A concrete strength prediction system and method based on SVR

Abstract

The present application relates to the technical field of concrete quality control, in particular to a concrete strength prediction system and method based on SVR; the present application collects multi-dimensional data of rebound test, compressive strength test and carbonation depth value determination through a data acquisition module, provides rich input features for the model, and enhances the prediction ability of the model; the Grubbs criterion is used to eliminate abnormal data, ensuring the quality and consistency of the input data and reducing the influence of noise on model training. The present application respectively establishes a concrete rebound curved surface regression equation and a support vector regression model, both of which serve as prediction criteria, and data fusion is carried out through a Kalman filter, further improving the prediction accuracy and reliability. Through the automatic and standardized data processing process, including data acquisition, preprocessing, regression prediction, SVR prediction and data fusion, the present application realizes the rapid prediction of concrete strength and improves the prediction efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of concrete quality control, in particular to a concrete strength prediction system and method based on SVR. BACKGROUND

[0002] Concrete quality control technology is the key link to ensure that the performance of concrete meets the design requirements and guarantees the quality and safety of the project, including raw material quality control, mix proportion design, production process control, curing environment control and post-detection. Concrete strength is the core content of concrete structure quality control, an important basis for structure design and construction, and one of the most important properties of concrete, so the innovation and application of its detection method are particularly important.

[0003] The traditional test method of concrete strength is to use the compressive strength test of standard cubic test blocks as the main basis for evaluating the strength of concrete. However, in actual engineering applications, the number of test blocks may be insufficient, the structure entity environment may be different from the storage environment of the test blocks, and the compressive strength test results of the standard cubic test blocks may not accurately reflect the concrete strength of the structure entity. Therefore, to avoid the above problems, the method of field detection is often used to detect the concrete strength in engineering. Common detection methods include local damage detection method and non-destructive detection method.

[0004] In addition, in the process of rapid development of modern science and technology, machine learning algorithms are becoming more mature and play an important role in actual engineering applications. Machine learning is a branch of artificial intelligence that uses computer programs to automatically learn patterns and rules from data and optimize the performance of prediction models to make accurate predictions or decisions on unknown data. The support vector machine regression algorithm (SVR) is a branch of SVM, which can effectively solve the nonlinear regression problem, map the low-dimensional nonlinear problem to high-dimensional linear, complete linear regression in high-dimensional space, and return the regression result to the space through inverse mapping until the regression result of the entire sample data set is completed. It can solve the problems of limited sample data and nonlinear data relationship, which is exactly the current situation of concrete compressive strength prediction. At the same time, SVR requires fewer parameter definitions, making modeling more convenient and reducing computing costs.

[0005] The existing test method of concrete strength has many shortcomings. First, the local damage detection method will cause certain damage to the structure during the detection process, which will cause certain damage to the building structure. In addition, after detection, the structure needs to be maintained, which increases the cost and is not conducive to large-scale detection of the structure.

[0006] Secondly, the detection accuracy of the commonly used non-destructive testing method is limited, and the results of some non-destructive testing methods need to be manually interpreted and analyzed by professional engineers, which may lead to subjectivity of the interpretation results and limit the interpretability of the data.

[0007] Finally, although the prior art has begun to try to establish an artificial intelligence-based concrete non-destructive testing method, there are still many use condition restrictions. The existing detection data set cannot cover all types of concrete and different working conditions, resulting in poor generalization ability of the model when facing new scenes and new samples, and prone to overfitting and underfitting problems.

[0008] In view of the above problems, it is necessary to propose a concrete strength prediction system and method based on SVR. SUMMARY

[0009] The purpose of the present application is to solve the problems in the background art and propose a concrete strength prediction system and method based on SVR.

[0010] The purpose of the present application can be achieved by the following technical solutions:

[0011] A concrete strength prediction system and method based on SVR.

[0012] In a first aspect, the present application provides a concrete strength prediction system based on SVR, comprising a data acquisition module, a data preprocessing module, a regression prediction module, an SVR prediction module and a data fusion module.

[0013] The data acquisition module collects the results of the rebound test, the compressive strength test and the carbonation depth value determination of the concrete test block.

[0014] Each concrete test block is numbered and subjected to a compressive strength test, the load at the time of failure is recorded, and the compressive strength is calculated.

[0015] Test points are arranged on each test block, rebound test and carbonation depth determination are performed, and the average rebound value and the average carbonation depth are calculated.

[0016] The data preprocessing module inputs all rebound values, compressive strength values and carbonation depth values into an outlier processing program, and removes abnormal data through the Grubbs criterion, including rebound value, compressive strength value and carbonation depth value.

[0017] The outlier processing results are renumbered and the aligned data are integrated to generate a data vector. Subsequently, the data vector is sent to the regression prediction module.

[0018] The data vector is subjected to Z-score standardization, including the standardization of rebound value, compressive strength, carbonation depth, age, and design strength grade. The standardization result is sent to the SVR prediction module.

[0019] The regression prediction module performs relationship fitting on the compressive strength, rebound value, and carbonation depth through a mathematical function according to the data vector.

[0020] A concrete rebound curved surface regression equation is established, and the optimal fitting parameter is solved through the least square method.

[0021] The fitting curved surface image is drawn, and the concrete rebound curved surface regression equation is saved as the first prediction criterion.

[0022] The SVR prediction module constructs a support vector regression model, including an input layer, a kernel function, and an output layer.

