Non-destructive testing method and system for rebound of compressive strength of fast-hardening cement-based materials

Abstract

The present application belongs to the technical field of nondestructive testing of building materials, and particularly relates to a rebound nondestructive testing method and system for the compressive strength of rapidly hardened cement-based materials. The technical solution is as follows: sample data of rapidly hardened cement-based materials is collected to construct a sample data set; the sample data set is subjected to outlier rejection, normalization processing and feature weight mapping to form a multi-dimensional input feature vector; based on the multi-dimensional input feature vector, a compressive strength prediction model is constructed, and the model parameters are optimized through k-fold cross-validation; the rebound value, curing age and mix proportion parameters of the rapidly hardened cement-based material to be tested are input into the trained compressive strength prediction model to output the corresponding compressive strength detection value. The present application can comprehensively consider sample data, establish a nonlinear mapping relationship between multi-dimensional features and compressive strength, thereby realizing rapid, continuous and high-precision evaluation of the compressive strength of rapidly hardened cement-based materials without damaging the materials or structure.

Description

Technical Field

[0001] This invention belongs to the field of non-destructive testing technology for building materials, specifically relating to a rebound non-destructive testing method and system for the compressive strength of rapidly hardening cement-based materials. Background Technology

[0002] Rapid-hardening cement-based materials are characterized by rapid setting and hardening, rapid early strength development, and quick recovery of load-bearing capacity after construction. They are widely used in scenarios such as rapid repair of roads and bridges, local reinforcement of existing structures, emergency structural repairs, and post-disaster emergency engineering. These projects typically require high early-age compressive strength of the materials, necessitating a rapid assessment of whether the materials have reached the strength requirements for reopening to traffic, restoring use, or proceeding to the next construction stage. Therefore, establishing a rapid, continuous, and reliable compressive strength testing method suitable for rapid-hardening cement-based materials is of great significance for project quality control and construction schedule planning.

[0003] Currently, the compressive strength testing of cement-based materials mainly employs methods such as standard specimen compressive strength testing and core sampling. Standard specimen compressive strength testing can accurately obtain the material's compressive strength, but it requires pre-forming, curing, and destructive loading. The test results often lag behind the on-site construction process, making it difficult to meet the need for rapid early strength assessment of fast-curing cement-based materials. Core sampling testing can reflect the local strength of a solid structure, but the core sampling process can cause irreversible damage to the structure, and it suffers from limitations such as limited sample quantity, long testing cycles, and complex operating procedures, making it unsuitable for large-scale, continuous testing in emergency repair, reinforcement, and other emergency projects.

[0004] The rebound method is a commonly used non-destructive testing method for the compressive strength of concrete, offering advantages such as ease of operation, rapid testing speed, and minimal structural damage. Current rebound methods typically calculate compressive strength by measuring the rebound value of the material surface and combining this with empirical formulas or strength curves. However, compared to ordinary concrete, fast-curing cementitious materials exhibit more intense early hydration reactions and a faster strength increase rate. The relationship between rebound value and compressive strength is influenced by various factors, including curing age, mix proportions, admixture dosage, and material composition, exhibiting significant non-linear and time-varying characteristics. If a single rebound value or ordinary empirical formula is still used for strength conversion, it can easily lead to increased prediction errors, making it difficult to meet the accuracy and reliability requirements for on-site strength assessment of fast-curing cementitious materials.

[0005] With the development of data processing technology and machine learning methods, combining multi-source detection parameters with prediction models has provided a new approach to improve the accuracy of non-destructive testing of the compressive strength of cement-based materials. However, existing methods are mostly designed for ordinary concrete or materials of conventional age, and do not fully consider the impact of rapid strength growth in early-age cement-based materials, strong mix proportion sensitivity, and the dispersion of rebound data on prediction results. This results in insufficient applicability of these methods in rapid repair, reinforcement, and emergency engineering projects.

[0006] In summary, the existing technology has the following technical problems:

[0007] (i) Existing standard specimen compressive strength tests and core sampling tests are destructive or semi-destructive testing methods, which have problems such as long testing cycles, insufficient on-site continuous testing capabilities, and potential damage to structural integrity. They are difficult to meet the needs of rapid hardening cement-based materials for rapid assessment of early-age strength in structural repair, reinforcement, and emergency engineering.

