UHPC entity core sample strength determination method and system based on multi-source data fusion

Abstract

The application discloses a UHPC entity core sample strength determination method and system based on multi-source data fusion, belongs to the technical field of concrete nondestructive testing, and aims to solve the problems of parameter solidification of traditional core strength detection, strength calculation only by relying on macro load, neglect of determination error caused by fiber distribution, internal defects and damage mode. It comprises collecting core sample basic characteristics, predicting mechanics parameters, adaptively configuring test parameters, synchronously collecting load, displacement and DIC image, coupling two-stage strength correction of DIC mesoscopic analysis, and regional strength statistical evaluation. The application fuses ultrasonic, machine vision and digital image correlation technology, relies on real material characteristics to adaptively match the detection scheme, and completes the strength correction based on the quantitative defects and failure forms of the mesoscopic strain. The method greatly reduces the UHPC entity core sample strength determination error, and the detection result can be completely traced and closed-loop iterative optimization of the database is realized.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology for concrete, and more specifically to a method and system for determining the strength of UHPC solid core samples based on multi-source data fusion. Background Technology

[0002] The compressive strength of ultra-high performance concrete (UHVPC) structures is a core control indicator for evaluating structural load-bearing capacity, construction quality, and service safety. Currently, the strength testing of UHVPC structures in engineering projects primarily employs traditional core drilling methods, relying on macroscopic load test results from core samples to calculate structural strength. This is currently the most mainstream method for verifying structural quality. However, in actual engineering testing, UHVPC materials are highly sensitive to construction processes, molding conditions, and curing environments. Fluctuations in on-site pouring, vibration, and curing processes can easily lead to significant individual differences in the internal density and fiber distribution of core samples. Furthermore, inherent defects such as micropores and microcracks are unavoidable, resulting in significant heterogeneity in the mechanical properties of the core samples. Traditional testing methods rely solely on standardized and fixed test parameters, calculating strength values ​​based only on ultimate loads and geometric dimensions. They fail to consider the impact of internal defects, microscopic damage evolution, and differences in failure modes on strength, thus failing to accurately reflect the true mechanical properties of UHVPC core samples and easily leading to significant deviations in strength measurement results. Meanwhile, existing testing technologies struggle to achieve refined performance evaluation of individual core samples and comprehensive assessment of the overall quality of the entire structural area. The traceability, accuracy, and systematic nature of the testing data are insufficient, failing to meet the high-precision, standardized, and multi-dimensional strength testing and quality assessment requirements of ultra-high performance concrete (UHPC) solid structures in current major engineering projects. Therefore, to overcome these limitations, this invention proposes a method and system for determining the strength of UHPC solid core samples based on multi-source data fusion. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for measuring the strength of UHPC solid core samples based on multi-source data fusion, thereby solving the problems of large deviations in traditional UHPC core strength testing results, inability to quantify the impact of internal defects and failure modes on load-bearing capacity, and lack of comprehensive evaluation methods for material properties of the entire structural area.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for determining the strength of UHPC solid core samples based on multi-source data fusion includes:

[0006] All UHPC core samples obtained from drilling in the target area are identified and encoded, and non-destructive testing is performed to obtain the basic feature set of each UHPC core sample.

[0007] Using the basic feature set and the geometric dimension information of the UHPC solid core sample as input features, the prior mechanical parameter set of the UHPC solid core sample is predicted to match the loading parameters and DIC detection parameters corresponding to the UHPC solid core sample, and mechanical loading control commands and DIC acquisition control commands are generated.

[0008] A detection pattern is prepared on the surface of the UHPC solid core sample, and the load data, displacement data and surface sequence digital images of the UHPC solid core sample are collected and stored in real time based on the mechanical loading control command and the DIC acquisition control command.

[0009] The load and displacement data are preprocessed and standardized to obtain the uncorrected strength value of the UHPC solid core sample. DIC analysis is performed on the surface sequence digital images to generate a full-field principal strain cloud map and extract microscopic feature parameters to obtain the strength correction coefficient. The measured strength value of the UHPC solid core sample is then calculated.

[0010] Based on the measured strength values, failure mode types, and equivalent defect diameters of all UHPC solid core samples within the target area, statistical analysis is performed to obtain representative values ​​of UHPC solid strength in the target area.

[0011] Specifically, the steps for calculating the measured strength value of the UHPC solid core sample include:

[0012] The load and displacement data are preprocessed and standardized to obtain the uncorrected strength values ​​of the UHPC solid core sample.

[0013] DIC analysis was performed on the surface sequence digital images to generate the full-field displacement field and full-field principal strain cloud map at each moment during the entire loading process, and the microscopic feature parameters were extracted.

[0014] Extract the mesoscopic characteristic parameters at the ultimate load moment, and identify the failure mode type of the UHPC solid core sample based on the standard characteristic patterns of failure mode types;

[0015] Retrieve the full-field principal strain cloud map at the moment when the load equals the elastic boundary load, calculate the regional average strain and strain standard deviation of the principal strain of all pixels in the region of interest, so as to identify the defect strain concentration area and quantify the equivalent diameter of the defect.

[0016] Based on the failure mode type and equivalent diameter of the defect in the UHPC solid core sample, correction coefficients are matched to the uncorrected strength value to generate the measured strength value of the UHPC solid core sample.

[0017] Specifically, DIC analysis is performed on the surface sequence digital images to generate the full-field displacement field and full-field principal strain contour map at each moment during the entire loading process, and mesoscopic feature parameters are extracted, including:

[0018] The region of interest for DIC calculation of the UHPC solid core sample is defined, and the surface sequence digital images are cropped and segmented to obtain each frame of the image of interest.

[0019] Feature point tracking and matching are performed on adjacent frames of interest images to solve for the horizontal and vertical displacement components of each pixel, and the full-field displacement field corresponding to each frame of interest image is generated.

[0020] Based on the full-field displacement field of each frame, differential solution is performed to calculate the principal strain value of each pixel, which is mapped to a gradient visualization color spectrum to generate a full-field principal strain cloud map.

[0021] By traversing all frames of the full-field principal strain contour map and synchronously recorded load data, mesoscopic characteristic parameters are extracted.

[0022] Specifically, the full-field displacement field corresponding to each frame of the image of interest is generated, including:

[0023] The image of interest in the current frame is divided into grids to obtain a computational subset of the image of interest in the current frame;

[0024] Using the computational subsets of the previous frame of the image of interest as the matching benchmark, the gray-level similarity between each pair of computational subsets is calculated, and the normalized correlation coefficient between the two sets of computational subsets is obtained by zero-mean normalization.

[0025] The normalized correlation coefficient is used as the matching discrimination index, and a global traversal search is performed to complete the feature tracking and matching of all computational subsets between adjacent frames of interest.

[0026] Based on the coordinate offset of the matched subset, interpolation is used to calculate the horizontal and vertical displacement components of each pixel in the grid where the subset is located. All pixel displacement data are integrated to output the full-field displacement field corresponding to each frame of the image of interest.

[0027] Specifically, mesoscopic characteristic parameters at the ultimate load time are extracted, and the failure mode type of the UHPC solid core sample is identified based on the standard characteristic patterns of failure mode types, including:

[0028] From detailed characteristic parameters, input indicators for determining the failure mode at the ultimate load moment are selected.

[0029] Obtain the standard feature patterns corresponding to each pre-stored damage mode type, wherein the standard feature patterns include a combination of judgment input index intervals for the corresponding damage mode type;

[0030] The mesoscopic characteristic parameters extracted from the current UHPC solid core sample at the ultimate load moment are compared one by one with the standard characteristic patterns corresponding to various pre-stored failure mode types to determine the failure mode type of the current UHPC solid core sample.

[0031] Specifically, all UHPC core samples drilled from the target area are identified and encoded, and non-destructive testing is performed to obtain the basic feature set of each UHPC core sample, including:

[0032] Each UHPC entity core sample is identified and coded, and its basic information is bound to it;

[0033] Non-destructive testing was performed on each UHPC solid core sample, including:

[0034] Digital images of both ends of a UHPC solid core sample are acquired and input into a pre-trained first image recognition model to identify the contours of all steel fibers within the end faces and to statistically analyze the geometric parameters of the steel fibers. Based on the geometric parameters of the steel fibers, the actual fiber volume content and fiber orientation distribution coefficient of the UHPC solid core sample are calculated.

[0035] The actual fiber volume content refers to the percentage of actual volume occupied by steel fibers in a unit volume UHPC solid core sample; the fiber orientation distribution coefficient is calculated by statistically analyzing the distribution variance of the projection angles of all steel fiber end faces.

[0036] The longitudinal wave velocity of the UHPC solid core sample is measured, and combined with the actual fiber volume content and fiber orientation distribution coefficient, a set of basic features corresponding one-to-one with the identification code of the UHPC solid core sample is generated.

[0037] Specifically, the mechanical loading control commands and DIC acquisition control commands are generated, including:

[0038] The basic feature set and the geometric dimension information of the UHPC solid core sample are used as input features and fed into a pre-trained mechanical parameter prediction model. The output is the set of prior mechanical parameters corresponding to the UHPC solid core sample.

[0039] The prior mechanical parameter set is input into the pre-built detection parameter mapping relationship library to match the loading parameters and DIC detection parameters corresponding to the UHPC solid core sample;

[0040] Based on the loading parameters, mechanical loading control commands are generated, and based on the DIC detection parameters, DIC acquisition control commands are generated. The mechanical loading control commands and DIC acquisition control commands are then synchronized and started and run according to a preset time base.

[0041] Specifically, the mechanical parameter prediction model is used to establish a nonlinear mapping relationship between input features and prior mechanical parameters. Its training and prediction process includes:

[0042] Multiple sets of UHPC solid core samples with different input characteristics were obtained, and a standard uniaxial compressive strength test was performed on each UHPC solid core sample to obtain the corresponding measured mechanical parameters.

[0043] The basic feature set, geometric dimension information and corresponding measured mechanical parameters of each UHPC solid core sample are associated and stored to form a training dataset;

[0044] Initialize the mechanical parameter prediction model to fit the nonlinear relationship between the input features and the prior mechanical parameters;

[0045] The training dataset is divided into a training subset and a validation subset. The initialized mechanical parameter prediction model is trained using the training subset, and the trained mechanical parameter prediction model is validated using the validation subset, thus obtaining the trained mechanical parameter prediction model.

[0046] The basic feature set and geometric dimension information of the current UHPC solid core sample are input into the trained mechanical parameter prediction model, and the prior mechanical parameter set corresponding to the current UHPC solid core sample is output.