[0023] The hyperparameters are set, including a regularization parameter, an error tolerance, a kernel function parameter, an optimization objective function, and a constraint condition. The model parameter is solved through an SMO optimization algorithm, and the support vector regression model is trained as the second prediction criterion.

[0024] The data fusion module performs data fusion on the output results of the first prediction criterion and the second prediction criterion.

[0025] The two prediction results are weighted and fused through a Kalman filter to calculate an optimal estimate.

[0026] The filtering result is recorded as the final concrete strength prediction value distribution.

[0027] In a second aspect, the present application provides a concrete strength prediction method based on SVR, including the following steps:

[0028] Step one, material analysis and data collection;

[0029] The results of the rebound test, compressive strength test, and carbonation depth value determination of each concrete test piece i are collected, including the rebound value Ri, the compressive strength fi, and the carbonation depth hi.

[0030] Step two, data preprocessing and cleaning;

[0031] All the collected rebound values Ri, compressive strength fi, and carbonation depths hi are input into an abnormal value processing program, and abnormal data is removed through the Grubbs criterion to obtain an abnormal value processing result, i.e., a data vector (Rj, fj, hj).

[0032] Step three, data fitting and regression model establishment;

[0033] S301, regression model data fitting;

[0034] According to the data vector (Rj, fj, hj), the relationship fitting of the compressive strength, the rebound value and the carbonation depth is obtained by a mathematical function, to obtain a concrete rebound curved surface regression equation The parameters k, b and c in the equation.

[0035] The concrete rebound curved surface regression equation with known parameters is denoted as a first prediction criterion.

[0036] S302, a support vector regression model is constructed

[0037] The concrete age tj and the design strength grade Cj corresponding to the concrete block number symbol j in the data vector (Rj, fj, hj) are obtained, and a new data vector (tj, Cj, Rj, fj, hj) is constructed.

[0038] According to the new data vector (tj, Cj, Rj, fj, hj), the support vector regression model is trained.

[0039] The support vector regression model comprises an input layer, a kernel function and an output layer.

[0040] The input layer comprises four nodes for inputting Rj, hj, tj and Cj, respectively; the kernel function layer; the kernel function maps the input data to a high-dimensional space and gives a nonlinear relationship between the input data.

[0041] The output layer comprises one output node for outputting the predicted compressive strength fj.

[0042] The hyperparameters are set: the regularization parameter C, the error tolerance ε and the hyperparameter γ.

[0043] The insensitive loss coefficient of the support vector regression model, the kernel function, the optimization objective function and the constraint condition are set.

[0044] The new data vector (tj, Cj, Rj, fj, hj) is substituted, and the optimization objective function and the constraint condition are combined to solve the parameters of the model by the SMO optimization algorithm, and the solving result is recorded as the final parameters of the trained support vector regression model.

[0045] The trained support vector regression model is used as a second prediction criterion.

[0046] Step four, concrete strength prediction;

[0047] Take the concrete for predicting the compressive strength to prepare a plurality of concrete test blocks.

[0048] The basic data of the plurality of concrete test blocks are obtained, including the rebound value, the carbonation depth, the age and the design strength grade.

[0049] input the rebound value and carbonation depth of the several concrete test blocks into the first prediction criterion to obtain an output result of the first prediction criterion;

[0050] input the rebound value, carbonation depth, age and design strength grade of the several concrete test blocks into the second prediction criterion to obtain an output result of the second prediction criterion.

[0051] Step five, data fusion;

[0052] perform data fusion on the output result of the first prediction criterion and the output result of the second prediction criterion through a Kalman filter to obtain a final concrete strength prediction result.

[0053] Compared with the prior art, the present application has the following beneficial effects:

[0054] 1. The present application collects multi-dimensional data of rebound test, compressive strength test and carbonation depth value determination through the data acquisition module, provides rich input features for the model, enhances the prediction ability of the model, adopts Grubbs criterion to eliminate abnormal data, ensures the quality and consistency of the input data, and reduces the influence of noise on model training.

[0055] The present application respectively establishes a concrete rebound curved surface regression equation and a support vector regression model, both of which are used as prediction criteria, and data fusion is performed through a Kalman filter, further improving the prediction accuracy and reliability.

[0056] The present application realizes rapid prediction of concrete strength through an automatic and standardized data processing procedure including data acquisition, preprocessing, regression prediction, SVR prediction and data fusion, and improves the prediction efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0057] In order to facilitate understanding of those skilled in the art, the present application is further described below in combination with the drawings:

[0058] Figure 1 is a flow chart of the method of the present application;

[0059] Figure 2 is a schematic diagram of the test point arrangement of the concrete test block.

[0060] Figure 3 is a schematic diagram of the abnormal value processing program of the present application;

[0061] Figure 4 is a schematic diagram of the fitting surface of the concrete rebound curved surface regression equation;

[0062] Figure 5 is a schematic diagram of the structure topology of the support vector regression model;

[0063] Figure 6 A Kalman filtering result diagram of the present application;

[0064] Figure 7 A Kalman filtering result probability distribution diagram of the present application;

[0065] Figure 8 A method flowchart of the present application. DETAILED DESCRIPTION

[0066] The technical solutions of the present application will be described clearly and completely below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0067] Please refer to Figure 1 The concrete strength prediction system based on SVR shown in the figure comprises a data acquisition module, a data preprocessing module, a regression prediction module, an SVR prediction module and a data fusion module.