[0008] (ii) Existing rebound methods usually estimate compressive strength based on a single rebound value or empirical strength curve, which is difficult to reflect the characteristics of rapid early strength growth, sensitivity to mix ratio changes and complex material composition of fast-hardening cement-based materials, resulting in insufficient accuracy of compressive strength prediction and engineering applicability.

[0009] (iii) Existing data-driven strength prediction methods are insufficient in integrating multiple factors such as rebound test data, curing age, mix proportion parameters and rebound dispersion, making it difficult to accurately establish the nonlinear mapping relationship between rebound value and compressive strength under the multi-factor coupling effect of rapid hardening cement-based materials. Summary of the Invention

[0010] This invention provides a rebound non-destructive testing method and system for the compressive strength of rapidly hardening cement-based materials. It can comprehensively consider factors such as rebound test data, curing age, mix proportion parameters and rebound dispersion, and establish a non-linear mapping relationship between multi-dimensional characteristics and compressive strength. Thus, it can achieve rapid, continuous and high-precision evaluation of the compressive strength of rapidly hardening cement-based materials without damaging the material or structure.

[0011] The technical solution of the present invention is as follows:

[0012] A rebound non-destructive testing method for the compressive strength of rapidly hardening cement-based materials includes the following steps:

[0013] S1. Obtain basic data of rapid-hardening cement-based material samples, including rebound test data, rebound dispersion index, curing age, mix proportion parameters, and corresponding measured compressive strength values ​​of rapid-hardening cement-based material samples under different mix proportion parameters and curing ages; construct a sample dataset;

[0014] S2. Preprocess the basic data, including outlier identification and removal, data normalization, and standardization of the mixing ratio parameters.

[0015] S3. Based on the preprocessed basic data, construct a multi-dimensional input feature vector, which includes rebound value, rebound dispersion index, maintenance age and mixing ratio parameter;

[0016] S4. Based on the multidimensional input feature vector and the corresponding measured compressive strength value, construct a compressive strength prediction model. The compressive strength prediction model is a random forest regression prediction model that introduces feature weighting, cascaded random forest and error feedback mechanism.

[0017] S5. The compressive strength prediction model is trained using the sample dataset, and the model parameters are optimized through k-fold cross-validation to obtain the trained compressive strength prediction model.

[0018] S6. Obtain the rebound test data, curing age information and mix proportion parameters of the rapid hardening cement-based material to be tested, and input them into the trained compressive strength prediction model to output the compressive strength test value of the rapid hardening cement-based material to be tested.

[0019] Furthermore, in the described method for non-destructive testing of the compressive strength of rapidly hardening cement-based materials, step S1, the acquisition of rebound test data includes the following process:

[0020] (1) Select a flat area on the surface of the rapid-hardening cement-based material sample as the rebound test surface;

[0021] (2) Calibrate and test the rebound hammer;

[0022] (3) Multiple rebound test points are arranged on the rebound test surface, and the net distance between adjacent rebound test points is not less than 20mm;

[0023] (4) Perform a rebound test with the rebound tester perpendicular to the rebound test surface to obtain the rebound value of each rebound test point;

[0024] (5) Statistical processing is performed on multiple rebound values ​​on the same test surface to obtain representative rebound values ​​and rebound dispersion index.

[0025] Furthermore, in the method for rebound non-destructive testing of the compressive strength of the rapidly hardening cement-based material, the mix proportion parameters include the amount of cementitious material, the amount of aggregate, the water-cement ratio, the amount of expanding agent and / or the amount of water-reducing agent.

[0026] Furthermore, in the method for non-destructive testing of the compressive strength of rapidly hardening cement-based materials, in step S2, outlier identification and removal adopts the Pauta criterion, box plot criterion, or statistical discrimination method based on standard deviation to identify and remove outlier data in the rebound test data and measured compressive strength values.

[0027] Furthermore, in the described method for non-destructive testing of the compressive strength of rapidly hardening cement-based materials, the multidimensional input feature vector in step S3 is represented as follows:

[0028]

[0029] Among them, X i Let R be the multidimensional input feature vector of the i-th sample. i Let be the representative rebound value of the i-th sample. Let t be the rebound dispersion index for the i-th sample. i Let i be the maintenance age of the i-th sample. This is the standardized and weighted parameter vector of the mix proportions.

