area, and the reduction ratio is the real-time damage variable. Therefore, subtracting the real-time damage variable from 1 and multiplying it by the complete cross-sectional area gives the effective area. Substituting the effective area into the formula for calculating the corrected tensile stress, we get the corrected tensile stress as the difference between the nominal tensile stress divided by 1 and the real-time damage variable.

[0030] To illustrate the relationship between the modified tensile stress and the cumulative ringing count, Table 1 is constructed as follows: Table 1. Variation of modified tensile stress with cumulative ring count As shown in Table 1, when the cumulative ring count is below 980, it corresponds to the microcrack initiation stage. At this time, the external load is low, acoustic emission activity is sparse, and the cumulative ring count increases slowly. Elastic deformation is the main process at this stage. When the cumulative ring count is between 980 and 6200, it corresponds to the crack propagation stage. At this time, the load continues to increase, microcracks begin to merge and propagate along the weak path, acoustic emission events increase sharply, and the cumulative ring count rises rapidly. When the cumulative ring count is between 6200 and 22200, it corresponds to the crack acceleration penetration zone. At this time, acoustic emission events increase explosively, the material is highly damaged, the remaining effective load-bearing section is greatly reduced, and it is close to the load-bearing limit. When the cumulative ring count is above 22200, it corresponds to the pre-failure stage. At this time, the cumulative ring count continues to increase significantly, but the corrected tensile stress is close to the limit value of 4.0 MPa and hardly increases anymore. The material has lost its structural integrity, and the corrected tensile stress gradually reaches its upper limit.

[0031] Figure 2 The curve is a fitted curve of the modified tensile stress as a function of cumulative ringing count, plotted based on Table 1. Figure 2 It was found that the curve variation pattern completely corresponds to the data in Table 1: the corrected tensile stress increases with the increase of the cumulative ring count, and the rate of increase of the corrected tensile stress gradually decreases.

[0032] Using the cumulative ring count as the independent variable and the modified tensile stress as the dependent variable, multiple sets of discrete data sample points with one-to-one correspondence are formed. By using the Boltzmann function to perform nonlinear fitting on discrete sample points, a single-point coupling relationship between the cumulative ring count and tensile stress at a single acoustic emission test point is obtained.

[0033] By combining the strain field data of the specimen, this single-point relationship is extended to a two-dimensional plane to calculate the true tensile stress of all pixels on the specimen surface. The specific principle is as follows: The equivalent elastic modulus corresponding to the current degree of frost heave damage is retrieved. Combined with the horizontal and vertical normal strains of each pixel, the elastic stress of the pixel is solved according to the plane Hooke's law. The plane shear strain is solved from the horizontal and vertical normal strains. The Poisson's ratio of the concrete is measured in advance. The horizontal and vertical normal stress components are solved by combining the equivalent elastic modulus, Poisson's ratio, and the two types of normal strains. The shear stress components are solved by combining the elastic modulus, Poisson's ratio, and shear strain. Substituted into the plane principal stress calculation formula, the elastic principal stress of each pixel without considering local damage is obtained. For each acoustic emission acquisition point, the damage correction factor specific to that point is obtained by dividing the elastic principal stress of that point by the obtained corrected tensile stress. Radial basis function interpolation is used, with the damage correction factor of each acquisition point as the interpolation constraint, to solve for the damage correction factor corresponding to all pixels on the surface of the specimen. The damage correction factor obtained by interpolation of each pixel is multiplied by the elastic principal stress of that pixel to obtain the true tensile stress of each pixel considering local damage.

[0034] The formula for calculating the horizontal elastic stress components is: ; The formula for calculating the vertical elastic stress components is: ; The formula for calculating shear strain is: ; The formula for calculating shear stress is: ; in, Represents the elastic stress component in the horizontal direction. This represents the equivalent static elastic modulus corresponding to the current degree of frost heave damage. This represents the Poisson's ratio of the concrete specimen. This represents the horizontal normal strain of a pixel. This represents the vertical normal strain of a pixel. Indicates shear strain. These represent the horizontal strain field and the vertical strain field, respectively. This represents the rate of change of horizontal displacement along the y-direction. This represents the rate of change of vertical displacement along the x-direction. This represents shear stress.