[0068] The data acquisition module collects the result data of the rebound test, the compressive strength test and the carbonation depth value determination of the concrete test blocks.

[0069] All the concrete test blocks are numbered, and the number is i, i=1, 2, 3,..., n; n is the total number of the concrete test blocks. The compressive strength test is performed on each concrete test block i, and the compressive strength fi of each concrete test block i is obtained.

[0070] The compressive strength test is specifically as follows:

[0071] The pressure testing machine uniformly loads along the central axis of each concrete test block i to the upper and lower compression surfaces at a preset loading speed until the test block is damaged. The load Fi of the concrete test block i at the time of damage is recorded, and the compressive strength fi of the concrete test block i is calculated by the formula . Wherein Ai is the compression area of the concrete test block i.

[0072] Please refer to Figure 2 The rebound test and the carbonation depth determination are performed on each test point, and the rebound value and the carbonation depth of each test point are obtained. The average value of the rebound values of all the test points is calculated to obtain the rebound value Ri of the entire concrete test block i; the average value of the carbonation depths of all the test points is calculated to obtain the carbonation depth hi of the entire concrete test block i.

[0073] The rebound test is specifically as follows: the rebound hammer is vertically placed on the test point to ensure that the probe of the instrument is in full contact with the surface. Then, the indenter of the rebound hammer is pressed with force to compress the spring inside and rebound, and the rebound value displayed by the rebound hammer is recorded.

[0074] The carbonation depth value determination is specifically as follows: after the rebound value test and the compressive strength test are completed, a hole with a preset depth and a preset diameter is drilled at each test point on the surface of the broken test block using a drilling tool. The debris and powder in the hole are cleaned, and the phenolphthalein alcohol solution is applied to the inner wall of the hole. When a clear color boundary line appears on the inner wall of the hole, the boundary line is determined to be the boundary line between the carbonation region and the non-carbonation region of the concrete. The carbonation depth is measured by a carbonation depth measuring instrument.

[0075] The data preprocessing module inputs all the rebound values Ri, the compressive strengths fi and the carbonation depths hi collected into an outlier processing program, and removes abnormal data by the Grubbs criterion.

[0076] Please refer to Figure 3 The outlier processing program is specifically as follows:

[0077] STEP1: Start the program;

[0078] Instruction: clear the program cache and running data in the previous round of operation;

[0079] STEP2: Rebound value data acquisition;

[0080] Instruction: call all rebound value data and record the data quantity n1;

[0081] STEP3: Compressive strength data acquisition;

[0082] Instruction: call all compressive strength data and record the data quantity n2;

[0083] STEP4: Carbonation depth data acquisition;

[0084] Instruction: call all carbonation depth data and record the data quantity n3;

[0085] STEP5: Data number distinction;

[0086] Instruction: use the number symbol Ri1 to number the rebound value data, i1∈(1, n1);

[0087] Use the number symbol fi2 to number the compressive strength data, i2∈(1, n2);

[0088] Use the number symbol hi3 to number the carbonation depth data, i3∈(1, n3);

[0089] STEP6: Resilience value statistical feature extraction;

[0090] Instruction: Calculate the average value of the resilience value data by the formula and the variance ;

[0091] Feedback: Jump to STEP9;

[0092] STEP7: Compressive strength statistical feature extraction;

[0093] Instruction: Calculate the average value of the compressive strength data by the formula and the variance ;

[0094] Feedback: Jump to STEP12;

[0095] STEP8: Carbonization depth statistical feature extraction;

[0096] Instruction: Calculate the average value of the carbonization depth data by the formula and the variance ;

[0097] Feedback: Jump to STEP15;

[0098] STEP9: Calculate the resilience value upper data amount;

[0099] Instruction: Calculate the Grubbs criterion outlier G(Ri1) of each resilience value Ri1 by the formula

[0100] Feedback1: If G(Ri1) is greater than the preset threshold Gmax, execute STEP10;

[0101] Feedback2: If G(Ri1) is not greater than the preset threshold Gmax, execute STEP18;

[0102] STEP10: Resilience value abnormal data marking;

[0103] Instruction: Mark the resilience value Ri1 corresponding to the Grubbs criterion outlier G(Ri1) greater than the preset threshold Gmax as resilience value abnormal data, and obtain the resilience value abnormal data amount k1;

[0104] STEP11: Resilience value abnormal data elimination

[0105] Instruction: Delete the resilience value abnormal data from all resilience values Ri1, and let n1=n1-k1;

[0106] Feedback: Return to STEP6: ​​​​

[0107] STEP 12: Calculate the amount of data on the upper side of the compressive strength;

[0108] Instructions: Calculate the Grubbs criterion outlier G (fi2) of each compressive strength fi2 by the formula

[0109] Feedback 1: If there is G (fi2) greater than the preset threshold Gmax, execute STEP 13;

[0110] Feedback 2: If there is no G (fi2) greater than the preset threshold Gmax, execute STEP 19;

[0111] STEP 13: Compressive strength abnormal data marking;

[0112] Instructions: Mark the compressive strength fi2 corresponding to the Grubbs criterion outlier G (fi2) greater than the preset threshold Gmax as compressive strength abnormal data, and obtain the number of compressive strength abnormal data k2;

[0113] STEP 14: Compressive strength abnormal data elimination

[0114] Instructions: Delete the compressive strength abnormal data from all compressive strength fi2, and let n2=n2-k2;