[0030] The mix proportion parameter vector after standardization and weight mapping Obtained through the following formula:

[0031]

[0032]

[0033] Among them, P i Let C be the original combination ratio parameter vector for the i-th sample. i A represents the amount of cementitious material used. i For aggregate usage, W i For water-cement ratio, E i S represents the dosage of the expanding agent. i denoted as the water-reducing agent dosage, and W as the parameter weight matrix.

[0034] Furthermore, in the described method for rebound non-destructive testing of compressive strength of rapidly hardening cement-based materials, in step S4, the compressive strength prediction model includes a feature weighting module, a cascaded random forest module, an error feedback optimization module, and an integrated prediction module. The feature weighting module is used to assign weights to different features in the multi-dimensional input feature vector to form a weighted feature vector. The cascaded random forest module includes multiple random forest sub-models connected in a cascaded manner, and the input of the subsequent random forest sub-model includes the output result of the previous random forest sub-model. The error feedback optimization module is used to update the weights of the training samples or the weights of the random forest sub-models according to the error between the predicted value and the measured compressive strength value. The integrated prediction module is used to perform weighted integration of the output results of each random forest sub-model to obtain the predicted compressive strength value.

[0035] Furthermore, in the method for non-destructive testing of the compressive strength of rapidly hardening cement-based materials, in step S5, k-fold cross-validation involves dividing the sample dataset into k subsets, where k-1 subsets are used for model training and the remaining subset is used for model validation. The optimal parameter combination of the compressive strength prediction model is determined through multiple rounds of training and validation; where k is an integer from 5 to 10.

[0036] Furthermore, in the described method for rebound non-destructive testing of the compressive strength of rapidly hardening cement-based materials, in step S6, the predicted compressive strength value is obtained according to the following formula:

[0037]

[0038] in, This is the predicted compressive strength value. For the ensemble weights of the m-th level random forest submodel, This represents the prediction result output by the m-th level random forest sub-model, and each ensemble weight is determined based on the prediction error of the corresponding random forest sub-model.

[0039] A rebound non-destructive testing system for the compressive strength of rapidly hardening cement-based materials includes a calculation and processing unit, which is installed in a computer device, mobile terminal, tablet terminal, cloud server, or embedded processing device; used to implement the above-mentioned rebound non-destructive testing method for the compressive strength of rapidly hardening cement-based materials.

[0040] The computational processing unit includes a data acquisition module, a data processing module, a model building module, and a prediction calculation module. The data acquisition module acquires rebound test data, curing age information, mix proportion parameters, and corresponding measured compressive strength values ​​of the rapid-hardening cement-based material. The data processing module performs outlier identification and removal, normalization, and mix proportion parameter standardization on the rebound test data, curing age information, mix proportion parameters, and measured compressive strength values, and constructs a multi-dimensional input feature vector. The model building module constructs and trains a compressive strength prediction model based on the multi-dimensional input feature vector and the corresponding measured compressive strength values. The prediction calculation module inputs the rebound test data, curing age information, and mix proportion parameters of the rapid-hardening cement-based material to be tested into the trained compressive strength prediction model and outputs the measured compressive strength values.

[0041] Furthermore, in the described rebound non-destructive testing system for the compressive strength of rapidly hardening cement-based materials, the data acquisition module includes a rebound detection unit and a parameter acquisition unit. The rebound detection unit is used to acquire the rebound values ​​of multiple rebound test points on the surface of the rapidly hardening cement-based material specimen or the structure to be tested, and to calculate a representative rebound value and a rebound dispersion index based on the rebound values ​​of the multiple rebound test points. The parameter acquisition unit is used to acquire or input the curing age information and mix proportion parameters of the rapidly hardening cement-based material.

[0042] The model building module includes a feature weighting submodule, a cascaded random forest submodule, an error feedback optimization submodule, and an ensemble prediction submodule. The feature weighting submodule is used to perform weight mapping on the multi-dimensional input feature vector to obtain a weighted feature vector. The cascaded random forest submodule includes multiple random forest sub-models connected in a cascaded manner, used to output the compressive strength prediction result step by step based on the weighted feature vector. The error feedback optimization submodule is used to update the weights of the training samples or the weights of the random forest sub-models according to the error between the prediction result and the measured compressive strength value. The ensemble prediction submodule is used to perform weighted ensemble of the output results of multiple random forest sub-models to obtain the predicted compressive strength value.