[0047] Specifically, the detection parameter mapping database is used to store combinations of loading parameters and DIC detection parameters corresponding to different prior mechanical parameter ranges. The establishment and matching process of the detection parameter mapping database includes:

[0048] Select the stratification factor parameters from the prior set of mechanical parameters and divide the stratification factor parameters into multiple mechanical property ranges;

[0049] For each mechanical performance range, representative UHPC solid core samples within the corresponding mechanical performance range were selected for comparative tests. Each set of comparative tests used a different combination of test parameters to be optimized.

[0050] Using the comprehensive evaluation value of detection accuracy and detection efficiency as the test evaluation index, a regression model between the combination of detection parameters to be optimized and the comprehensive evaluation value is established. The extreme value of the regression model is solved to obtain the combination of loading parameters and DIC detection parameters corresponding to each mechanical performance range.

[0051] Each mechanical performance range and its corresponding loading parameters are associated and stored with the DIC test parameters to form a test parameter mapping relationship library;

[0052] By comparing the stratification factor parameters corresponding to the prior mechanical parameter set of the current UHPC solid core sample with all mechanical property ranges in the detection parameter mapping relationship library, the mechanical property range to which the current UHPC solid core sample belongs is determined, and the loading parameters and DIC detection parameters corresponding to the current UHPC solid core sample are obtained.

[0053] Specifically, the steps for real-time acquisition and storage of load data, displacement data, and surface sequence digital images of UHPC solid core samples include:

[0054] A detection pattern suitable for DIC detection is prepared on the side of a UHPC solid core sample. The detection pattern is used to provide surface feature markings for the DIC detection system.

[0055] The mechanical loading control command and the DIC acquisition control command are executed synchronously. According to the mechanical loading control command, the load data and displacement data of the UHPC solid core sample are acquired and stored in real time. According to the DIC acquisition control command, the surface sequence digital images of the UHPC solid core sample are acquired and stored in real time.

[0056] Specifically, the load and displacement data are preprocessed and standardized to obtain the uncorrected strength values ​​of the UHPC solid core sample, including:

[0057] Based on the preprocessed load and displacement data, load-displacement curves are plotted, and the ultimate load value is determined by peak extraction method.

[0058] Based on the geometric dimensions and end-face flatness of the UHPC solid core sample, the ultimate load value is standardized and corrected to obtain the uncorrected strength value of the UHPC solid core sample; the standardization correction includes aspect ratio correction and end-face flatness correction.

[0059] The UHPC solid core sample intensity measurement system based on multi-source data fusion includes:

[0060] The information acquisition module is used to identify and encode all UHPC core samples drilled in the target area, and to perform non-destructive testing to obtain the basic feature set of each UHPC core sample.

[0061] The parameter configuration module is used to take the basic feature set and the geometric dimension information of the UHPC solid core sample as input features, predict the prior mechanical parameter set of the UHPC solid core sample, match the loading parameters and DIC detection parameters corresponding to the UHPC solid core sample, and generate mechanical loading control commands and DIC acquisition control commands.

[0062] The synchronous test acquisition module is used to prepare test patterns on the surface of the UHPC solid core sample, and based on the mechanical loading control command and the DIC acquisition control command, to acquire and store the load data, displacement data and surface sequence digital images of the UHPC solid core sample in real time.

[0063] The strength calculation module is used to preprocess and standardize the load and displacement data to obtain the uncorrected strength value of the UHPC solid core sample, and to perform DIC analysis on the surface sequence digital image to generate a full-field principal strain cloud map and extract microscopic feature parameters to obtain the strength correction coefficient, and calculate the measured strength value of the UHPC solid core sample.

[0064] The performance statistics module is used to perform statistical analysis based on the measured strength values, failure mode types, and equivalent defect diameters of all UHPC solid core samples within the target area to obtain representative values ​​of UHPC solid strength in the target area.

[0065] The beneficial effects of this invention are:

[0066] This invention achieves precise acquisition of the true fiber distribution and internal density characteristics of UHPC core samples through unique identification encoding and multi-dimensional non-destructive feature acquisition. This overcomes the information distortion problems caused by traditional testing methods that rely on design theoretical parameters. By leveraging a trained mechanical parameter prediction model and a mapping database of testing parameters, it enables adaptive intelligent matching of testing parameters and synchronous triggering of dual systems. This effectively solves the problems of fixed testing parameters, poor adaptability, and asynchronous macro- and micro-scale data in traditional methods. Furthermore, by combining DIC full-field deformation analysis technology, it accurately extracts the micro-damage and strain evolution characteristics of the core sample throughout the loading process, intelligently identifies the core sample failure mode, and quantifies the internal defect size. Through the relationship between failure mode and defect size... The dual correction mechanism provides refined calibration of core sample strength, effectively overcoming the shortcomings of traditional methods that rely solely on macroscopic load calculations and neglect the influence of microscopic damage and intrinsic defects, resulting in large measurement errors. Simultaneously, by performing stratified statistical analysis on multi-core sample test data of the target area and removing abnormal defect samples, a representative value for regional strength is obtained. This achieves an organic combination of precise single-core sample evaluation and comprehensive evaluation of the construction quality of the entire structural area. Furthermore, the entire testing system can continuously iterate and optimize the model and parameter matching rules based on measured sample data, effectively improving the accuracy, stability, and versatility of UHPC solid core sample strength testing, and ensuring the authenticity, reliability, and traceability of UHPC solid structural strength test results for major projects. Attached Figure Description

[0067] Figure 1 This is a flowchart of the method for measuring the strength of UHPC solid core samples based on multi-source data fusion according to the present invention;

[0068] Figure 2 This is a flowchart illustrating the process of obtaining the basic feature set of each UHPC solid core sample according to the present invention;

[0069] Figure 3 This is a flowchart illustrating the generation of mechanical loading control commands and DIC acquisition control commands for this invention.

[0070] Figure 4 This is a flowchart illustrating the calculation of the measured strength values ​​of the UHPC solid core sample according to the present invention. Detailed Implementation

[0071] Example 1:

[0072] Please see Figure 1This embodiment provides a method for determining the strength of UHPC solid core samples based on multi-source data fusion. It is applicable to core drilling strength testing of ultra-high performance concrete (UHPC) solid structures in fields such as bridge engineering, building engineering, and marine engineering. The method includes:

[0073] Step S1: Identify and encode all UHPC core samples obtained from drilling in the target area, and record the basic information corresponding to each UHPC core sample; perform non-destructive testing on each UHPC core sample through the basic feature acquisition unit to obtain the basic feature set of each UHPC core sample.

[0074] The target area mentioned in this embodiment refers to a pre-defined continuous testing area in the UHPC solid structure to be tested, which has the same design mix ratio and construction process, such as the mid-span area of ​​a certain span of the main beam of a UHPC bridge, or a certain floor area of ​​the core tube of a UHPC in a high-rise building.

[0075] Please see Figure 2 Furthermore, the steps for obtaining the basic feature set of each UHPC entity core sample include:

[0076] Step S11: Identify and encode each UHPC core sample and bind its basic information; the identification code adopts a combination of numbers and letters, and the code content includes at least the project number, target area number, and UHPC core sample drilling sequence number; the basic information includes at least the drilling location coordinates, design mix ratio information, design strength grade, UHPC core sample geometric dimension information, and end face processing quality information, and the UHPC core sample geometric dimension information includes the UHPC core sample diameter, UHPC core sample height, and UHPC core sample height-to-diameter ratio.

[0077] Step S12: Non-destructive testing is performed on each UHPC solid core sample through the basic feature acquisition unit. This is to obtain multi-dimensional feature parameters that can reflect the true material composition, fiber distribution state and internal density of the UHPC solid core sample without destroying the structural integrity of the UHPC solid core sample. This provides a reliable input basis for subsequent prediction of a priori mechanical parameters. The basic feature acquisition unit includes an image acquisition unit and an ultrasonic detection unit.

[0078] Step S121: Acquire digital images of both ends of the UHPC solid core sample using the image acquisition unit. In one specific implementation, the image acquisition unit employs an industrial area array camera with a resolution of no less than 20 megapixels. During acquisition, the end faces of the UHPC solid core sample are placed horizontally directly below the camera. The camera's focal length and lighting conditions are adjusted to ensure that the steel fiber outline is clearly discernible in the image, without significant reflections or shadows.

[0079] Step S122: Input the acquired digital image into the pre-trained first image recognition model to identify the contours of all steel fibers within the end face and statistically analyze the geometric parameters of the steel fibers, including the length, diameter, and end face projection angle of each steel fiber. As a specific implementation, the first image recognition model uses a Mask R-CNN instance segmentation model, trained using no fewer than 10,000 UHPC end face images labeled with steel fiber contours.

[0080] Step S123: Based on the statistically obtained steel fiber geometric parameters, calculate the actual fiber volume content and fiber orientation distribution coefficient of the UHPC solid core sample. The actual fiber volume content refers to the percentage of actual volume occupied by steel fibers in a unit volume of the UHPC solid core sample, estimated using the end-face steel fiber statistical method. Specifically, based on the average length, average diameter, and number of all steel fibers identified within the end face of the UHPC solid core sample, calculate the total volume of steel fibers within the end face. Then, combining the end face area of ​​the UHPC solid core sample and the effective depth of the steel fibers, calculate the steel fiber volume content within a unit volume of the UHPC solid core sample. The effective depth refers to the maximum depth range perpendicular to the end face within the UHPC solid core sample corresponding to steel fibers that can be clearly identified and have their outlines completely extracted from the end-face digital image. It is the core correction parameter for converting the two-dimensional end-face fiber statistical results into three-dimensional volume content. Its value is determined through pre-test calibration and image resolution matching. As a specific implementation method, the effective depth is taken as 3 to 5 times the average diameter of the steel fibers. Commonly used straight steel fibers with a diameter of 0.2 mm have an effective working depth ranging from 0.6 mm to 1.0 mm. The fiber orientation distribution coefficient is a dimensionless parameter characterizing the uniformity of steel fiber orientation within the end face of the UHPC solid core sample. Its value ranges from 0 to 1. A value closer to 1 indicates a more random and uniform fiber orientation, while a value closer to 0 indicates a more concentrated fiber orientation in a certain direction. It is obtained by statistically analyzing the distribution variance of the projection angles of all steel fiber end faces. Specifically, the projection angle range of 0° to 180° is divided into several consecutive angle intervals. The number of steel fibers in each angle interval is counted, and the actual variance of the fiber quantity distribution in each interval is calculated. Then, the actual variance is divided by the maximum theoretical variance under this statistical method for normalization to obtain the fiber orientation distribution coefficient. As a specific implementation method, 0° to 180° is divided into 18 angle intervals of 10° each. The maximum theoretical variance is the variance value when all steel fibers are concentrated in a single angle interval.