[0035] Based on the principles of continuum damage mechanics, the nominal tensile stress is corrected to the effective value of the actual tensile stress according to the proportion of the effective bearing area, thereby establishing a single-point coupling relationship. This corrected tensile stress eliminates the artificially increased elastic stress caused by cumulative damage such as microcracks and pores, and is closer to the true mechanical state of the material's microstructure. Furthermore, using the full-field strain data obtained by digital image correlation methods and the equivalent elastic modulus after frost heave damage, the elastic principal stress of each pixel is solved according to the plane Hooke's law. Then, the ratio of the elastic principal stress at the discrete acoustic emission measurement point to the corrected tensile stress is used as the damage correction factor. This factor is extended to all pixels through radial basis function interpolation. Finally, the elastic principal stress of each pixel is multiplied by the corresponding damage correction factor. Thus, while retaining the analytical framework of elasticity mechanics, the weakening effect of local non-uniform damage on bearing capacity is introduced into the full-field stress calculation. This approach avoids the global bias caused by directly using a fixed elastic modulus or nominal tensile stress, and makes full use of the complementary advantages of multi-source monitoring data in space. This allows the final obtained true tensile stress to reflect both the stiffness degradation caused by frost heave damage and the local stress redistribution caused by crack propagation.

[0036] Step 4: Using the composite aggregate replacement rate and acoustic emission characteristic parameters as input features and the tensile stress of each pixel as output, construct multiple machine learning prediction models. Use grid search combined with K-fold cross-validation algorithm to optimize the model hyperparameters and select the model with the highest prediction accuracy as the concrete lining stress prediction model. In this embodiment, a grid search combined with K-fold cross-validation algorithm is used to optimize the model hyperparameters and select the model with the highest prediction accuracy. The specific principle is as follows: Using composite aggregate replacement rate, cumulative ring count, energy rate, amplitude, average frequency, b-value, and entropy as model input features, and the actual tensile stress of each pixel as the output label, a pixel-level supervised regression dataset is constructed. Six machine learning models were constructed, including logistic regression, support vector machine, artificial neural network, random forest, extreme gradient boosting tree and adaptive boosting tree, and multiple sets of hyperparameter combinations to be traversed were preset for each model. A grid search is used to traverse all hyperparameter combinations, and K-fold cross-validation is used to evaluate the performance of each set of parameters: the dataset is divided into K non-overlapping subsets, K-1 subsets are selected in turn to train the model, and the remaining 1 subset is used as the validation set. After K complete cycles, the average of the K rounds of validation metrics is taken as the performance score of the hyperparameter set; after traversing all combinations, the optimal hyperparameter combination for each model is determined. For each candidate machine learning model, a list of candidate values ​​for each hyperparameter is pre-defined based on its structural characteristics and tuning sensitivity, thus forming a multi-dimensional hyperparameter space. A grid search strategy is then used to exhaustively traverse all possible hyperparameter combinations within this space. For each set of hyperparameters, K-fold cross-validation is used to robustly evaluate its generalization performance: the entire training dataset is randomly divided into K non-overlapping subsets of approximately equal size. K-1 subsets are selected sequentially as the training set for model fitting, and the remaining subset is used as the validation set to calculate the prediction error. This process is repeated K times, ensuring that each subset has one opportunity to serve as the validation set. After K rounds of validation, the arithmetic mean square error of the validation sets in each round is calculated as the performance score of each set of hyperparameter values ​​for each model. After traversing all pre-defined hyperparameter combinations, the hyperparameter combination with the smallest root mean square error is selected as the actual hyperparameter values ​​for the model.

[0037] Each model was trained using optimal hyperparameters. The coefficient of determination and root mean square error between the model's predicted tensile stress and the actual tensile stress were calculated. The two evaluation indicators were normalized and weighted by setting weighting coefficients. A unified accuracy judgment index was obtained by calculating the weighted difference. Compare the accuracy judgment index of all models and select the model with the highest accuracy judgment index value.

[0038] After determining the optimal hyperparameters for each candidate model, a comprehensive evaluation and horizontal comparison of the model prediction accuracy are conducted to select the globally optimal stress prediction model. The model is retrained based on the complete training set, and two fundamental evaluation metrics are calculated using an independent test set: the coefficient of determination (COD) and the root mean square error (RMSE). The COD reflects the model's explanatory power for the variance of the target variable; a value closer to 1 indicates higher fitting accuracy. The RMSE measures the average deviation between the predicted and actual values; a smaller value indicates lower prediction error. Since the COD and RMSE have different dimensions and numerical ranges, they are mapped to different values ​​after being subjected to maximum-minimum normalization. The intervals are used to make the two comparable; then, the accuracy judgment index is obtained by weighted difference calculation. The higher the accuracy judgment index, the better the model performs in terms of high fit and low prediction error. The model with the highest accuracy judgment index is selected from all models. This model is the optimal model that meets the requirements of prediction bias and fitting accuracy.