[0043] The beneficial effects of this invention are as follows:

[0044] 1. Based on rebound test data, this invention enables compressive strength testing without damaging the material or structure under test. It can reduce the impact of standard specimen destructive testing or core drilling on testing efficiency and structural integrity, and is suitable for rapid on-site testing.

[0045] 2. This invention uses representative rebound value, rebound dispersion index, curing age and mix proportion parameters as input features, which can comprehensively reflect the rapid development of early strength of fast-hardening cement-based materials and the influence of multiple factors coupled on compressive strength. Compared with the single rebound parameter conversion method, it has better applicability.

[0046] 3. This invention reduces the adverse effects of test errors, differences in feature dimensions, and changes in the mixing ratio on the prediction results by removing outliers, normalizing the process, and mapping the weights of the mixing ratio parameters, thereby improving the stability of model training.

[0047] 4. This invention employs a random forest regression prediction model that incorporates feature weighting, cascaded random forests, and error feedback mechanisms. This model can establish a nonlinear mapping relationship between rebound test data, curing age, mix proportion parameters, and compressive strength, thereby improving the accuracy and reliability of compressive strength prediction for rapidly hardening cement-based materials.

[0048] 5. This invention can be deployed in computer equipment, mobile terminals, cloud servers or embedded processing devices, and can be used for on-site quality control and strength assessment during structural repair, structural reinforcement, rapid repair of roads and bridges and emergency engineering construction. Attached Figure Description

[0049] Figure 1 A flowchart illustrating the non-destructive testing method for the compressive strength rebound of rapidly hardening cement-based materials;

[0050] Figure 2 A schematic diagram of the structure of a non-destructive testing system for the compressive strength rebound of rapidly hardened cement-based materials;

[0051] Figure 3 This is a schematic diagram of the compressive strength prediction model structure based on feature weighting, cascaded random forest, and error feedback mechanism;

[0052] Figure 4 The results show the comparison between measured and predicted values ​​of compressive strength in the training set.

[0053] Figure 5 This is a comparison of the measured and predicted values ​​of the compressive strength of the test set. Detailed Implementation

[0054] Example 1

[0055] like Figure 1 As shown, the rebound non-destructive testing method for the compressive strength of rapidly hardening cement-based materials includes the following steps:

[0056] S1. Obtaining basic parameters;

[0057] The mix proportion parameters include the amount of cementitious materials, the amount of aggregates, the water-cement ratio, the amount of expanding agent, and the amount of water-reducing agent. Four rapid-hardening cement-based material specimens with different mix proportions were prepared, and their mix proportions are shown in Table 1.

[0058] Table 1. Mix proportions of rapidly hardening cement-based materials

[0059] Rebound tests were conducted on the rapid-curing cement-based material specimens at different curing ages. The curing ages could be determined based on the strength development characteristics of the rapid-curing cement-based material, and were selected as 1d, 1.5d, 2d, 2.5d, 3d, 4d, 4.5d, 6d, 7d, 11d, 14d, 21d, and 28d.

[0060] During the rebound test, the side of the specimen surface that is flat, dry, and without obvious defects is selected as the rebound test surface. Before the rebound test, the rebound hammer is calibrated and inspected using a standard steel anvil. Multiple rebound test points are arranged on the same rebound test surface, with a net distance of not less than 20 mm between adjacent rebound test points. The rebound test points avoid the edges, holes, cracks, and areas with obvious defects of the specimen. The rebound hammer is perpendicular to the rebound test surface to perform the rebound test, and the rebound value of each rebound test point is recorded.

[0061] Statistical processing was performed on multiple rebound values ​​obtained from the same test surface to obtain a representative rebound value R. i and rebound dispersion index Among them, the representative rebound value R i The rebound dispersion index is the average of multiple effective rebound values. The standard deviation or coefficient of variation of multiple effective rebound values;

[0062] After completing the rebound test, a compressive strength test was performed on the corresponding specimen to obtain the measured compressive strength value Y. i The compressive strength is calculated using the following formula:

[0063]

[0064] Among them, Y i Let F be the measured compressive strength of the i-th sample, F be the maximum load when the specimen fails, and A be the area of ​​the specimen subjected to pressure.