[0081] Step S124: The longitudinal wave velocity of the UHPC solid core sample is measured using the transmission method via the ultrasonic detection unit. In one specific implementation, the transmission frequency of the ultrasonic detection unit is preset to a range of 50kHz to 100kHz, with the specific value adjusted according to the diameter of the UHPC solid core sample. The larger the diameter of the UHPC solid core sample, the lower the transmission frequency. During measurement, coupling agent is applied to both ends of the UHPC solid core sample. The ultrasonic transmitting probe and receiving probe are respectively placed close to the center positions of both ends of the UHPC solid core sample. Measurements are taken three times consecutively, and the average value is taken as the longitudinal wave velocity of the UHPC solid core sample.

[0082] Step S13: Integrate the actual fiber volume content, fiber orientation distribution coefficient obtained in step S123 and the longitudinal wave velocity obtained in step S124 to generate a set of basic features that correspond one-to-one with the identification code of the UHPC physical core sample, and store the set of basic features in association with the basic information recorded in step S11.

[0083] Specifically, step S1 is the fundamental data standardization and real feature extraction stage of the entire UHPC solid core sample strength determination method based on multi-source data fusion. Its core design addresses key issues in existing technologies, such as unreliable prior information, difficulty in tracing UHPC solid core sample data, and the highly destructive and low-precision methods for acquiring material characteristics. It provides a high-quality, traceable, and reliable input foundation for all subsequent core steps, including intelligent prediction of prior mechanical parameters, adaptive detection scheme generation, and microscopic feature strength correction. In UHPC solid core sample testing scenarios, traditional methods generally directly use theoretical parameters such as design mix proportions and design fiber content as the basis for subsequent calculations. However, UHPC materials are extremely sensitive to construction processes. During on-site mixing, pouring, and vibration, uneven distribution of steel fibers, local clumping, or missing fibers are easily observed, resulting in a significant deviation between the actual fiber volume content and the design value. Furthermore, fiber orientation exhibits significant anisotropy due to the influence of the pouring flow direction. Simultaneously, fluctuations in on-site curing conditions lead to significant differences in the internal density of the UHPC solid core sample. These deviations in real material characteristics are the root cause of large errors in subsequent strength determinations. Furthermore, engineering testing often requires processing multiple UHPC core samples from different regions and batches simultaneously. Without a unified identification and information management system, problems such as UHPC core sample confusion, data mismatch, and untraceable results can easily arise, affecting the standardization and authority of the testing work. Step S1 establishes a unique digital identity for each UHPC core sample by binding the identification code and basic information from step S11, ensuring traceability throughout the entire process from UHPC core sample drilling and testing to report generation, fundamentally avoiding UHPC core sample confusion and data mismatch issues. Step S12 utilizes a multi-dimensional non-destructive testing technology system to rapidly acquire real material characteristics without damaging the structural integrity of the UHPC core sample or affecting subsequent mechanical tests. Step S13 integrates and associates features, unifying all basic information and testing features under the corresponding UHPC core sample's digital identity, forming a complete UHPC core sample basic file. This ensures that all subsequent steps can be performed based on the same real and unified data source for calculation and analysis.

[0084] Step S2: Using the basic feature set and the geometric dimension information of the UHPC solid core sample as input features, predict the prior mechanical parameter set of the UHPC solid core sample to match the loading parameters and DIC detection parameters corresponding to the UHPC solid core sample, and generate mechanical loading control commands and DIC acquisition control commands.

[0085] Please see Figure 3 Furthermore, the steps for generating mechanical loading control commands and DIC acquisition control commands include:

[0086] Step S21: The basic feature set obtained in step S1 and the geometric dimension information of the UHPC solid core sample are used as input features and input into the pre-trained mechanical parameter prediction model to output the prior mechanical parameter set corresponding to the UHPC solid core sample. The input features include the actual fiber volume content, fiber orientation distribution coefficient, longitudinal wave velocity, UHPC solid core sample diameter, and UHPC solid core sample height-to-diameter ratio. The prior mechanical parameter set includes at least the estimated ultimate compressive strength, the estimated initial cracking load, and the estimated elastic modulus.

[0087] Furthermore, the mechanical parameter prediction model is used to establish a nonlinear mapping relationship between input features and prior mechanical parameters, and its training and prediction process specifically includes:

[0088] Step S211: Construct a training dataset for the mechanical parameter prediction model; obtain multiple sets of UHPC solid core samples with different input features, perform a standard uniaxial compressive strength test on each UHPC solid core sample, and obtain the corresponding measured mechanical parameters, including the measured ultimate compressive strength, the measured initial crack load, and the measured elastic modulus; associate and store the basic feature set, geometric dimension information, and corresponding measured mechanical parameters of each UHPC solid core sample to form a training dataset; perform preprocessing operations on the training dataset, including outlier removal and feature normalization.

[0089] As a specific implementation method, the training dataset contains sample data of no less than a preset number. The preset number of samples is set based on the following criteria: it must cover all UHPC mix proportion ranges, fiber content ranges, curing condition ranges, and UHPC core sample size ranges applicable to this method, and the number of samples must meet the generalization ability requirements of the ensemble learning model to avoid overfitting due to insufficient sample size; for example, the preset number of samples is set to 500 groups. The outlier removal uses the Grubbs test to remove abnormal sample data that deviates from the statistical distribution; the feature normalization process uses the min-max normalization method to map all input features to the interval between 0 and 1, eliminating the influence of different feature dimensions on the model prediction results.

[0090] Step S212: Initialize the mechanical parameter prediction model, which adopts an ensemble learning model to fit the nonlinear relationship between input features and prior mechanical parameters.

[0091] As a specific implementation method, the ensemble learning model adopts a gradient boosting tree model, whose initial hyperparameters include the learning rate, the number of decision trees, the depth of the decision trees, and the minimum number of leaf node samples. The initial hyperparameters are set based on the following: the learning rate needs to balance the model training speed and convergence stability; the number and depth of the decision trees need to balance the model fitting ability and generalization ability; and the minimum number of leaf node samples is used to prevent the model from overfitting. For example, the initial learning rate is set to 0.1, the number of decision trees is set to 100, the decision tree depth is set to 4 layers, and the minimum number of leaf node samples is set to 5.

[0092] Step S213: Divide the training dataset into a training subset and a validation subset. Use the training subset to train the initialized mechanical parameter prediction model and use the validation subset to validate the trained mechanical parameter prediction model. Use a preset model performance evaluation index to evaluate the prediction accuracy of the model. If the model accuracy does not meet the preset requirements, adjust the model hyperparameters and retrain until the model accuracy meets the preset requirements, and obtain the trained mechanical parameter prediction model.

[0093] As a specific implementation, the ratio of the training subset to the validation subset is 7:3; the model performance evaluation metrics include the coefficient of determination and the mean absolute percentage error; the preset requirement is that the mean absolute percentage error of the model does not exceed a preset error threshold, and the preset error threshold is set based on the following: it is necessary to ensure that the prediction error of the prior mechanical parameters is less than half of the width of the mechanical performance range in the subsequent detection parameter mapping relationship library, so as to ensure that the detection parameter matching error will not be caused by the prediction error; for example, the preset error threshold is set to 10%.

[0094] Step S214: Input the basic feature set and geometric dimension information of the current UHPC solid core sample obtained in step S1 into the trained mechanical parameter prediction model. The model automatically calculates and outputs the estimated ultimate compressive strength, estimated initial crack load and estimated elastic modulus corresponding to the UHPC solid core sample, forming a priori mechanical parameter set.

[0095] Step S22: Input the set of prior mechanical parameters obtained in step S21 into the pre-constructed detection parameter mapping relationship library to match the loading parameters and DIC detection parameters corresponding to the UHPC solid core sample; the loading parameters include at least the loading rate; the DIC detection parameters include at least the staged image acquisition frequency and the camera magnification.

[0096] Furthermore, the detection parameter mapping database is used to store combinations of loading parameters and DIC detection parameters corresponding to different prior mechanical parameter ranges. Its establishment and matching process specifically includes:

[0097] Step S221: Select the stratification factor parameters from the prior set of mechanical parameters, and divide the stratification factor parameters into multiple continuous and non-overlapping mechanical property intervals according to the mechanical property distribution characteristics of UHPC materials; select the loading parameters and DIC detection parameters as the detection parameters to be optimized, and set multiple level values ​​covering the engineering standard range for each detection parameter to be optimized.

[0098] As a specific implementation method, the stratification factor parameter is selected based on the estimated ultimate compressive strength. The estimated ultimate compressive strength is the most critical mechanical performance indicator of the UHPC solid core sample, and its value directly determines the failure rate, crack propagation rate, and DIC acquisition frequency requirements of the UHPC solid core sample. The mechanical performance range is divided based on the following: the range width must be greater than the strength fluctuation range corresponding to the preset error threshold in step S213, to ensure that the mechanical performance difference of the UHPC solid core sample within a single range does not cause significant changes in the optimal detection parameters. For example, the estimated ultimate compressive strength is divided into three mechanical performance ranges: a low-strength range of less than 140 MPa, a medium-strength range of 140 MPa to 160 MPa, and a high-strength range of greater than 160 MPa. The detection parameters to be optimized include loading rate, phased image acquisition frequency, and camera magnification. The loading rate is set to 0.6 MPa / s, 0.9 MPa / s, 1.2 MPa / s, and 1.5 MPa / s. The phased image acquisition frequency is set according to the loading stage: 1 frame / second and 2 frames / second for the elastic stage, 5 frames / second and 10 frames / second for the initial cracking stage, and 15 frames / second and 25 frames / second for the extreme stage. The camera magnification is set to 1x, 1.5x, and 2x.

[0099] Step S222: For each mechanical performance range, select a representative UHPC solid core sample within that range and conduct comparative experiments according to orthogonal experiments. Each comparative experiment uses a different combination of detection parameters to be optimized. Using the comprehensive evaluation value of detection accuracy and detection efficiency as the experimental evaluation index, analyze the influence of each detection parameter to be optimized on the comprehensive evaluation value through response surface methodology, establish a regression model between the combination of detection parameters to be optimized and the comprehensive evaluation value, and solve for the extreme value of the regression model to obtain the loading parameters and DIC detection parameter combination corresponding to each mechanical performance range.