[0039] In engineering, it is desirable for the model to simultaneously satisfy high goodness of fit and low prediction error. Therefore, the accuracy judgment index is positively correlated with the goodness of fit and negatively correlated with the prediction error, and the two have relatively independent effects on the accuracy judgment index. Therefore, the accuracy judgment index is calculated by weighted subtraction of the coefficient of determination and the root mean square error. The coefficient of determination is used to measure whether the model can accurately reflect the change of stress with damage evolution, while the root mean square error is used to reflect the deviation between the predicted stress value and the true value at each pixel, directly reflecting the prediction accuracy of the model. Therefore, the weight of the root mean square error is greater than the weight of the coefficient of determination to ensure the prediction accuracy of the model.

[0040] As one implementation method, the weight of the root mean square error ranges from 1 to 1. The weight range of the coefficient of determination is: The specific value is to be set by those skilled in the art based on the actual situation, and is not limited here.

[0041] Step 5: Collect the horizontal and vertical strain fields of the concrete lining structure in the cold and arid region to be tested, reconstruct the full-field strain field of the structure, complete the pixel-level spatial registration of the full-field strain field data and the acoustic emission monitoring data, input the composite aggregate replacement rate of the lining to be tested and the acoustic emission parameters matched by each pixel into the trained stress prediction model, and output the tensile stress of each pixel in the full domain of the concrete lining.

[0042] In this embodiment, the full-field strain field of the structure is reconstructed, and the full-field strain field data is spatially registered with the acoustic emission monitoring data at the pixel level. The specific principle is as follows: A non-contact visual measurement method, identical to that used for constructing the strain field on the surface of the specimen, was employed to acquire images of the concrete lining surface to be measured. The horizontal and vertical normal strains of each pixel in the lining were then calculated. The horizontal and vertical normal strains of a single pixel were integrated into a two-dimensional strain vector, and the set of two-dimensional strain vectors corresponding to all pixels constituted the full-field strain field of the lining. Several acoustic emission monitoring points were set up on the surface of the lining to be tested, and acoustic emission characteristic parameters of each point were collected. The inverse distance weighted interpolation algorithm was used, with the parameters of each monitoring point as the interpolation constraint, to solve the acoustic emission characteristic parameters corresponding to each pixel in the entire lining area. A unified image pixel coordinate system for the full-field strain field and interpolated pixel-level acoustic emission characteristic parameters is established to achieve one-to-one correspondence between pixels in the same space for two types of multi-source monitoring data, thus completing pixel-level spatial registration.

[0043] Due to the limited number of monitoring points, it is difficult to cover every pixel in the entire area. Therefore, an inverse distance weighted interpolation algorithm is adopted. The measured acoustic emission parameters of each monitoring point are used as interpolation constraints to spatially interpolate all pixels on the surface of the lining under test, and the acoustic emission characteristic parameter values ​​corresponding to each pixel are obtained. On this basis, the image pixel coordinate system of the whole-field strain field and the pixel coordinate system of the interpolated acoustic emission characteristic parameters are unified to ensure that the two are based on the same spatial reference frame, so as to achieve one-to-one correspondence between the two types of multi-source monitoring data at the same pixel position, that is, to complete the pixel-level spatial registration. The composite aggregate replacement rate of the lining under test and the registered acoustic emission parameter vector of each pixel are input into the stress prediction model. The stress prediction model outputs the tensile stress prediction value pixel by pixel. By arranging the prediction results of all pixels according to the original image coordinates, a tensile stress distribution map of all pixels in the entire area of ​​the concrete lining under test can be generated, realizing continuous and visual prediction of the stress in the entire field of the lining surface.