[0065] Through the rebound test and compressive strength test described above, the rebound value, rebound dispersion index, curing age, mix proportion parameters, and measured compressive strength values ​​under different mix proportions and curing ages were obtained, forming the original sample dataset. Table 2 shows some of the measured data on the rebound value and compressive strength of rapid-hardening cementitious materials with different mix proportions under different curing ages.

[0066] Table 2. Measured data of rebound value and compressive strength of rapidly hardening cement-based materials at different ages.

[0067] S2. Preprocess the basic data;

[0068] The original sample dataset is preprocessed; the preprocessing includes outlier identification and removal, data normalization, and mix proportion parameter standardization.

[0069] Outlier identification and removal can be done using the Pauta criterion. For rebound or compressive strength data under the same mix ratio and curing age, calculate the mean and standard deviation. When the deviation of a data point from the mean exceeds a preset multiple of the standard deviation, the data point is identified as an outlier and removed.

[0070] After removing outliers, the data is normalized to reduce the impact of differences in the units of different features on model training. Normalization can be performed using the min-max normalization method, expressed as:

[0071]

[0072] Where x is the original feature value, x min x is the minimum value of this feature. max The maximum value of this feature. These are the normalized eigenvalues.

[0073] S3. Construction of multi-dimensional input feature vectors;

[0074] Based on the preprocessed basic data, construct a multidimensional input feature vector X. i The matching ratio parameter vector P of the i-th sample i Represented as:

[0075]

[0076] Among them, C i A represents the amount of cementitious material used. i For aggregate usage, W i For water-cement ratio, E i S represents the dosage of the expanding agent. i This refers to the dosage of water-reducing agent;

[0077] For the mix proportion parameter vector P i Standardization and weight mapping are performed to obtain the weighted mix ratio parameter vector. :

[0078]

[0079] Where W is the parameter weight matrix;

[0080] Based on the representative rebound value R i Rebound dispersion index Maintenance period t i and weighted combination ratio parameter vector Construct a multidimensional input feature vector together:

[0081]

[0082] Among them, X i Let be the multidimensional input feature vector of the i-th sample;

[0083] By incorporating rebound test information, age information, and mix proportion information into the compressive strength prediction process, the model can comprehensively reflect the impact of rapid early-age strength growth and mix proportion changes on the compressive strength of fast-hardening cement-based materials.

[0084] S4. Construction of compressive strength prediction model;

[0085] like Figure 3 As shown, based on the multidimensional input feature vector X i and measured compressive strength Y i A compressive strength prediction model was constructed. The compressive strength prediction model is a random forest regression prediction model that incorporates feature weighting, cascaded random forest and error feedback mechanism.

[0086] The compressive strength prediction model includes a feature weighting module, a cascaded random forest module, an ensemble prediction module, and an error feedback optimization module;

[0087] The feature weighting module is used to weight the multidimensional input feature vector X. i We perform weight mapping on different features to obtain a weighted feature vector. : .in, Let be the weighted feature vector of the i-th sample, α be the feature weight vector, and ⊙ denote element-wise multiplication;

[0088] The cascaded random forest module consists of multiple random forest sub-models connected in a cascaded manner; the input of the m-th level random forest sub-model includes a weighted feature vector. and the predicted output of the previous level random forest submodel; the output of the m-th level random forest submodel is expressed as:

[0089]

[0090] in, f represents the prediction result output by the m-th level random forest submodel. m Let m be the nonlinear mapping function corresponding to the m-th level random forest submodel. This represents the prediction results from the previous level of the random forest sub-model. This means concatenating the weighted feature vector with the predicted output of the previous random forest sub-model.

[0091] The error feedback optimization module updates the training sample weights or random forest submodel weights based on the error between the prediction result and the measured compressive strength value; the prediction error is expressed as:

[0092]

[0093] Among them, e i Let Y be the prediction error for the i-th sample. i This is the measured value of compressive strength. This is the predicted compressive strength value;

[0094] The ensemble prediction module is used to perform weighted ensemble of the outputs of multiple random forest sub-models to obtain the final predicted compressive strength value:

[0095]

[0096] in, This is the predicted compressive strength value. For the ensemble weights of the m-th level random forest submodel, This represents the prediction result output by the m-th level random forest sub-model.