[0100] As a specific implementation method, the orthogonal experiment adopts a three-factor, four-level orthogonal array, and conducts no less than 16 sets of comparative tests in each mechanical property range. The comprehensive evaluation value is calculated by weighted summation, where the weight of detection accuracy is set to 0.7 and the weight of detection efficiency is set to 0.3. The basis for this setting is that detection accuracy is the core requirement of strength measurement and should be given priority, while detection efficiency is taken into consideration on this basis. The detection accuracy is characterized by the average relative error between the strain value measured by DIC and the strain value measured by the standard strain gauge. The detection efficiency is characterized by the sum of the total detection time of a single UHPC solid core sample and the total data processing time. The response surface methodology is used to establish a quadratic polynomial regression model between the detection parameters to be optimized and the comprehensive evaluation value. By taking the partial derivative of the regression model and setting it equal to zero, the combination of loading parameters and DIC detection parameters that maximizes the comprehensive evaluation value is obtained.

[0101] Step S223: Associate and store each mechanical property range and its corresponding loading parameters with DIC detection parameters to form a structured detection parameter mapping relationship library; compare the hierarchical factor parameters corresponding to the prior mechanical parameter set of the current UHPC solid core sample obtained in step S21 with all mechanical property ranges in the detection parameter mapping relationship library to determine the mechanical property range to which the current UHPC solid core sample belongs, and extract the loading parameters and DIC detection parameter combination corresponding to the mechanical property range as the loading parameters and DIC detection parameter combination corresponding to the UHPC solid core sample.

[0102] As one specific implementation method, the detection parameter mapping relationship library is stored in a relational database, which supports subsequent updates and optimizations of the content in the library by adding new experimental data.

[0103] Step S23: Generate mechanical loading control commands based on the loading parameters matched in step S22, and generate DIC acquisition control commands based on the DIC detection parameters; establish a time synchronization association between the mechanical loading system and the DIC detection system through a hardware synchronization trigger unit to ensure that the mechanical loading control commands and the DIC acquisition control commands can be started and run synchronously according to the preset time base.

[0104] As a specific implementation method, the mechanical loading system refers to a material mechanical property testing device capable of applying uniaxial compressive load to a UHPC solid core sample at a preset loading rate and acquiring load and displacement values ​​in real time. Its accuracy level must meet the requirements of the current national standard for concrete compressive strength testing. The DIC detection system refers to a non-contact optical measurement system based on digital image correlation technology, which calculates the full-field displacement and strain fields of the specimen surface by acquiring a sequence of digital images of the specimen surface. It includes at least two industrial area array cameras, an image acquisition card, an illumination source, and supporting image analysis software. The hardware synchronization triggering unit adopts a high-precision pulse signal generator, and its synchronization accuracy is not lower than a preset synchronization accuracy threshold. The preset synchronization accuracy threshold is set based on the following: it must be less than the single-frame acquisition time interval of the DIC detection system at the highest acquisition frequency to ensure that the corresponding error between the macroscopic load data and the DIC microscopic image data on the time axis does not exceed one acquisition cycle. For example, the preset synchronization accuracy threshold is set to 1 microsecond.

[0105] Specifically, the core design of step S2 aims to address key issues in existing technologies, such as rigid testing schemes, the inability to balance accuracy and efficiency, and asynchrony among multiple systems. It achieves personalized matching between the testing scheme and the mechanical properties of UHPC solid core samples, providing precise control commands for subsequent synchronous testing. In UHPC solid core sample testing scenarios, traditional methods generally employ uniform loading rates and DIC acquisition parameters, failing to consider the heterogeneity of mechanical properties caused by differences in material composition, fiber distribution, and curing conditions among different core samples. For high-strength, high-toughness core samples, excessively low image acquisition frequencies miss critical moments of crack initiation and propagation, resulting in an inability to accurately capture the microscopic failure process. For low-strength, low-toughness core samples, excessively high acquisition frequencies generate a large amount of redundant data, increasing the time and computational cost of subsequent data processing. Simultaneously, a uniform loading rate can lead to insufficient failure processes or excessively rapid loading of some core samples, resulting in inertial effects and affecting the accuracy of strength measurements. Step S2, through multi-feature fusion mechanical parameter prediction in step S21, accurately predicts the mechanical performance range of the core sample based on the real material characteristics obtained in step S1 using a pre-trained ensemble learning model, avoiding prediction bias caused by reliance on design parameters. Step S22, through adaptive matching of detection parameters, generates personalized loading and acquisition parameters for core samples with different mechanical properties based on a detection parameter mapping relationship library established by orthogonal experiments and response surface methodology. This minimizes redundant data and improves detection efficiency while ensuring no loss of key microscopic information. Step S23, through control command generation and hardware synchronization triggering, achieves high-precision synchronization between the mechanical loading system and the DIC detection system, ensuring strict correspondence between macroscopic and microscopic data on the time axis, laying the foundation for subsequent multi-source data fusion.

[0106] Step S3: Prepare a detection pattern on the surface of the UHPC solid core sample, and start the mechanical loading test and DIC microscopic detection based on the mechanical loading control command and the DIC acquisition control command. Collect and store the load data, displacement data and surface sequence digital images of the UHPC solid core sample in real time until the UHPC solid core sample is completely destroyed.

[0107] In this embodiment, complete failure of the UHPC solid core sample refers to the state in which the load value of the UHPC solid core sample drops below a preset proportion of the ultimate load under uniaxial compressive load, or macroscopic cracks appear through the core sample and it can no longer bear the load; as a specific implementation, the preset proportion is set to 85%.

[0108] Further, step S3 includes:

[0109] Step S31: Prepare a detection pattern suitable for DIC detection on the side of the UHPC solid core sample. The detection pattern is used to provide identifiable surface feature markings for the DIC detection system to achieve accurate measurement of surface deformation.

[0110] As a specific implementation method, the detection pattern is a matte black and white speckle pattern, and its preparation steps are as follows: First, a layer of matte white primer is uniformly sprayed on the side of the UHPC solid core sample. After the primer is completely dry, matte black dots are uniformly sprayed through a spray gun, and the speckle particle size is controlled within the range of 3 to 5 pixels to ensure that the speckle distribution is random and uniform, without obvious agglomeration or blank areas. The purpose of using matte material is to eliminate the high reflectivity of the UHPC solid core sample surface and avoid DIC image recognition failure due to reflection.

[0111] Step S32: Install the UHPC solid core sample with the prepared detection pattern onto the loading platform of the mechanical loading system, and adjust the centering accuracy of the UHPC solid core sample to within the preset centering accuracy threshold; fix the DIC detection system at the preset observation position, complete the camera calibration and system calibration, and ensure that the calibration error does not exceed the preset calibration error threshold.

[0112] As a specific implementation method, the mechanical loading system adopts a universal pressure testing machine; the preset centering accuracy threshold is set to 0.5 degrees, which is based on the fact that when the centering accuracy deviation exceeds 0.5 degrees, the core sample will experience significant eccentric compression, resulting in a lower strength measurement result; the preset calibration error threshold is set to 0.01 pixels, which is based on the fact that calibration error directly affects the strain measurement accuracy of the DIC testing system, and a calibration error of 0.01 pixels can ensure that the strain measurement accuracy meets the engineering testing requirements.

[0113] Step S33: Simultaneously send the mechanical loading control command and DIC acquisition control command generated in step S2 to the mechanical loading system and the DIC detection system, and realize the synchronous start of the two through the hardware synchronization trigger unit; during the test, the mechanical loading system continuously loads according to the mechanical loading control command, and collects and stores the load data and displacement data of the UHPC solid core sample in real time; the DIC detection system collects and stores the surface sequence digital images of the UHPC solid core sample in real time according to the preset stage image acquisition frequency in the DIC acquisition control command, until the UHPC solid core sample is completely destroyed, and then stops loading and image acquisition.

[0114] Specifically, step S3 is a crucial link connecting front-end intelligent decision-making with back-end data fusion analysis. Its core design addresses issues in existing technologies such as non-standard test preparation, asynchronous data timing, and loss of key damage information. It provides high-quality, time-aligned macroscopic and mesoscopic multi-source data for subsequent strength correction based on microscopic features. Step S3 ensures the basic accuracy of DIC measurement through step S31; through strict alignment control and system calibration in step S32, it eliminates the influence of eccentric compression, ensuring the consistency and repeatability of test conditions; and through hardware synchronous triggering and phased adaptive acquisition in step S33, it achieves microsecond-level time alignment of load data, displacement data, and mesoscopic image data while ensuring complete acquisition of key mesoscopic information such as crack initiation and propagation. This provides a reliable data foundation for establishing a quantitative mapping relationship between mesoscopic damage features and macroscopic strength.

[0115] Step S4: Preprocess and standardize the load and displacement data obtained in Step S3 to obtain the uncorrected strength value of the UHPC solid core sample; perform DIC analysis on the surface sequence digital image obtained in Step S3 to generate a full-field principal strain cloud map of the entire loading process and extract mesoscopic feature parameters; automatically identify the failure mode type and defect equivalent diameter of the UHPC solid core sample based on the full-field principal strain cloud map; obtain the corresponding first strength correction coefficient according to the failure mode type, and calculate the second strength correction coefficient according to the defect equivalent diameter; combine the uncorrected strength value, the first strength correction coefficient, and the second strength correction coefficient to calculate the measured strength value of the UHPC solid core sample.

[0116] Please see Figure 4 Furthermore, the calculated measured strength values ​​of the UHPC solid core sample include:

[0117] Step S41: Perform preprocessing on the load data and displacement data obtained in step S3, and standardize and correct the preprocessed load data and displacement data to obtain the uncorrected strength value of the UHPC solid core sample.

[0118] Further, step S41 includes:

[0119] Step S411: Perform data preprocessing operations to remove high-frequency noise and outliers from the load and displacement data; the preprocessing operations include moving average filtering and outlier removal. Moving average filtering is used to eliminate random noise generated by mechanical vibration and electromagnetic interference during the test, and outlier removal is used to remove data points that deviate significantly from the normal trend due to sudden equipment failure.

[0120] As a specific implementation method, the moving average filtering adopts the 5-point moving average method, and the outlier removal adopts the 3σ criterion, that is, outliers that deviate from the mean by more than 3 times the standard deviation are removed.

[0121] Step S412: Based on the load and displacement data after preprocessing, plot the load-displacement curve and determine the ultimate load value by peak extraction method; the load-displacement curve refers to a two-dimensional curve plotted with the axial displacement of the UHPC solid core sample as the abscissa and the axial load applied by the pressure testing machine as the ordinate, which is used to intuitively characterize the entire mechanical response process of the core sample under uniaxial compressive load.

[0122] As a specific implementation method, the peak extraction method is as follows: traverse all data points of the load-displacement curve, and take the maximum value of the load value as the ultimate load value of the core sample; if the load-displacement curve has multiple peaks, take the first peak value as the ultimate load value.