[0115] Feedback: Return to STEP 7:

[0116] STEP 15: Calculate the amount of data on the upper side of the carbonization depth;

[0117] Instructions: Calculate the Grubbs criterion outlier G (hi3) of each carbonization depth hi3 by the formula

[0118] Feedback 1: If there is G (hi3) greater than the preset threshold Gmax, execute STEP 16;

[0119] Feedback 2: If there is no G (hi3) greater than the preset threshold Gmax, execute STEP 20;

[0120] STEP 16: Carbonization depth abnormal data marking;

[0121] Instructions: Mark the carbonization depth hi3 corresponding to the Grubbs criterion outlier G (hi3) greater than the preset threshold Gmax as carbonization depth abnormal data, and obtain the number of carbonization depth abnormal data k3;

[0122] STEP 17: Carbonization depth abnormal data elimination

[0123] Instructions: Delete the carbonization depth abnormal data from all carbonization depth hi3, and let n3=n3-k3;

[0124] ​​Feedback: Return to STEP 8:

[0125] STEP 18: Output the result of the resilience processing;

[0126] Instruction: Generate end signal one;

[0127] Feedback: Record all remaining Ri1;

[0128] STEP 19: Output the result of the compressive strength processing;

[0129] Instruction: Generate end signal two;

[0130] Feedback: Record all remaining fi2;

[0131] STEP 20: Output the result of the carbonization depth processing;

[0132] Instruction: Generate end signal three;

[0133] Feedback: Record all remaining hi3;

[0134] STEP 21: Data integration and alignment;

[0135] Instruction: Identify the items in the resilience data Ri1, compressive strength data fi2, and carbonization depth data hi3 that have the same number i1, i2, and i3, and mark them as aligned data;

[0136] Feedback: Integrate the aligned data and assign a new number j to generate the data vector (Rj, fj, hj) = (Ri1, fi2, hi3); where i1 = i2 = i3; where j = 1, 2, 3,..., m; where m is the total number of data vectors obtained after integrating and aligning the data;

[0137] STEP 22: Output and end;

[0138] Instruction: Output all data vectors (Rj, fj, hj); where j = 1, 2, 3,..., m; output the average value of the resilience data in the last loop and variance ; output the average value of the compressive strength data in the last loop and variance ; output the average value of the carbonization depth data in the last loop and variance ;

[0139] Feedback: End all operation processes;

[0140] Copy all the output data vectors (Rj, fj, hj) into two copies, one of which is sent to the regression prediction module and the other is input to the normalization processing program.

[0141] The normalization processing procedure is specifically:

[0142] The mean value μ(R) and standard deviation σ(R) of Rj are called, the mean value μ(f) and standard deviation σ(f) of fj are called, and the mean value μ(h) and standard deviation σ(h) of hj are calculated;

[0143] The Z-score standardization is performed through the formula to convert Rj, fj and hj into preprocessed data Rj', fj' and hj' with a mean value of 0 and a standard deviation of 1.

[0144] Further, the age tj and design strength grade Cj of the concrete test block corresponding to each data vector (Rj, fj, hj) are collected.

[0145] And the Z-score standardization is performed through the formula to convert tj and Cj into preprocessed data tj' and Cj' with a mean value of 0 and a standard deviation of 1. Wherein is the mean value of the age, is the age variance; wherein is the mean value of the design strength grade, is the design strength grade variance.

[0146] All Rj', fj', hj', tj' and Cj' are combined into a new data vector (Rj', fj', hj', tj', Cj') and output to the SVR prediction module.

[0147] It should be noted that when using SVR support vector regression, since SVR is very sensitive to the scale of features, input data normalization is a key step to ensure effective training of the model. The purpose of standardization is to convert each feature to the same scale range, usually converting it to a distribution with a mean value of 0 and a variance of 1, or scaling it to a fixed range. After standardization, the influence of various input data on the model output result will be more balanced, thereby improving the effect of the model.

[0148] The regression prediction module performs relationship fitting on the compressive strength, rebound value and carbonation depth through mathematical functions according to the data vector (Rj, fj, hj).

[0149] The concrete rebound curved surface regression equation is established as follows:

[0150] Where k is the influence coefficient; b and c are the influence exponents; wherein f is the compressive strength, R is the rebound value, and h is the carbonation depth.

[0151] Substitute the experimental data, through the least square method to develop linear regression operation, get the influence coefficient k and the specific value of influence index b and c.

[0152] The specific operation process of the least square method is:

[0153] First, take the logarithmic form of the concrete rebound curve regression equation: ,

[0154] Subsequently, the concrete rebound curve regression equation in logarithmic form is changed by variable substitution, let =y', let lnk=a; let lnR=x, let lnh=z, get the concrete rebound curve regression equation after substitution: y'=bx+cz+a.

[0155] Subsequently, set the loss function , respectively, a, b and c partial derivative equal to 0, that is .

[0156] Expand the equation to get: ;

[0157] Substitute ;

[0158] get ;

[0159] Finally, substitute the data vector (Rj, fj, hj) to expand the operation, calculate the specific value of k, b and c.

[0160] For example, substitute the following data Rj, fj and hj:

[0161]

[0162] The best fitting parameters are obtained: k=0.0326; b=1.9143; c=-0.0902.

[0163] That is .

[0164] Please refer to Figure 4 , draw the fitting surface image of the concrete rebound curve regression equation .

[0165] Save the concrete rebound curve regression equation as the first prediction criterion.