[0097] S5, Model Training and Cross-Validation;

[0098] The sample dataset is divided into a training set and a validation set. The training set is used to train the compressive strength prediction model. During the model training process, the model parameters are optimized by k-fold cross-validation. The sample dataset is divided into k subsets. Each time, k-1 subsets are selected as training data and the remaining subset is used as validation data. The training and validation are repeated k times to obtain the prediction performance of the model under different data partitioning conditions.

[0099] Model evaluation metrics may include root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²). 2 By minimizing RMSE and MAE and improving R 2 The optimal parameter combination of the compressive strength prediction model is determined, and the trained compressive strength prediction model is obtained.

[0100] S6. Prediction of compressive strength of the material to be tested;

[0101] During the on-site testing phase, the rebound test data, curing age information, and mix proportion parameters of the rapid-hardening cement-based material to be tested are obtained; the test data is preprocessed according to the methods described in S2 and S3 and a multi-dimensional input feature vector is constructed; the multi-dimensional input feature vector is input into the trained compressive strength prediction model, and the compressive strength test value of the rapid-hardening cement-based material to be tested is output.

[0102] To improve the stability of field applications, the prediction results are corrected based on the reasonable strength range of the rapid-hardening cement-based material. If the predicted value is lower than the preset minimum reasonable strength value, it is corrected to the minimum reasonable strength value; if the predicted value is higher than the preset maximum reasonable strength value, it is corrected to the maximum reasonable strength value. The corrected prediction results are expressed as follows:

[0103]

[0104] in, This is the corrected compressive strength test value. The predicted compressive strength output by the model. To preset the minimum reasonable strength value, This is the preset maximum reasonable strength value;

[0105] Finally, the corrected compressive strength test value is used as the on-site compressive strength assessment result of the rapid hardening cement-based material under the corresponding curing age; the result is used to determine whether the material meets the strength requirements for structural repair, reinforcement, rapid repair of roads and bridges or emergency projects to enter the next construction process, restore traffic or restore use.

[0106] To verify the prediction effect of the method of the present invention, the predicted compressive strength value output by the trained compressive strength prediction model is compared with the measured compressive strength value; for example... Figure 4 The image shows a comparison between the measured and predicted values ​​of the compressive strength in the training set; as shown... Figure 5 The figure shows the comparison between the measured and predicted compressive strength values ​​of the test set; the horizontal axis represents the measured compressive strength value, and the vertical axis represents the predicted compressive strength value; the scatter points represent the correspondence between the predicted and measured compressive strength values ​​of different samples, and the dashed line represents the ideal reference line where the predicted value equals the measured value; the closer the scatter points are to the ideal reference line, the closer the model prediction result is to the measured compressive strength value; if the predicted points are close to the y=x reference line, and the RMSE and MAE are small, R 2 If the value is ≥0.96, it indicates that the compressive strength prediction model can well reflect the strength development law of rapidly hardening cement-based materials under different curing ages and different mix proportions.

[0107] Compared with empirical conversion methods based solely on a single rebound value, this invention simultaneously introduces representative rebound values, rebound dispersion index, curing age, and mix proportion parameters. Furthermore, it establishes a nonlinear prediction relationship through feature weighting, cascaded random forests, and error feedback mechanisms. Therefore, it can improve the accuracy, stability, and engineering applicability of nondestructive testing of the compressive strength of rapidly hardening cement-based materials.

[0108] Example 2

[0109] like Figure 2 As shown, a rebound non-destructive testing system for the compressive strength of rapidly hardening cement-based materials includes a calculation and processing unit, which is installed in a computer device, mobile terminal, tablet terminal, cloud server, or embedded processing device; used to implement the above-mentioned rebound non-destructive testing method for the compressive strength of rapidly hardening cement-based materials.

[0110] The computational processing unit includes a data acquisition module, a data processing module, a model building module, and a prediction calculation module. The data acquisition module acquires rebound test data, curing age information, mix proportion parameters, and corresponding measured compressive strength values ​​of the rapid-hardening cement-based material. The data processing module performs outlier identification and removal, normalization, and mix proportion parameter standardization on the rebound test data, curing age information, mix proportion parameters, and measured compressive strength values, and constructs a multi-dimensional input feature vector. The model building module constructs and trains a compressive strength prediction model based on the multi-dimensional input feature vector and the corresponding measured compressive strength values. The prediction calculation module inputs the rebound test data, curing age information, and mix proportion parameters of the rapid-hardening cement-based material to be tested into the trained compressive strength prediction model and outputs the measured compressive strength values.