[0123] Step S413: Based on the geometric dimensions and end-face flatness of the UHPC solid core sample, the ultimate load value is standardized and corrected to obtain the uncorrected strength value of the UHPC solid core sample. The end-face flatness refers to the degree of unevenness and undulation of the overall plane of the upper and lower bearing end faces of the UHPC solid core sample and the degree of inclination of the end face relative to the axis, obtained through end-face contour detection data accompanying the end-face processing quality information collected and stored in step S1.

[0124] Specifically, the standardization correction includes two aspects: aspect ratio correction and end face flatness correction. First, based on the actual aspect ratio of the UHPC core sample recorded in step S11, the aspect ratio correction coefficient obtained in advance through standard comparison calibration test is retrieved to correct the ultimate load value, eliminating the strength calculation deviation caused by the difference in the geometric dimensions of the core sample. Then, based on the end face flatness of the UHPC core sample, the end face flatness correction coefficient obtained in advance is retrieved to correct the load value after aspect ratio correction, eliminating the uneven bearing error caused by processing defects such as uneven end face and end face tilt. Finally, the load value after two corrections is divided by the actual bearing area of ​​the UHPC core sample to obtain the uncorrected strength value.

[0125] As a specific implementation method, the aspect ratio correction coefficient and end face flatness correction coefficient are obtained by fitting multiple sets of UHPC core sample calibration tests with the same mix ratio, different sizes, and different end face processing conditions. For example, UHPC mixes with uniform steel fiber content and uniform water-cement ratio are selected, and standard specimens with various diameters and height combinations are cast. At the same time, multiple end face concave and convex processing conditions and end face tilt processing conditions are artificially set, and uniaxial compressive strength calibration tests are carried out. The test strength of the standard regular core sample is used as the benchmark value, and the correction coefficients corresponding to different aspect ratios and different flatness levels are fitted. The actual bearing area of ​​the UHPC solid core sample is calculated based on the core sample diameter recorded in step S11. The uncorrected strength value refers to the preliminary calculated value of the core sample compressive strength that only corrects the systematic errors caused by the geometric dimensions and end face processing quality of the core sample, and does not consider the differences in failure modes and the influence of internal defects.

[0126] Step S42: Perform DIC analysis on the surface sequence digital images obtained in step S3 to generate the full-field displacement field and full-field principal strain cloud map at each moment during the entire loading process, and extract the microscopic feature parameters that can reflect the microscopic failure process of the UHPC solid core sample.

[0127] Further, step S42 includes:

[0128] Step S421: Delineate the region of interest (ROI) for DIC calculation of the UHPC core sample, import the surface sequence digital images acquired in step S3 frame by frame, and perform cropping and segmentation based on the delineated ROI to obtain each frame of ROI image; specifically, the ROI for DIC calculation of the UHPC core sample can be delineated by manual selection or threshold contour recognition algorithm. As a specific implementation method, the complete pressure observation area of ​​the side of the UHPC core sample excluding the chamfers of the upper and lower end faces is selected as the ROI for DIC calculation; the original digital image is cropped using an image mask to remove irrelevant background pixels such as chamfers and the pressure head of the testing machine, and only the effective observation surface image block of the core sample is retained as a single frame of ROI image.

[0129] Step S422: Perform feature point tracking and matching on adjacent frames of interest, solve for the horizontal and vertical displacement components of each pixel, and generate the full-field displacement field corresponding to each frame of interest.

[0130] Further, step S422 includes:

[0131] Step S4221: The image of interest in the current frame is divided into a grid, dividing the entire image of interest into a large number of overlapping computational subsets. Each computational subset contains several pixels, and the grayscale information of the detected image within the computational subset is used as a feature identifier. The computational subset refers to a square local image block formed by expanding outward from a single pixel grid. It is used to carry the grayscale features of the local detected image and serves as the smallest computational unit for cross-frame matching. The setting is based on the following: if the subset size is too small, the internal detected image features will be insufficient, and matching confusion will easily occur; if the subset size is too large, local small deformation information will be lost, and the computational load will be greatly increased. The overlapping step size balances the continuity of the overall displacement and the computational efficiency. As a specific implementation, the size of the computational subset is set to 31×31 pixels, and the computational step size between adjacent subsets is set to 3 pixels, which reduces the overall computation time while ensuring the accuracy of displacement solution.

[0132] Step S4222: Using each computational subset of the previous frame of interest as a matching benchmark, calculate the gray-level similarity between each pair of computational subsets, and obtain the normalized correlation coefficient of the gray-level distribution of the two sets of computational subsets through zero-mean normalization. The gray-level similarity is used to quantify the degree of overlap of the speckle gray-level distribution within the two sets of computational subsets in the previous and next frames. The closer the gray-level distribution is, the higher the similarity value.

[0133] The specific grayscale similarity calculation process is as follows: traverse all pixels within the subset, calculate the square of the grayscale difference of each pixel, and sum the squares of all pixel differences to obtain the sum of squared grayscale differences. The smaller this value, the higher the overlap of grayscale distribution between the two subsets, and the higher the grayscale similarity. Substitute the sum of squared grayscale differences into the zero-mean normalized correlation coefficient calculation formula to eliminate the matching interference caused by the overall brightness change of the subset, and output the normalized correlation coefficient with a value range of [-1, 1]. The closer the coefficient is to 1, the higher the matching degree between the two subsets.

[0134] Step S4223: Using the normalized correlation coefficient as the matching discrimination index, perform a global traversal search to complete the feature tracking and matching of all computational subsets between adjacent frames of interest; specifically, slide search other computational subsets in the entire domain of the current frame of interest, select the computational subset with the largest normalized correlation coefficient value as the matching computational subset, and establish a one-to-one matching mapping relationship between the computational subsets of the previous and next frames.

[0135] Step S4224: Based on the coordinate offsets of the matched subsets, interpolate the horizontal and vertical displacement components of each pixel within the grid of the subset, integrate all pixel displacement data, and output the full-field displacement field corresponding to each frame of the image of interest. The horizontal displacement component represents the positional offset of the pixel along the horizontal axis of the image; the vertical displacement component represents the positional offset of the pixel along the vertical axis of the image. The specific interpolation solution includes: first, extracting the horizontal and vertical coordinate offsets of the center points of all matched subsets as discrete displacement sampling points; using a bicubic spline interpolation algorithm to perform global interpolation fitting on the discrete displacement sampling points, constraining the interpolation surface with the displacement data of multiple adjacent subsets to eliminate displacement abrupt changes caused by single subset matching errors; traversing the coordinates of each pixel in the image of interest, substituting them into the interpolation surface function to solve for the precise horizontal and vertical displacement components of that pixel; storing all pixel coordinates and corresponding dual-component displacement data in pixel row and column order to form a structured full-field displacement field.

[0136] Step S423: Based on the full-field displacement field of each frame, perform differential solution to calculate the principal strain value of each pixel, map the principal strain value to a gradient visualization color spectrum, and generate a full-field principal strain cloud map corresponding to each frame of the loading process; the full-field principal strain cloud map can intuitively display the location of strain concentration, strain concentration range and strain evolution trend on the core sample surface during the loading process.

[0137] Step S424: Traverse the full-field principal strain cloud map and synchronously recorded load data of all frames to extract mesoscopic characteristic parameters; the mesoscopic characteristic parameters refer to physical indicators that can quantitatively characterize the mesoscopic strain evolution, crack initiation and propagation, and energy dissipation laws of UHPC solid core samples during the entire loading process. The mesoscopic characteristic parameters include at least the average principal strain in the elastic stage, the ultimate principal strain corresponding to the ultimate load, the full-field strain concentration factor, the total number of effective cracks, the equivalent length of the principal crack, the average crack propagation rate, and the energy dissipation of the load-principal strain curve integral.

[0138] Specifically, the elastic stage refers to the initial loading stage of the UHPC solid core sample, where stress and strain exhibit a linear proportional relationship, and there is no irreversible plastic damage or microcrack initiation. An elastic boundary load factor is used to divide the elastic stage and the plastic development stage. This elastic boundary load factor is set to 0.3, based on the following: when the load is 30% below the ultimate load, the UHPC material does not experience debonding at the matrix-fiber interface, there is no accumulation of plastic strain, and the mechanical response is entirely within the linear elastic range. This effectively avoids interference from plastic deformation on the extraction of elastic characteristic parameters. For example, if the ultimate load of the core sample is 200 kN, then the elastic boundary load is 60 kN. The average principal strain of the elastic stage is obtained by selecting the instantaneous full-field principal strain cloud map corresponding to the load value equal to the elastic stage boundary load, traversing the principal strain values ​​of all effective pixels within the region of interest in the DIC, and calculating the regional average principal strain through arithmetic mean. This accurately characterizes the overall uniform deformation characteristics of the core sample during the elastic stage. The ultimate principal strain corresponding to the ultimate load is obtained by extracting the maximum principal strain value of all pixels in the region of interest from the full-field principal strain cloud map corresponding to the ultimate bearing moment when the load reaches its peak. This represents the maximum local deformation degree of the core sample surface before failure and is a core indicator for determining the ductile and brittle failure of the core sample. The full-field strain concentration factor is the ratio of the ultimate principal strain to the average principal strain in the elastic stage. It is used to quantitatively characterize the degree of non-uniformity of surface strain concentration when the core sample is loaded to the failure limit state. The larger the value, the more significant the local stress concentration and the more prominent the core sample defects and damage problems. The total number of effective cracks is obtained by adaptive strain threshold segmentation, contour recognition, and removal of excessively small areas from each frame of the full-field principal strain cloud map. The noise strain points and the total number of independent continuous strain concentration zones are obtained; the equivalent length of the main crack is obtained by extracting the full-field principal strain cloud map at the peak of the ultimate load, screening the main crack region with the largest profile length among all effective crack strain zones, and quantifying the length of the main crack region; the average crack propagation rate is obtained by recording the time and crack length corresponding to the first crack appearance frame and the ultimate load frame, and dividing the length difference between the two frames by the time difference; the load-principal strain curve integral dissipation energy is obtained by plotting a curve with the average principal strain of the elastic stage as the strain abscissa and the synchronous load as the ordinate, and calculating the integral area enclosed by the curve and the abscissa using the numerical integration method, which is taken as the deformation energy dissipated per unit area.

[0139] Step S43: Extract the mesoscopic characteristic parameters at the ultimate load moment, and identify the failure mode type of the UHPC solid core sample according to the standard characteristic pattern of the failure mode type; the standard characteristic pattern refers to the combination of fixed quantitative parameter value ranges corresponding to various failure mode types, which is composed of the mesoscopic characteristic parameter ranges corresponding to each failure mode type, for example, including the ultimate principal strain range, the full-field strain concentration factor range, the effective crack number range, and the crack average propagation rate range.