[0044] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0045] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0046] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0047] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for predicting the overall stress of concrete lining in cold and arid regions, characterized in that, The specific steps include: Step 1: Prepare concrete specimens with different pumice-coal gangue composite aggregate replacement rates, and conduct multiple freeze-thaw cycles on each concrete specimen to obtain the dynamic elastic modulus of the concrete specimen after each freeze-thaw cycle. Based on this, determine the elastic modulus of the concrete specimen under different damage levels. Step 2: Conduct fracture tests on concrete specimens that have undergone different numbers of frost heave cycles. Construct the horizontal and vertical strain fields of each pixel on the surface of each concrete specimen through non-contact visual measurement. At the same time, collect the acoustic emission characteristic parameters of the concrete specimens throughout the fracture test. Step 3: Establish a single-point coupling relationship between acoustic emission parameters and tensile stress of the specimen at the test measurement point, and extend this single-point relationship to a two-dimensional plane by combining the strain field data of the specimen, and calculate the true tensile stress of all pixels on the surface of the specimen; Step 4: Using the composite aggregate replacement rate and acoustic emission characteristic parameters as input features and the tensile stress of each pixel as output, construct multiple machine learning prediction models. Use grid search combined with K-fold cross-validation algorithm to optimize the model hyperparameters and select the model with the highest prediction accuracy as the concrete lining stress prediction model. Step 5: Collect the horizontal and vertical strain fields of the concrete lining structure in the cold and arid region to be tested, reconstruct the full-field strain field of the structure, complete the pixel-level spatial registration of the full-field strain field data and the acoustic emission monitoring data, input the composite aggregate replacement rate of the lining to be tested and the acoustic emission parameters matched by each pixel into the trained stress prediction model, and output the tensile stress of each pixel in the full domain of the concrete lining.

2. The method for predicting the overall stress of concrete lining in arid and cold regions according to claim 1, characterized in that: The dynamic elastic modulus of concrete specimens after each freeze-suspension cycle is obtained, and the elastic modulus of concrete specimens under different degrees of damage is calibrated accordingly. The specific principle is as follows: The static elastic modulus of each concrete specimen without frost heave cycle test is used as the reference value of elastic modulus. After each frost heave cycle, the dynamic elastic modulus of the concrete specimen under the current damage level is measured by resonance method. The quotient of the dynamic elastic modulus and the elastic modulus reference value is calculated by subtracting 1 from the dynamic elastic modulus and the elastic modulus reference value to obtain the damage quantification value. The mapping relationship between dynamic elastic modulus, damage quantification value and equivalent static elastic modulus is established in advance. The equivalent static elastic modulus under the corresponding damage level is obtained by interpolation based on the measured dynamic elastic modulus. The equivalent static elastic modulus is used as the elastic modulus of the concrete specimen.

3. The method for predicting the overall stress of concrete lining in cold and arid regions according to claim 1, characterized in that: The principle of constructing the horizontal and vertical strain fields of each pixel on the surface of each concrete specimen through non-contact visual measurement is as follows: The fracture test was a three-point bending fracture test. Before the fracture test, surface images of each concrete specimen were acquired as reference images. Random speckle patterns were sprayed onto the surface of the concrete specimens as feature markers. The three-point bending fracture test was then performed on the concrete specimens. During the test, sequential digital images of the concrete specimen surface were acquired according to a preset sampling frequency. The reference image was divided into several rectangular subsets. A zero-mean normalized cross-correlation function was used to match the subsets of the sequential digital images with the reference image. The displacement vectors of the center points of each matrix subset in the reference image were calculated. Spatial interpolation was performed on the displacement vectors of all subset center points to expand the horizontal and vertical displacement fields of the entire pixel area of ​​the specimen. Spatial differentiation was performed on the horizontal and vertical displacement fields to obtain the horizontal and vertical normal strains of each pixel. The horizontal and vertical normal strains of all pixels were integrated to construct the horizontal and vertical strain fields of each pixel on the surface of the concrete specimen. Acoustic emission characteristic parameters are collected at preset measuring points on concrete specimens. These acoustic emission characteristic parameters include cumulative ring count, energy rate, amplitude, average frequency, b-value, and entropy value.

4. The method for predicting the overall stress of concrete lining in cold and arid regions according to claim 3, characterized in that: The principle for establishing a single-point coupling relationship between acoustic emission parameters at the test measurement point and the tensile stress of the specimen is as follows: In the three-point bending fracture test, several acquisition times were set at equal intervals. The cumulative ring count and applied load value at each acoustic emission measuring point were collected at each acquisition time, and the final value of the cumulative ring count when the concrete specimen completely fractured was recorded. The cross-sectional width, cross-sectional height, and support span of the rectangular concrete specimen are measured. At any sampling time, the load value is multiplied by the support span and then divided by 4 to obtain the mid-span bending moment of the concrete specimen. The cube of the cross-sectional height is multiplied by the cross-sectional width and then divided by 12 to obtain the rectangular cross-sectional moment of inertia of the concrete specimen. The mid-span bending moment is multiplied by half of the cross-sectional height and then divided by the rectangular cross-sectional moment of inertia to obtain the nominal tensile stress of the concrete specimen at each sampling time. Damage evolution is characterized by cumulative ring count: the cumulative ring count at each acquisition time is divided by the final value of the cumulative ring count to obtain the real-time damage variable at each acquisition time; the nominal tensile stress at each acquisition time is divided by 1 and the difference between the real-time damage variable to obtain the corrected tensile stress at each acquisition time. Using the cumulative ring count as the independent variable and the modified tensile stress as the dependent variable, multiple sets of discrete data sample points with one-to-one correspondence are formed. By using the Boltzmann function to perform nonlinear fitting on discrete sample points, a single-point coupling relationship between the cumulative ringing count and the corrected tensile stress at a single acoustic emission test point is obtained.