[0111] The data acquisition module includes a rebound detection unit and a parameter acquisition unit; the rebound detection unit is used to acquire the rebound values ​​of multiple rebound test points on the surface of the rapid-curing cement-based material specimen or the structure to be tested, and calculate the representative rebound value and rebound dispersion index based on the rebound values ​​of the multiple rebound test points; the parameter acquisition unit is used to acquire or input the curing age information and mix proportion parameters of the rapid-curing cement-based material.

[0112] The model building module includes a feature weighting submodule, a cascaded random forest submodule, an error feedback optimization submodule, and an ensemble prediction submodule. The feature weighting submodule is used to perform weight mapping on the multi-dimensional input feature vector to obtain a weighted feature vector. The cascaded random forest submodule includes multiple random forest sub-models connected in a cascaded manner, used to output the compressive strength prediction result step by step based on the weighted feature vector. The error feedback optimization submodule is used to update the weights of the training samples or the weights of the random forest sub-models according to the error between the prediction result and the measured compressive strength value. The ensemble prediction submodule is used to perform weighted ensemble of the output results of multiple random forest sub-models to obtain the predicted compressive strength value.

Claims

1. A rebound non-destructive testing method for the compressive strength of rapidly hardening cement-based materials, characterized in that, Includes the following steps: S1. Obtain basic data of rapid-hardening cement-based material samples, including rebound test data, rebound dispersion index, curing age, mix proportion parameters, and corresponding measured compressive strength values ​​of rapid-hardening cement-based material samples under different mix proportion parameters and curing ages; construct a sample dataset; S2. Preprocess the basic data, including outlier identification and removal, data normalization, and standardization of the mixing ratio parameters. S3. Based on the preprocessed basic data, construct a multi-dimensional input feature vector, which includes rebound value, rebound dispersion index, maintenance age and mixing ratio parameter; S4. Based on the multidimensional input feature vector and the corresponding measured compressive strength value, a compressive strength prediction model is constructed. The compressive strength prediction model is a random forest regression prediction model that introduces feature weighting, cascaded random forest, and error feedback mechanism. The compressive strength prediction model includes a feature weighting module, a cascaded random forest module, an error feedback optimization module, and an integrated prediction module. The feature weighting module is used to assign weights to different features in the multidimensional input feature vector to form a weighted feature vector. The cascaded random forest module includes multiple random forest sub-models connected in a cascaded manner. The input of the subsequent random forest sub-model includes the output result of the previous random forest sub-model. The error feedback optimization module is used to update the weights of the training samples or the weights of the random forest sub-models based on the error between the predicted value and the measured compressive strength; the ensemble prediction module is used to perform weighted ensemble of the outputs of each random forest sub-model to obtain the predicted compressive strength value. S5. The compressive strength prediction model is trained using the sample dataset, and the model parameters are optimized through k-fold cross-validation to obtain the trained compressive strength prediction model. S6. Obtain the rebound test data, curing age information and mix proportion parameters of the rapid hardening cement-based material to be tested, and input them into the trained compressive strength prediction model to output the compressive strength test value of the rapid hardening cement-based material to be tested.

2. The method for rebound non-destructive testing of the compressive strength of rapidly hardening cement-based materials according to claim 1, characterized in that, In step S1, the acquisition of rebound detection data includes the following process: (1) Select a flat area on the surface of the rapid-hardening cement-based material sample as the rebound test surface; (2) Calibrate and test the rebound hammer; (3) Multiple rebound test points are arranged on the rebound test surface, and the net distance between adjacent rebound test points is not less than 20mm; (4) Perform a rebound test with the rebound tester perpendicular to the rebound test surface to obtain the rebound value of each rebound test point; (5) Statistical processing is performed on multiple rebound values ​​on the same test surface to obtain representative rebound values ​​and rebound dispersion index.

3. The method for rebound non-destructive testing of the compressive strength of rapidly hardening cement-based materials according to claim 1, characterized in that, The mix proportion parameters include the amount of cementitious materials, the amount of aggregates, the water-cement ratio, the amount of expanding agent and / or the amount of water-reducing agent.