[0140] Further, step S43 includes:

[0141] Step S431: From the mesoscopic characteristic parameters output in step S424, select the failure mode determination input indicators corresponding to the ultimate load moment. The failure mode determination input indicators refer to mesoscopic mechanical parameters that can quantitatively distinguish the compressive failure mode of UHPC solid core samples and characterize the degree of mesoscopic damage and crack propagation characteristics. For example, the ultimate principal strain, the overall strain concentration factor, the total number of effective cracks, and the average crack propagation rate are used as the failure mode determination input indicators.

[0142] Step S432: Obtain the pre-stored standard characteristic patterns corresponding to each failure mode type. All standard characteristic patterns are obtained through statistical calibration of standard compressive strength tests on a large batch of regular UHPC solid core samples with the same mix ratio and no processing defects. Each failure mode type is matched with a unique set of non-overlapping judgment input index intervals. For example, the specific division of each standard characteristic pattern is as follows: In the fiber pull-out ductile failure standard characteristic pattern, the ultimate principal strain is greater than 0.003, the global strain concentration factor is less than 3, the total number of effective cracks is greater than 3, and the average crack propagation rate is less than 10 mm / s. This mode corresponds to the ductile failure characteristics of multiple diffuse cracks and full fiber pull-out energy consumption. In the matrix crushing brittle failure standard characteristic pattern, the ultimate principal strain is less than 0.001, the global strain concentration factor is greater than 10, the total number of effective cracks is equal to 1, and the average crack propagation rate is greater than 10 mm / s. This mode corresponds to the brittle failure characteristics of a single main crack instantaneously penetrating and the matrix crushing first. In the standard characteristic mode of interface transition zone failure, the ultimate principal strain is between 0.001 and 0.003, the global strain concentration factor is between 3 and 10, the total number of effective cracks is 2 to 3, and the average crack propagation rate is between 5 and 10 mm / s. This mode corresponds to failure characteristics where cracks extend along the weak aggregate-matrix interface, the degree of damage is moderate, and the deformation and energy dissipation performance is between ductility and brittleness. In the standard characteristic mode of defect-induced failure, a significant strain concentration zone has appeared in the global principal strain contour map corresponding to the elastic boundary load. The initial crack initiation location coincides with this strain concentration zone, and the four core parameters at the ultimate load time have no fixed range constraints. This mode corresponds to abnormal failure characteristics where internal defects induce stress concentration prematurely.

[0143] Step S433: Compare the mesoscopic feature parameters extracted from the current UHPC solid core sample at the ultimate load moment with the standard feature patterns corresponding to various pre-stored failure mode types one by one. Match the standard feature patterns in which all failure mode determination input indicators fall into the corresponding determination input indicator range to determine the failure mode type of the current UHPC solid core sample.

[0144] As a specific implementation method, if the current core sample parameters simultaneously meet multiple sets of interval conditions, the defect-induced failure standard feature pattern is matched first; if there is no matching standard feature pattern, it is marked as an undefined mixed failure mode.

[0145] Step S44: Retrieve the full-field principal strain cloud map at the moment when the load equals the elastic boundary load, identify the strain concentration area of ​​the defect, and quantify the equivalent diameter of the defect;

[0146] Further, step S44 includes:

[0147] Step S441: Retrieve the full-field principal strain contour map at the moment when the load equals the elastic boundary load, and calculate the regional average strain and strain standard deviation of all principal strains of all pixels within the region of interest. Specifically, iterate through all valid pixels within the region of interest calculated by DIC, read the principal strain value corresponding to each pixel one by one, sum the principal strains of all pixels, and divide by the total number of valid pixels to obtain the regional average strain; calculate the variance based on the square of the difference between the principal strain of each pixel and the regional average strain, and then take the square root of the variance to obtain the strain standard deviation.

[0148] Step S442: Based on the regional average strain and strain standard deviation, identify defect strain concentration areas. For example, mark all pixel-connected regions where the principal strain value is greater than the sum of the regional average strain and 3 times the strain standard deviation as defect strain concentration areas. The strain distribution of homogeneous and defect-free UHPC solid core samples in the elastic stage follows a normal distribution. Areas exceeding 3 times the standard deviation are statistically significant anomalies, caused by local stiffness reduction due to defects such as internal pores, microcracks, and fiber clusters.

[0149] Step S443: For each defect strain concentration area, find the minimum bounding rectangle and use the length of the diagonal of the bounding rectangle as the equivalent diameter of the defect; if there are multiple defect strain concentration areas, take the maximum equivalent diameter as the equivalent diameter of the core sample; if there is no strain concentration area that meets the conditions, it is determined that the core sample has no obvious internal defects and the equivalent diameter of the defect is recorded as 0.

[0150] Step S45: Based on the failure mode type and defect equivalent diameter of the UHPC solid core sample, match the correction coefficient of the uncorrected strength value to generate the measured strength value of the UHPC solid core sample.

[0151] Specifically, the pre-established failure mode-strength correction coefficient database is queried, and the corresponding first strength correction coefficient is obtained according to the identified failure mode type; the second strength correction coefficient is calculated according to the identified defect equivalent diameter and the preset defect-strength influence relationship; the uncorrected strength value, the first strength correction coefficient and the second strength correction coefficient are multiplied to calculate the final measured strength value of the UHPC solid core sample.

[0152] In one specific implementation, the failure mode-strength correction coefficient database is obtained through regression analysis of a large amount of standard comparative test data. The first strength correction coefficient for fiber pull-out ductile failure is 1.0; the first strength correction coefficient for matrix crushing brittle failure is 0.92-0.98; the first strength correction coefficient for interface transition zone failure is 0.85-0.92; and the first strength correction coefficient for defect-induced failure is 0.75-0.88. The preset defect-strength influence relationship is as follows: when the equivalent diameter of the defect is less than 1 / 10 of the core sample diameter, the second strength correction coefficient is 1.0; when the equivalent diameter of the defect is greater than or equal to 1 / 10 of the core sample diameter but less than 1 / 5, the second strength correction coefficient is 0.9-0.95; and when the equivalent diameter of the defect is greater than or equal to 1 / 5 of the core sample diameter, the second strength correction coefficient is 0.8-0.89.

[0153] Specifically, step S4 is the multi-source data fusion and strength precision correction stage of the entire UHPC solid core sample strength measurement method based on multi-source data fusion. Its core design addresses key issues in existing technologies, such as relying solely on macroscopic load data to calculate strength, the inability to distinguish the impact of failure modes and internal defects on strength, large measurement errors, and the inability to explain the intrinsic causes of strength dispersion. It achieves a seamless transition from macroscopic strength measurement to microscopic mechanism analysis, significantly improving the accuracy and comprehensiveness of strength measurement. In the UHPC solid core sample testing scenario, traditional methods can only obtain a single strength value from the load-displacement curve, failing to identify differences in core sample failure modes or quantitatively assess the degree of strength reduction caused by internal defects. The actual mechanical properties of core samples corresponding to different failure modes differ significantly, with internal defects leading to substantial strength reduction. However, traditional methods do not consider these factors, resulting in large deviations between the measured results and the true strength of the core sample. Furthermore, traditional methods cannot explain the large strength dispersion of core samples within the same batch, failing to provide targeted guidance for construction quality control. Step S4, through macroscopic data preprocessing and initial correction in step S41, obtains the basic strength value that meets the standard requirements; through DIC microscopic analysis in step S42, it obtains strain evolution information of the entire core sample loading process and extracts characteristic parameters that can reflect the microscopic failure behavior of the material; through intelligent identification in step S43, it realizes automated and quantitative identification of failure modes and internal defects, eliminating the subjective error of manual judgment; through defect quantification analysis in step S44, it accurately identifies the defect strain concentration area in the elastic stage and quantifies the equivalent diameter of the defect, clarifying the degree of influence of internal defects on load-bearing performance, and providing a quantitative basis for the microscopic defect dimension for subsequent strength correction; through dual-coefficient strength correction in step S45, it establishes a quantitative mapping relationship between microscopic failure characteristics and macroscopic strength, reducing the error of strength measurement. Without step S4, only the uncorrected macroscopic strength value can be obtained, making it impossible to distinguish between normal and abnormal failure, assess the influence of internal defects, and resulting in poor accuracy of strength measurement results, and failing to reveal the intrinsic reasons for strength dispersion.

[0154] Step S5: Based on the measured strength values, failure mode types, and equivalent defect diameters of all UHPC solid core samples within the target area, perform statistical analysis to obtain representative values ​​of UHPC solid strength in the target area; generate and output a complete measurement report; store all data obtained from this measurement into the sample database for updating and optimizing the mechanical parameter prediction model and the mapping relationship library of the detection parameters.

[0155] Further, step S5 includes:

[0156] Step S51: Collect the identification codes, uncorrected intensity values, measured intensity values, damage mode types, defect equivalent diameters, mesoscopic feature parameters, and basic feature sets of all UHPC solid core samples under the same target area to construct a regional sample dataset;

[0157] Step S52: Perform hierarchical statistical analysis on the regional sample dataset, and calculate the representative value of the UHPC solid strength in the target area after removing abnormal failure samples. Specifically, this includes: screening out severe defect core samples whose failure mode is defect-induced failure and whose defect equivalent diameter reaches 1 / 5 or more of the core sample diameter. Such core samples cannot represent the uniform material properties of the structural body.

[0158] The arithmetic mean, standard deviation, and coefficient of variation of the measured strength values ​​of the remaining UHPC solid core samples are calculated. The average measured strength of the effective core samples is used as the representative value of the UHPC solid strength in the target area. The proportion of various failure modes and the distribution of the number of core samples in different defect size ranges are statistically analyzed to quantitatively reflect the overall casting, molding, and curing quality of the target area.

[0159] Step S53: Automatically generate a complete and standardized UHPC solid core sample strength test report; the test report includes the following contents: target area overview, original data collected from each UHPC solid core sample, non-destructive foundation feature detection results, adaptive loading and DIC acquisition parameter configuration records, load-displacement original curves, typical principal strain cloud maps for the entire field, summary of microscopic feature parameters, failure mode determination results, internal defect quantification results, calculation process of two-level strength correction coefficients, final measured strength of a single core sample, statistical analysis results of target area strength, representative value of area strength, and construction quality analysis suggestions corresponding to core sample defects and failure modes.

[0160] As one specific implementation method, the report outputs both an editable electronic document and a printed paper document. All original images, curves, and cloud maps are embedded as attachments in the electronic report, supporting traceability and verification.