[0166] The SVR prediction module constructs a support vector regression model and trains the model.

[0167] The support vector regression model is specifically:

[0168] Please refer to Figure 5 As shown in the figure, the support vector regression model includes an input layer, a kernel function, and an output layer.

[0169] The input layer includes 4 nodes for inputting Rj, hj, tj, and Cj, respectively; the kernel function layer; the kernel function in the kernel function layer maps the input data to a high-dimensional space, giving a non-linear relationship between the input data.

[0170] The output layer includes 1 output node for outputting the predicted compressive strength fj.

[0171] It should be noted that support vector regression is a regression method based on support vector machine algorithm, mainly used for processing regression problems. The basic idea of SVR is to map the original data from low-dimensional space to high-dimensional space through a mapping function, and then find the optimal hyperplane in the high-dimensional space to predict the output.

[0172] The specific process of the model training is as follows:

[0173] Set the hyperparameters: regularization parameter C, error tolerance ε, and hyperparameter γ.

[0174] It should be noted that the regularization parameter C controls the balance between model complexity and training error. A larger C value will make the model pay more attention to the training error, which may lead to overfitting. A smaller C value will make the model smoother, increasing the generalization ability; the error tolerance ε controls the smoothness of the regression function. A larger ε value means that larger errors can be tolerated, and a smaller ε value means that the regression model will strictly fit the training data; the hyperparameter γ is the training parameter of the kernel function, which controls the influence range of a single training sample. A larger γ value means a smaller influence range of a single sample, which may lead to overfitting; a smaller γ value will make the model smooth, which may lead to underfitting.

[0175] Set the insensitive loss coefficient Where fj is the actual value, i.e. the measured value of the compressive strength of concrete block j; is the predicted value, i.e. the actual output value of the support vector regression model; is the preset tolerance error threshold.

[0176] Set the kernel function: Where x and xw are the input values of the kernel function, including Rj, hj, tj, and Cj; where γ is the hyperparameter. Where w = 1, 2, 3,..., N. N is the kernel function node number.

[0177] Set the optimization objective function: Where w is the weight vector; is the regularization term of the weight vector; where is the penalty error term of the set optimization objective; where and is a slack variable, represents the error degree of the jth sample; wherein C is a regularization parameter.

[0178] The input data is mapped to a high bit space by a set kernel function K(xi, xj), and the mapping formula is: ; wherein and is a support vector; is a preset kernel function; and b is a bias coefficient.

[0179] The constraint condition is set as: .

[0180] According to the optimization objective function and the constraint condition, the parameters of the model are solved by the SMO optimization algorithm, and the solution result is recorded as the final parameter of the trained support vector regression model.

[0181] The trained support vector regression model is used as the second prediction criterion.

[0182] The data fusion module performs data fusion on the output results of the first prediction criterion and the second prediction criterion.

[0183] Please refer to Figure 6 and Figure 7 , the two prediction results are fused by a Kalman filter, and the first prediction criterion and the second prediction criterion are fused into a prediction result with a small mean and variance.

[0184] The specific process of data fusion is: obtaining the output result of the first prediction criterion: f1j={f11, f12,..., f1m}; obtaining the output result of the second prediction criterion: f2j={f21, f22,..., f2m}.

[0185] The mean μ(f1j) and μ(f2j) of f1j and f2j are calculated respectively; the variance and of f1j and f2j are calculated respectively.

[0186] The two data sets are weighted and fused by a Kalman filter, and the optimal estimate is calculated.

[0187] The Kalman filtering formula is: .

[0188] wherein K1 and K2 are both Kalman gain coefficients; wherein is the error estimation variance of the previous number j, and its update formula is .

[0189] record the filtering result , and determine the distribution of the predicted value of the final concrete strength.

[0190] Referring to Figure 8 , a concrete strength prediction method based on SVR includes the following steps:

[0191] Step 1, material analysis and data collection;

[0192] Collect the results of the rebound test, compressive strength test and carbonation depth value determination of each concrete test block i, including the rebound value Ri, the compressive strength fi and the carbonation depth hi.

[0193] Step 2, data preprocessing and cleaning;

[0194] Input all the collected rebound values Ri, compressive strengths fi and carbonation depths hi into the outlier processing program, and remove abnormal data through the Grubbs criterion to obtain the outlier processing result, that is, the data vector (Rj, fj, hj).

[0195] Step 3, data fitting and regression model establishment;

[0196] S301, regression model data fitting;

[0197] According to the data vector (Rj, fj, hj), the relationship between the compressive strength, the rebound value and the carbonation depth is fitted through a mathematical function, and the concrete rebound curved surface regression equation The parameters k, b and c in the equation are obtained.

[0198] The concrete rebound curved surface regression equation with known parameters is recorded as the first prediction criterion.

[0199] S302, construct a support vector regression model

[0200] Obtain the concrete age tj and design strength grade Cj corresponding to the concrete test block number symbol j in the data vector (Rj, fj, hj), and construct a new data vector (tj, Cj, Rj, fj, hj).

[0201] According to the new data vector (tj, Cj, Rj, fj, hj), the support vector regression model is trained.

[0202] The support vector regression model includes an input layer, a kernel function and an output layer.

[0203] The input layer includes 4 nodes for inputting Rj, hj, tj and Cj, respectively; a kernel function layer; the kernel function maps the input data to a high-dimensional space and gives a nonlinear relationship between the input data.