4. The method for rebound non-destructive testing of the compressive strength of rapidly hardening cement-based materials according to claim 1, characterized in that, In step S2, outlier identification and removal employs the Pauta criterion, box plot criterion, or statistical discrimination method based on standard deviation to identify and remove outlier data in the rebound test data and measured compressive strength values.

5. The method for rebound non-destructive testing of the compressive strength of rapidly hardening cement-based materials according to claim 3, characterized in that, In step S3, the multidimensional input feature vector is represented as follows: Among them, X i Let R be the multidimensional input feature vector of the i-th sample. i Let be the representative rebound value of the i-th sample. Let t be the rebound dispersion index for the i-th sample. i Let i be the maintenance age of the i-th sample. This is the standardized and weighted parameter vector of the mix proportions. The mix proportion parameter vector after standardization and weight mapping Obtained through the following formula: Wherein, P i is the original mix proportion parameter vector of the i th sample, C i is the cement content, A i is the aggregate content, W i is the water-cement ratio, E i is the expansion agent content, S i is the water-reducing agent content, and W is the parameter weight matrix.

6. The method for rebound non-destructive testing of the compressive strength of rapidly hardening cement-based materials according to claim 1, characterized in that, In step S5, k-fold cross-validation involves dividing the sample dataset into k subsets, where k-1 subsets are used for model training and the remaining subset is used for model validation. The optimal parameter combination of the compressive strength prediction model is determined through multiple rounds of training and validation, where k is an integer from 5 to 10.

7. The method for rebound non-destructive testing of the compressive strength of rapidly hardening cement-based materials according to claim 1, characterized in that, In step S6, the predicted compressive strength value is obtained according to the following formula: in, This is the predicted compressive strength value. For the ensemble weights of the m-th level random forest submodel, This represents the prediction result output by the m-th level random forest sub-model, and each ensemble weight is determined based on the prediction error of the corresponding random forest sub-model.

8. A rebound non-destructive testing system for the compressive strength of rapidly hardening cement-based materials, characterized in that, Includes a computing processing unit, which is disposed in a computer device, mobile terminal, tablet terminal, cloud server or embedded processing device; used to implement the rebound non-destructive testing method for the compressive strength of rapidly hardening cement-based materials as described in any one of claims 1-7; The computing and processing unit is equipped with a data acquisition module, a data processing module, a model building module, and a prediction calculation module. The data acquisition module is used to acquire rebound test data, curing age information, mix proportion parameters, and corresponding measured compressive strength values ​​of the rapid-hardening cement-based material. The data processing module is used to identify and remove outliers, normalize and standardize mix proportion parameters on the rebound test data, curing age information, mix proportion parameters, and measured compressive strength values, and to construct a multi-dimensional input feature vector. The model building module is used to build and train a compressive strength prediction model based on the multidimensional input feature vector and the corresponding measured compressive strength value; the prediction calculation module is used to input the rebound test data, curing age information and mix proportion parameters of the fast-hardening cement-based material to be tested into the trained compressive strength prediction model and output the compressive strength test value.

9. The rebound non-destructive testing system for the compressive strength of rapidly hardening cement-based materials according to claim 8, characterized in that, The data acquisition module includes a rebound detection unit and a parameter acquisition unit; the rebound detection unit is used to acquire the rebound values ​​of multiple rebound test points on the surface of the rapid-curing cement-based material specimen or the structure to be tested, and calculate the representative rebound value and rebound dispersion index based on the rebound values ​​of the multiple rebound test points; the parameter acquisition unit is used to acquire or input the curing age information and mix proportion parameters of the rapid-curing cement-based material. The model building module includes a feature weighting submodule, a cascaded random forest submodule, an error feedback optimization submodule, and an ensemble prediction submodule. The feature weighting submodule is used to perform weight mapping on the multi-dimensional input feature vector to obtain a weighted feature vector. The cascaded random forest submodule includes multiple random forest sub-models connected in a cascaded manner, used to output the compressive strength prediction result step by step based on the weighted feature vector. The error feedback optimization submodule is used to update the weights of the training samples or the weights of the random forest sub-models according to the error between the prediction result and the measured compressive strength value. The ensemble prediction submodule is used to perform weighted ensemble of the output results of multiple random forest sub-models to obtain the predicted compressive strength value.

Images