[0161] Step S54: Upload the complete sample data of all core samples tested in this test to the global sample database to complete incremental storage; the complete sample data includes the core sample basic characteristics, geometric dimensions, prior mechanical prediction values, adaptive test parameters, macroscopic load data, displacement data, DIC microscopic parameters, failure mode, defect equivalent diameter, and final measured strength.

[0162] Step S55: Based on the newly added sample data, iteratively update the mechanical parameter prediction model and the mapping relationship library of detection parameters to achieve continuous self-optimization of the model and the mapping library;

[0163] The basic features and geometric dimensions of the newly added samples are used as new input samples, and the actual measured strength, measured initial crack load and measured elastic modulus are used as the true output labels. They are then incorporated into the original training dataset in step S211. The training subset and validation subset are re-divided, and the model training, hyperparameter tuning and accuracy verification processes are repeated to update and obtain a gradient boosting tree mechanical parameter prediction model with stronger generalization ability.

[0164] Based on the estimated ultimate compressive strength range of the newly added samples, the optimal matching loading rate and the combination of staged DIC acquisition frequency, the orthogonal test sample size of the corresponding mechanical performance range is supplemented, and the optimal combination of detection parameters for each range is solved again by response surface regression. After covering more working conditions, the parameter matching rules stored in the database are updated.

[0165] Specifically, step S5 involves data aggregation, regional evaluation, and closed-loop self-optimization. Steps S51 and S52 complete the stratified statistical analysis of multi-core samples in the region, calculate the representative strength value of the region after removing severely defective and abnormal samples, and simultaneously statistically analyze the distribution patterns of damage and defects, achieving a comprehensive evaluation from the strength value of a single core sample to the material performance of the entire structural region. Step S53 outputs a complete test report containing all the original data and mechanism analysis of the entire process. The entire process of calculation, judgment, and correction logic is traceable, meeting the compliance requirements of engineering testing. Steps S54 and S55 establish a closed-loop optimization mechanism for test data. Each batch of field test samples automatically expands the model training library and parameter mapping database, continuously reducing the prediction error of mechanical parameters and improving the rationality of adaptive DIC and loading parameter matching. As the number of test samples accumulates, the measurement accuracy and testing efficiency of the entire method can be continuously improved through iteration.

[0166] Example 2:

[0167] This embodiment introduces a UHPC solid core sample strength measurement system based on multi-source data fusion. The UHPC solid core sample strength measurement system based on multi-source data fusion is designed to match the above-mentioned UHPC solid core sample strength measurement method based on multi-source data fusion, and includes an information acquisition module, a parameter configuration module, a synchronous test acquisition module, a strength calculation module, and a performance statistics module.

[0168] Information Acquisition Module: This module establishes a unique identification code for all UHPC core samples drilled within the target detection area. It collects and stores a complete set of basic information for each UHPC core sample, including drilling location, design parameters, geometric dimensions, and end-face processing quality. Simultaneously, through a non-destructive testing system composed of an image acquisition unit and an ultrasonic testing unit, it acquires digital images of the end face of the UHPC core sample and internal longitudinal wave velocity data without damaging the structural integrity of the UHPC core sample. After image recognition and statistical analysis of steel fiber geometric parameters, it calculates the actual fiber volume content and fiber orientation distribution coefficient, integrating them to form a multi-dimensional basic feature set specific to each UHPC core sample. It also achieves the association and binding of basic information and basic feature sets, and traceable storage, providing real, reliable, and standardized raw input data for subsequent intelligent prediction of mechanical parameters.

[0169] The parameter configuration module retrieves the basic feature set and geometric dimension information of the UHPC solid core sample stored in the information acquisition module. This information is then imported as input features into a pre-trained mechanical parameter prediction model. Through the model's nonlinear mapping relationship, it outputs a set of prior mechanical parameters for the UHPC solid core sample, including the estimated ultimate compressive strength, estimated initial crack load, and estimated elastic modulus. Based on a pre-built detection parameter mapping relationship library, and according to the mechanical performance range of the UHPC solid core sample, it adaptively matches corresponding personalized detection parameters such as mechanical loading rate, staged image acquisition frequency, and camera magnification. This generates precise mechanical loading control commands and DIC acquisition control commands. Furthermore, a hardware synchronization trigger unit achieves high-precision time synchronization between the mechanical loading system and the DIC detection system, solving the problems of fixed parameters, poor adaptability, and asynchronous operation among multiple systems in traditional detection schemes.

[0170] The synchronous test acquisition module is used to complete the standardized preprocessing and test assembly work before the UHPC solid core sample is tested. It prepares a matte black and white speckle detection pattern on the side of the UHPC solid core sample that meets the recognition requirements of digital image related technologies. It accurately calibrates the installation centering accuracy of the UHPC solid core sample and the calibration error of the DIC detection system. It receives dual-channel control commands from the parameter configuration module and synchronously starts the mechanical loading test and microscopic image acquisition. According to the phased adaptive acquisition strategy, it continuously acquires load data, displacement data and surface sequence digital images in real time throughout the entire test process until the UHPC solid core sample is completely destroyed. It achieves microsecond-level time alignment between macroscopic mechanical data and microscopic deformation image data, providing high-quality and highly synchronous raw test data for subsequent multi-source data fusion analysis.

[0171] The strength calculation module performs preprocessing operations such as filtering and noise reduction, and outlier removal on the load and displacement data acquired by the synchronous test acquisition module. It then performs standardization corrections based on the geometry and end-face processing status of the UHPC solid core sample, calculating the uncorrected strength value of the UHPC solid core sample. Simultaneously, it performs digital image correlation analysis on the surface sequence digital images, generating full-field displacement and full-field principal strain cloud maps frame by frame. It quantitatively extracts a complete set of mesoscopic characteristic parameters, including the elastic stage mean principal strain, ultimate principal strain, strain concentration factor, and crack propagation parameters. Based on the elastic stage strain cloud map, it identifies and quantifies the equivalent diameter of internal defects in the UHPC solid core sample. Relying on a preset standard feature pattern, it intelligently determines the failure mode of the UHPC solid core sample. It matches the corresponding first strength correction coefficient according to the failure mode and calculates the corresponding second strength correction coefficient based on the equivalent diameter of the defect. Through dual-coefficient coupling correction, it obtains the final measured strength value of the UHPC solid core sample, achieving accurate strength correction based on the mesoscopic damage mechanism.

[0172] The performance statistics module is used to collect all data, including identity information, basic characteristics, test parameters, measured strength, failure mode, and defect parameters, of all UHPC core samples within the same target detection area. This constructs a complete regional sample dataset. Through hierarchical statistics and anomaly sample removal mechanisms, it calculates representative values ​​of UHPC strength in the target area. Simultaneously, it statistically analyzes the proportion of core sample failure modes and the distribution of defect sizes within the area, quantitatively evaluating the construction and maintenance quality of the region. It standardizes and generates a complete set of traceable and verifiable test reports. Simultaneously, it incrementally inputs the complete sample data from this test into the global sample database. Based on the newly added samples, it continuously iterates and optimizes the fitting accuracy of the mechanical parameter prediction model and the parameter matching rules of the detection parameter mapping relationship library, achieving a closed-loop self-optimization and upgrade of the entire detection method and system.

[0173] Working principle and its effects:

[0174] This invention relies on multi-source data fusion technology to conduct strength measurements based on the correlation between the microscopic real characteristics and macroscopic mechanical properties of UHPC solid core samples. Through a complete closed-loop mechanism of standardized information tracing, intelligent parameter adaptive matching, simultaneous acquisition of macroscopic and microscopic data, quantitative correction of microscopic damage, and regional statistical evaluation, it abandons the traditional single mode of fixed parameter testing and strength determination based solely on macroscopic loads. This eliminates the deviation between theoretical parameters and the actual performance of solid core samples, effectively improving the accuracy and scientific nature of UHPC solid core sample strength testing.

[0175] This invention first achieves end-to-end data traceability by uniquely encoding the UHPC core sample. Combined with non-destructive testing methods such as machine vision to identify the true distribution of steel fibers and ultrasonic testing to detect internal density, it collects the true microscopic features of the core sample, avoiding the data distortion problem caused by traditional reliance on design parameters. This provides reliable data support for subsequent mechanical prediction. Based on a mechanical parameter prediction model trained with massive samples and a mapping library of experimental parameters, it achieves adaptive matching of loading rates and DIC image acquisition parameters for core samples with different performance characteristics. High-precision synchronous control is used to complete the synchronization of mechanical load data and microscopic images. Synchronous data acquisition not only solves the problems of poor adaptability of traditional unified detection parameters, easy loss of key damage information or generation of redundant data, but also ensures the consistency of macro and micro data in the time dimension. On this basis, the full-field displacement and strain cloud map of the core sample is solved by DIC technology, which quantifies micro-damage characteristics, identifies failure modes and accurately calibrates the size of internal defects. Combined with the dual correction mechanism of failure mode and defect size, the uncorrected strength value after standardization is finely corrected, which effectively makes up for the inability of traditional detection to quantify the influence of micro-defects and failure modes on strength. The detection results have smaller errors and are closer to the actual stress state of the entity.

[0176] In summary, this invention achieves an upgrade in UHPC solid core sample strength testing from a single macroscopic calculation to microscopic mechanism correction and from single-core sample testing to regional overall quality evaluation by deeply integrating multi-dimensional microscopic features and macroscopic mechanical data. Simultaneously, relying on the continuous iterative optimization of the mechanical parameter prediction model and the experimental parameter mapping relationship library based on measured samples, the testing system possesses continuous self-optimization capabilities. This significantly improves the stability, versatility, and authority of UHPC solid structure strength testing under complex working conditions, accurately reflecting the true construction quality and load-bearing capacity of UHPC solid structures on-site, and meeting the application requirements for high-precision testing and evaluation of UHPC structures in major civil engineering projects.

[0177] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining the strength of UHPC solid core samples based on multi-source data fusion, characterized in that, include: All UHPC core samples obtained from drilling in the target area are identified and encoded, and non-destructive testing is performed to obtain the basic feature set of each UHPC core sample. The basic feature set and the geometric dimension information of the UHPC solid core sample are used as input features and input into a pre-trained mechanical parameter prediction model to predict the prior mechanical parameter set of the UHPC solid core sample, so as to match the loading parameters and DIC detection parameters corresponding to the UHPC solid core sample and generate mechanical loading control commands and DIC acquisition control commands; the mechanical parameter prediction model is used to establish a nonlinear mapping relationship between the input features and the prior mechanical parameters. A detection pattern is prepared on the surface of the UHPC solid core sample, and the load data, displacement data and surface sequence digital images of the UHPC solid core sample are collected and stored in real time based on the mechanical loading control command and the DIC acquisition control command. The load and displacement data are preprocessed and standardized to obtain the uncorrected strength value of the UHPC solid core sample. DIC analysis is performed on the surface sequence digital images to generate a full-field principal strain contour map and extract mesoscopic feature parameters. Based on the full-field principal strain contour map, the failure mode type and equivalent defect diameter of the UHPC solid core sample are automatically identified. A first strength correction coefficient is obtained according to the failure mode type, and a second strength correction coefficient is calculated based on the equivalent defect diameter. Combining the uncorrected strength value, the first strength correction coefficient, and the second strength correction coefficient, the measured strength value of the UHPC solid core sample is calculated. Based on the measured strength values, failure mode types, and equivalent defect diameters of all UHPC solid core samples within the target area, statistical analysis is performed to obtain representative values ​​of UHPC solid strength in the target area.

2. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 1, characterized in that, The steps for calculating the measured strength value of the UHPC solid core sample include: The load and displacement data are preprocessed and standardized to obtain the uncorrected strength values ​​of the UHPC solid core sample. DIC analysis was performed on the surface sequence digital images to generate the full-field displacement field and full-field principal strain cloud map at each moment during the entire loading process, and the microscopic feature parameters were extracted. Extract the mesoscopic characteristic parameters at the ultimate load moment, and identify the failure mode type of the UHPC solid core sample based on the standard characteristic patterns of failure mode types; Retrieve the full-field principal strain cloud map at the moment when the load equals the elastic boundary load, calculate the regional average strain and strain standard deviation of the principal strain of all pixels in the region of interest, so as to identify the defect strain concentration area and quantify the equivalent diameter of the defect. Based on the failure mode type and equivalent diameter of the defect in the UHPC solid core sample, correction coefficients are matched to the uncorrected strength value to generate the measured strength value of the UHPC solid core sample.

3. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 2, characterized in that, The DIC analysis of the surface sequence digital images generates the full-field displacement field and full-field principal strain contour map at each moment during the entire loading process, and extracts mesoscopic feature parameters, including: The region of interest for DIC calculation of the UHPC solid core sample is defined, and the surface sequence digital images are cropped and segmented to obtain each frame of the image of interest. Feature point tracking and matching are performed on adjacent frames of interest images to solve for the horizontal and vertical displacement components of each pixel, and the full-field displacement field corresponding to each frame of interest image is generated. Based on the full-field displacement field of each frame, differential solution is performed to calculate the principal strain value of each pixel, which is mapped to a gradient visualization color spectrum to generate a full-field principal strain cloud map. By traversing all frames of the full-field principal strain contour map and synchronously recorded load data, mesoscopic characteristic parameters are extracted.

4. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 3, characterized in that, The generation of the full-field displacement field corresponding to each frame of the image of interest includes: The image of interest in the current frame is divided into grids to obtain a computational subset of the image of interest in the current frame; Using the computational subsets of the previous frame of the image of interest as the matching benchmark, the gray-level similarity between each pair of computational subsets is calculated, and the normalized correlation coefficient between the two sets of computational subsets is obtained by zero-mean normalization. The normalized correlation coefficient is used as the matching discrimination index, and a global traversal search is performed to complete the feature tracking and matching of all computational subsets between adjacent frames of interest. Based on the coordinate offset of the matched subset, interpolation is used to calculate the horizontal and vertical displacement components of each pixel in the grid where the subset is located. All pixel displacement data are integrated to output the full-field displacement field corresponding to each frame of the image of interest.

5. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 2, characterized in that, The extraction of mesoscopic characteristic parameters at the ultimate load moment, and the identification of the failure mode type of the UHPC solid core sample based on the standard characteristic patterns of failure mode types, including: From detailed characteristic parameters, input indicators for determining the failure mode at the ultimate load moment are selected. Obtain the standard feature patterns corresponding to each pre-stored damage mode type, wherein the standard feature patterns include a combination of judgment input index intervals for the corresponding damage mode type; The mesoscopic characteristic parameters extracted from the current UHPC solid core sample at the ultimate load moment are compared one by one with the standard characteristic modes corresponding to various pre-stored failure mode types to determine the failure mode type of the current UHPC solid core sample.

6. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 1, characterized in that, The process involves identifying and encoding all UHPC core samples obtained from drilling in the target area, performing non-destructive testing, and acquiring a basic feature set for each UHPC core sample, including: Each UHPC entity core sample is identified and coded, and its basic information is bound to it; Non-destructive testing was performed on each UHPC solid core sample, including: Digital images of both ends of a UHPC solid core sample are acquired and input into a pre-trained first image recognition model to identify the contours of all steel fibers within the end faces and to statistically analyze the geometric parameters of the steel fibers. Based on the geometric parameters of the steel fibers, the actual fiber volume content and fiber orientation distribution coefficient of the UHPC solid core sample are calculated. The actual fiber volume content refers to the percentage of actual volume occupied by steel fibers in a unit volume UHPC solid core sample; the fiber orientation distribution coefficient is calculated by statistically analyzing the distribution variance of the projection angles of all steel fiber end faces. The longitudinal wave velocity of the UHPC solid core sample is measured, and combined with the actual fiber volume content and fiber orientation distribution coefficient, a set of basic features corresponding one-to-one with the identification code of the UHPC solid core sample is generated.

7. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 1, characterized in that, The generated mechanical loading control command and DIC acquisition control command include: The basic feature set and the geometric dimension information of the UHPC solid core sample are used as input features and fed into a pre-trained mechanical parameter prediction model. The output is the set of prior mechanical parameters corresponding to the UHPC solid core sample. The prior mechanical parameter set is input into the pre-built detection parameter mapping relationship library to match the loading parameters and DIC detection parameters corresponding to the UHPC solid core sample; Based on the loading parameters, mechanical loading control commands are generated, and based on the DIC detection parameters, DIC acquisition control commands are generated. The mechanical loading control commands and DIC acquisition control commands are then synchronized and started and run according to a preset time base.

8. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 7, characterized in that, The mechanical parameter prediction model is used to establish a nonlinear mapping relationship between input features and prior mechanical parameters. Its training and prediction process includes: Multiple sets of UHPC solid core samples with different input characteristics were obtained, and a standard uniaxial compressive strength test was performed on each UHPC solid core sample to obtain the corresponding measured mechanical parameters. The basic feature set, geometric dimension information and corresponding measured mechanical parameters of each UHPC solid core sample are associated and stored to form a training dataset; Initialize the mechanical parameter prediction model to fit the nonlinear relationship between the input features and the prior mechanical parameters; The training dataset is divided into a training subset and a validation subset. The initialized mechanical parameter prediction model is trained using the training subset, and the trained mechanical parameter prediction model is validated using the validation subset, thus obtaining the trained mechanical parameter prediction model. The basic feature set and geometric dimension information of the current UHPC solid core sample are input into the trained mechanical parameter prediction model, and the prior mechanical parameter set corresponding to the current UHPC solid core sample is output.

9. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 7, characterized in that, The detection parameter mapping database is used to store combinations of loading parameters and DIC detection parameters corresponding to different prior mechanical parameter ranges. The establishment and matching process of the detection parameter mapping database includes: Select the stratification factor parameters from the prior set of mechanical parameters and divide the stratification factor parameters into multiple mechanical property ranges; For each mechanical performance range, representative UHPC solid core samples within the corresponding mechanical performance range were selected for comparative tests. Each set of comparative tests used a different combination of test parameters to be optimized. Using the comprehensive evaluation value of detection accuracy and detection efficiency as the test evaluation index, a regression model between the combination of detection parameters to be optimized and the comprehensive evaluation value is established. The extreme value of the regression model is solved to obtain the combination of loading parameters and DIC detection parameters corresponding to each mechanical performance range. Each mechanical performance range and its corresponding loading parameters are associated and stored with the DIC test parameters to form a test parameter mapping relationship library; By comparing the stratification factor parameters corresponding to the prior mechanical parameter set of the current UHPC solid core sample with all mechanical property ranges in the detection parameter mapping relationship library, the mechanical property range to which the current UHPC solid core sample belongs is determined, and the loading parameters and DIC detection parameters corresponding to the current UHPC solid core sample are obtained.

10. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 1, characterized in that, The steps of real-time acquisition and storage of load data, displacement data, and surface sequence digital images of UHPC solid core samples include: A detection pattern suitable for DIC detection is prepared on the side of a UHPC solid core sample. The detection pattern is used to provide surface feature markings for the DIC detection system. The mechanical loading control command and the DIC acquisition control command are executed synchronously. According to the mechanical loading control command, the load data and displacement data of the UHPC solid core sample are acquired and stored in real time. According to the DIC acquisition control command, the surface sequence digital images of the UHPC solid core sample are acquired and stored in real time.

11. The method for determining the strength of UHPC solid core samples based on multi-source data fusion as described in claim 2, characterized in that, The preprocessing and standardization correction of the load and displacement data to obtain the uncorrected strength values ​​of the UHPC solid core sample includes: Based on the preprocessed load and displacement data, load-displacement curves are plotted, and the ultimate load value is determined by peak extraction method. Based on the geometric dimensions and end-face flatness of the UHPC solid core sample, the ultimate load value is standardized and corrected to obtain the uncorrected strength value of the UHPC solid core sample; the standardization correction includes aspect ratio correction and end-face flatness correction.

12. A UHPC solid core sample strength measurement system based on multi-source data fusion, used to implement the UHPC solid core sample strength measurement method based on multi-source data fusion as described in any one of claims 1-11, characterized in that, include: The information acquisition module is used to identify and encode all UHPC core samples drilled in the target area, and to perform non-destructive testing to obtain the basic feature set of each UHPC core sample. The parameter configuration module is used to take the basic feature set and the geometric dimension information of the UHPC solid core sample as input features, predict the prior mechanical parameter set of the UHPC solid core sample, match the loading parameters and DIC detection parameters corresponding to the UHPC solid core sample, and generate mechanical loading control commands and DIC acquisition control commands. The synchronous test acquisition module is used to prepare test patterns on the surface of the UHPC solid core sample, and based on the mechanical loading control command and the DIC acquisition control command, to acquire and store the load data, displacement data and surface sequence digital images of the UHPC solid core sample in real time. The strength calculation module is used to preprocess and standardize the load and displacement data to obtain the uncorrected strength value of the UHPC solid core sample, and to perform DIC analysis on the surface sequence digital image to generate a full-field principal strain cloud map and extract microscopic feature parameters to obtain the strength correction coefficient, and calculate the measured strength value of the UHPC solid core sample. The performance statistics module is used to perform statistical analysis based on the measured strength values, failure mode types, and equivalent defect diameters of all UHPC solid core samples within the target area to obtain representative values ​​of UHPC solid strength in the target area.

Images