5. The method for predicting the overall stress of concrete lining in cold and arid regions according to claim 4, characterized in that: By combining the strain field data of the specimen, this single-point relationship is extended to a two-dimensional plane to calculate the true tensile stress of all pixels on the specimen surface. The specific principle is as follows: Retrieve the equivalent elastic modulus corresponding to the current degree of frost heave damage, and combine it with the horizontal and vertical normal strains of each pixel to solve the pixel elastic stress according to the plane Hooke's law: solve the plane shear strain from the horizontal and vertical normal strains; The Poisson's ratio of concrete is measured in advance. The horizontal and vertical normal stress components are solved by combining the equivalent elastic modulus, Poisson's ratio and two types of normal strain. The shear stress components are solved by combining the elastic modulus, Poisson's ratio and shear strain. Substitute them into the formula for calculating the plane principal stress to obtain the elastic principal stress of each pixel without considering local damage. For each acoustic emission acquisition point, the damage correction factor specific to that point is obtained by dividing the elastic principal stress of that point by the obtained corrected tensile stress. Radial basis function interpolation is used, with the damage correction factor of each acquisition point as the interpolation constraint, to solve for the damage correction factor corresponding to all pixels on the surface of the specimen. The damage correction factor obtained by interpolation of each pixel is multiplied by the elastic principal stress of that pixel to obtain the true tensile stress of each pixel considering local damage.

6. The method for predicting the overall stress of concrete lining in cold and arid regions according to claim 1, characterized in that: The model hyperparameters are optimized using a grid search combined with K-fold cross-validation to select the model with the highest prediction accuracy. The specific principle is as follows: Using composite aggregate replacement rate, cumulative ring count, energy rate, amplitude, average frequency, b-value, and entropy as model input features, and the actual tensile stress of each pixel as the output label, a pixel-level supervised regression dataset is constructed. Six machine learning models were constructed, including logistic regression, support vector machine, artificial neural network, random forest, extreme gradient boosting tree and adaptive boosting tree, and multiple sets of hyperparameter combinations to be traversed were preset for each model. A grid search is used to traverse all hyperparameter combinations, and K-fold cross-validation is used to evaluate the performance of each set of parameters: the dataset is divided into K non-overlapping subsets, K-1 subsets are selected in turn to train the model, and the remaining 1 subset is used as the validation set. After K complete cycles, the average of the K rounds of validation metrics is taken as the performance score of the hyperparameter set; after traversing all combinations, the optimal hyperparameter combination for each model is determined. Each model was trained using optimal hyperparameters. The coefficient of determination and root mean square error between the model's predicted tensile stress and the actual tensile stress were calculated. The two evaluation indicators were normalized and weighted by setting weighting coefficients. A unified accuracy judgment index was obtained by calculating the weighted difference. Compare the accuracy judgment index of all models and select the model with the highest accuracy judgment index value.

7. The method for predicting the overall stress of concrete lining in arid and cold regions according to claim 1, characterized in that: The full-field strain field of the structure is reconstructed, and pixel-level spatial registration is performed between the full-field strain field data and the acoustic emission monitoring data. The specific principle is as follows: A non-contact visual measurement method, identical to that used for constructing the strain field on the surface of the specimen, was employed to acquire images of the concrete lining surface to be measured. The horizontal and vertical normal strains of each pixel in the lining were then calculated. The horizontal and vertical normal strains of a single pixel were integrated into a two-dimensional strain vector, and the set of two-dimensional strain vectors corresponding to all pixels constituted the full-field strain field of the lining. Several acoustic emission monitoring points were set up on the surface of the lining to be tested, and acoustic emission characteristic parameters of each point were collected. The inverse distance weighted interpolation algorithm was used, with the parameters of each monitoring point as the interpolation constraint, to solve the acoustic emission characteristic parameters corresponding to each pixel in the entire lining area. A unified image pixel coordinate system for the full-field strain field and interpolated pixel-level acoustic emission characteristic parameters is established to achieve one-to-one correspondence between pixels in the same space for two types of multi-source monitoring data, thus completing pixel-level spatial registration.

Images