[0204] The output layer includes one output node for outputting the predicted compressive strength fj.

[0205] Set hyperparameters: regularization parameter C, error tolerance ε and hyperparameter γ.

[0206] Set the insensitive loss coefficient, kernel function, optimization objective function and constraint condition of the support vector regression model.

[0207] Substitute the new data vector (tj, Cj, Rj, fj, hj), and combine the optimization objective function and constraint condition to solve the model parameters by SMO optimization algorithm, and record the solution results as the final parameters of the trained support vector regression model.

[0208] The trained support vector regression model is used as the second prediction criterion.

[0209] Step four, concrete strength prediction;

[0210] Take the concrete used for predicting compressive strength to prepare several concrete test blocks.

[0211] Obtain the basic data of the several concrete test blocks, including rebound value, carbonation depth, age and design strength grade.

[0212] Input the rebound value and carbonation depth of the several concrete test blocks into the first prediction criterion to obtain the output result of the first prediction criterion;

[0213] Input the rebound value, carbonation depth, age and design strength grade of the several concrete test blocks into the second prediction criterion to obtain the output result of the second prediction criterion.

[0214] Step five, data fusion;

[0215] Data fusion is performed on the output result of the first prediction criterion and the output result of the second prediction criterion by a Kalman filter to obtain the final concrete strength prediction result.

[0216] It should be understood that the terms "include" and "contain" used in the specification and claims of the present disclosure indicate the presence of the described features, whole, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, whole, steps, operations, elements, components and / or sets thereof.

[0217] It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used in this disclosure and the claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "and / or," as used herein, refer to and set forth any and all possible combinations of one or more of the associated listed items, including permutations of such combinations;

[0218] The preferred embodiments of the application disclosed above are only to help explain the present application. The preferred embodiments do not describe all the details of the present application, nor limit the present application to only the specific embodiments. Obviously, many modifications and variations of the present application are possible in light of the above teachings. The embodiments were chosen and described in order to provide the best illustration of the principles of the present application and its practical application, and to thereby enable one of ordinary skill in the art to best utilize the present application. The present application is only limited by the claims and the full scope of equivalents thereof.

Claims

1. A concrete strength prediction system based on SVR, comprising a data preprocessing module, a regression prediction module, and an SVR prediction module, characterized in that: The data preprocessing module inputs all rebound values, compressive strength values, and carbonization depth values ​​into the outlier handling program, and removes outlier data using the Grubbs criterion, including outlier data removal for rebound values, compressive strength values, and carbonization depth values. The outlier handling results are renumbered and the data is integrated and aligned to generate a data vector; then, the data vector is sent to the regression prediction module. The data vector is Z-score normalized, including the normalization of springback value, compressive strength, carbonization depth, age, and design strength grade; the normalization result is sent to the SVR prediction module. The regression prediction module fits the relationship between compressive strength, rebound value and carbonization depth using mathematical functions based on the data vector. Establish the regression equation for the concrete rebound surface and solve the best fitting parameters using the least squares method; Plot the fitted surface image and save the concrete rebound surface regression equation as the first prediction criterion; The SVR prediction module constructs a support vector regression model, including an input layer, a kernel function, and an output layer; Set hyperparameters, including regularization parameters, error tolerance, kernel function parameters, optimization objective function, and constraints; solve for the model parameters using the SMO optimization algorithm, and train a support vector regression model as the second prediction criterion; It also includes a data acquisition module and a data fusion module; The data acquisition module collects the results of rebound tests, compressive strength tests, and carbonation depth measurements of concrete test blocks; Each concrete test block is numbered using the symbol i, and its compressive strength is tested. The load at failure is recorded, and the compressive strength fi2 is calculated. Test points were arranged on each test block to conduct rebound tests and carbonization depth measurements, and the rebound value Ri1 and carbonization depth hi3 were calculated. The data fusion module performs data fusion on the output results of the first prediction criterion and the second prediction criterion; The optimal estimate is calculated by weighting and fusing the two prediction results using a Kalman filter. Record the filtering results and output them as the final predicted value of concrete strength.

2. The concrete strength prediction system based on SVR according to claim 1, characterized in that, The outlier handling procedure is as follows: STEP 1: Start the program; Command: Clear the program cache and running data from the previous round of calculations; STEP2: Rebound value data capture; Command: Retrieve all rebound value data and record the data volume n1; STEP3: Capture compressive strength data; Command: Retrieve all compressive strength data and record the data volume n2; STEP4: Carbonization depth data capture; Command: Retrieve all carbonization depth data, record data volume n3; STEP5: Distinguish data numbers; Instruction: Use the numbering symbol Ri1 to number the rebound value data, i1∈(1,n1); The compressive strength data are numbered using the identifier fi2, where i2∈(1,n2); The carbonization depth data are numbered using the identifier hi3, where i3∈(1,n3). STEP 6: Extract statistical features of rebound value; Instructions: via formula Calculate the average value of the rebound data and standard deviation ; Feedback: Skip to STEP 9; STEP7: Extraction of statistical features of compressive strength; Instructions: via formula Calculate the average value of the compressive strength data and standard deviation ; Feedback: Skip to STEP12; STEP8: Extraction of statistical features of carbonization depth; Instructions: via formula Calculate the average value of carbonization depth data and standard deviation ; Feedback: Skip to STEP 15; STEP9: Calculate the amount of data above the rebound value; Instructions: via formula Calculate the Grubbs criterion outlier G(Ri1) for each rebound value Ri1. Feedback 1: There exists a value G(Ri1) greater than the preset threshold Gmax → Execute STEP10; Feedback 2: No G(Ri1) is greater than the preset threshold Gmax → Execute STEP18; STEP10: Mark abnormal rebound value data; Instruction: Mark the rebound value Ri1 corresponding to the Grubbs criterion outlier G(Ri1) that is greater than the preset threshold Gmax as rebound value outlier data, and obtain the number of rebound value outlier data k1; STEP11: Remove abnormal rebound value data; Instruction: Remove abnormal rebound values ​​from all rebound values ​​Ri1, and set n1 = n1 - k1; Feedback: Return to STEP 6: STEP12: Calculate the upper side data of the compressive strength; Instructions: via formula Calculate the Grubbs criterion outlier G(fi2) for each compressive strength fi2. Feedback 1: There exists G(fi2) greater than the preset threshold Gmax → Execute STEP13; Feedback 2: No G(fi2) is greater than the preset threshold Gmax → Execute STEP19; STEP13: Marking abnormal compressive strength data; Instruction: Mark the compressive strength fi2 corresponding to the Grubbs criterion outlier G(fi2) that is greater than the preset threshold Gmax as compressive strength outlier data, and obtain the number of compressive strength outlier data k2; STEP14: Remove abnormal compressive strength data; Instruction: Remove abnormal compressive strength data from all compressive strength fi2, and let n2 = n2 - k2; Feedback: Return to STEP7: STEP15: Calculate the amount of data above the carbonization depth; Instructions: via formula Calculate the Grubbs criterion outlier G(hi3) for each carbonization depth hi3. Feedback 1: There is a value G(hi3) greater than the preset threshold Gmax → Execute STEP16; Feedback 2: No G(hi3) is greater than the preset threshold Gmax → Execute STEP20; STEP16: Marking anomaly data in carbonization depth; Instruction: Mark the carbonization depth hi3 corresponding to the Grubbs criterion outlier G(hi3) that is greater than the preset threshold Gmax as carbonization depth outlier data, and obtain the number of carbonization depth outlier data k3; STEP17: Remove abnormal carbonization depth data; Instruction: Remove abnormal carbonization depth data from all carbonization depths hi3, and let n3 = n3 - k3; Feedback: Return to STEP8: STEP18: Output the springback value processing result; Command: Generate end signal one; Feedback: Record all remaining Ri1; STEP19: Output the compressive strength processing results; Command: Generate end signal two; Feedback: Record all remaining fi2; STEP20: Outputs the carbonization depth processing results; Command: Generate end signal three; Feedback: Record all remaining hi3s; STEP 21: Data Integration and Alignment; Instruction: Identify the items in the springback data Ri1, compressive strength data fi2, and carbonization depth data hi3 that have completely identical number symbols (i1, i2, and i3) and mark them as aligned data. Feedback: Integrate and align the data and assign a new identifier j to generate a data vector (Rj, fj, hj) = (Ri1, fi2, hi3); where i1 = i2 = i3; where j = 1, 2, 3, ..., m; where m is the total number of data vectors obtained after integrating and aligning the data. STEP22: Output and End; Command: Output all data vectors (Rj, fj, hj); where j = 1, 2, 3, ..., m; Output the average value of the rebound data in the last loop. and standard deviation ; Output the average compressive strength data in the last loop. and standard deviation Output the average carbonization depth data from the last loop. and standard deviation ; Feedback: All calculation processes have ended.

3. The concrete strength prediction system based on SVR according to claim 2, characterized in that, The specific process of Z-score standardization is as follows: Retrieve the mean μ(R) and standard deviation σ(R) of Rj; retrieve the mean μ(f) and standard deviation σ(f) of fj; calculate the mean μ(h) and standard deviation σ(h) of hj. Through formula Z-score standardization was performed to transform Rj, fj, and hj into preprocessed data Rj', fj', and hj' with a mean of 0 and a standard deviation of 1. Furthermore, the age tj and design strength grade Cj of the concrete test blocks corresponding to each data vector (Rj, fj, hj) are collected; And through the formula Z-score standardization was performed to transform tj and Cj into preprocessed data tj' and Cj' with a mean of 0 and a standard deviation of 1; in For age mean, is the standard deviation of age; where For the average design strength level, The standard deviation of the design strength grade; All Rj', fj', hj', tj', and Cj' are combined to form a new data vector (Rj', fj', hj', tj', Cj') and output to the SVR prediction module.

4. The concrete strength prediction system based on SVR according to claim 1, characterized in that, The specific process for establishing the regression equation for the concrete rebound surface is as follows: Let the power function be... Where k is the influence coefficient; b and c are the influence indices; f is the compressive strength, R is the resilience value; and h is the carbonization depth. Substitute the experimental data and perform linear regression using the least squares method to obtain the specific values ​​of the influence coefficient k and the influence indices b and c.

5. The concrete strength prediction system based on SVR according to claim 1, characterized in that, The support vector regression model is specifically as follows: It consists of an input layer, a kernel function, and an output layer. The input layer includes four nodes, which are used to input Rj, hj, tj, and Cj, respectively. Rj is Ri1, representing the rebound value; hj is hi3, representing the carbonation depth; tj is the age of the concrete specimen; and Cj is the design strength grade. The kernel function layer maps the input data to a high-dimensional space, assigning nonlinear relationships between the input data. The output layer includes one output node, which is used to output the predicted compressive strength fj, which is fi2.

6. The concrete strength prediction system based on SVR according to claim 5, characterized in that, The specific process of training a support vector regression model is as follows: Define the hyperparameters: regularization parameter C, error tolerance ε, and hyperparameter γ; Set insensitive loss coefficient Where fj is the actual value, that is, the measured value of the compressive strength of concrete test block j; These are the predicted values, i.e., the actual output values ​​of the support vector regression model; The preset tolerance error threshold; Define the kernel function: Where x and xw are the input values ​​of the kernel function, including Rj, hj, tj and Cj; where γ is a hyperparameter; where w = 1, 2, 3, ..., N; N is the kernel function node number; Define the objective function for optimization: Where w is the weight vector; It is the regularization term of the weight vector; where The penalty error term is used to set the optimization objective; where and As slack variables, This indicates the degree of error for the j-th sample; Where C is the regularization parameter; The input data is mapped to the higher-order space using a predefined kernel function K(x, xw). The mapping formula is as follows: ;in and These are support vectors; is the preset kernel function; b is the bias coefficient; Set constraints: ; Based on the objective function and constraints, the parameters of the model are solved using the SMO optimization algorithm, and the solution results are recorded. , which serve as the final parameters of the trained support vector regression model.

7. The concrete strength prediction system based on SVR according to claim 1, characterized in that, The specific procedures for compressive strength testing, rebound testing, and carbonization depth determination are as follows: The pressure testing machine applies a load uniformly to the upper and lower pressure surfaces along the central axis of each concrete specimen i at a preset loading rate until the specimen fails; the load Fi at the failure of concrete specimen i is recorded, and the result is obtained using the formula... Calculate the compressive strength fi of concrete specimen i; where Ai is the compressive area of ​​concrete specimen i; Several test points were arranged on each concrete test block i; Place the rebound hammer vertically at the test point, ensuring that the probe of the instrument is in full contact with the surface; then, press the pressure head of the rebound hammer firmly to compress the spring inside and cause it to rebound, and record the rebound value displayed by the rebound hammer; After completing the rebound value test and compressive strength test, a hole with a preset depth and preset diameter is drilled at each test point on the surface of the fractured specimen using a drilling tool; debris and powder in the hole are cleaned, and phenolphthalein alcohol solution is applied to the inner wall of the hole; when a clear color boundary line appears on the inner wall of the hole, the boundary line is determined to be the boundary line between the carbonized area and the non-carbonized area of ​​the concrete; the carbonization depth is measured using a carbonization depth measuring instrument.

8. The concrete strength prediction system based on SVR according to claim 1, characterized in that, The specific process of weighted fusion of the two prediction results using a Kalman filter is as follows: Obtain the output of the first prediction criterion: f1j={f11, f12, ..., f1m}; Obtain the output of the second prediction criterion: f2j={f21, f22, ..., f2m}; Calculate the mean μ(f1j) and μ(f2j) of f1j and f2j respectively; calculate the variance of f1j and f2j respectively. and ; The optimal estimate is calculated by weighting and fusing the two datasets using a Kalman filter. ; The Kalman filter formula is: ; Where K1 and K2 are both Kalman gain coefficients; The error estimate variance for the previous number j is updated using the following formula: ; Record the filtering results, i.e., the final output. = The predicted value of the final concrete strength is determined and output.

9. A method for predicting concrete strength based on SVR, characterized in that, Includes the following steps: Step 1: Material Analysis and Data Acquisition; Collect the results of rebound test, compressive strength test and carbonation depth determination of each concrete specimen i, including rebound value Ri, compressive strength fi and carbonation depth hi; Step 2: Data preprocessing and cleaning; All collected rebound values ​​Ri, compressive strength fi, and carbonization depth hi are input into the outlier processing program, and outlier data are removed using the Grubbs criterion to obtain the outlier processing result, i.e., the data vector (Rj, fj, hj). Step 3: Data fitting and regression model establishment; S301, Regression model data fitting; Based on the data vector (Rj, fj, hj), the relationship between compressive strength, rebound value, and carbonation depth is fitted using mathematical functions to obtain the concrete rebound surface regression equation. The parameters k, b, and c are used; where k is the influence coefficient, and b and c are the influence indices. Regression equation for concrete rebound surface with known parameters This is designated as the first prediction criterion; S302. Construct a support vector regression model; Obtain the concrete age tj and design strength grade Cj corresponding to the concrete test block number j in the data vector (Rj, fj, hj), and construct a new data vector (tj, Cj, Rj, fj, hj). Train a support vector regression model based on the new data vector (tj, Cj, Rj, fj, hj); The support vector regression model includes an input layer, a kernel function, and an output layer; The input layer consists of four nodes, which are used to input Rj, hj, tj and Cj respectively; the kernel function layer; the kernel function maps the input data to a high-dimensional space and assigns non-linear relationships between the input data; The output layer includes one output node, which is used to output the predicted compressive strength fj; Define the hyperparameters: regularization parameter C, error tolerance ε, and hyperparameter γ; Define the insensitive loss coefficient, kernel function, optimization objective function, and constraints for the support vector regression model; Substitute the new data vector (tj, Cj, Rj, fj, hj), and combine it with the optimization objective function and constraints to solve for the model parameters using the SMO optimization algorithm. Record the solution results as the final parameters of the trained support vector regression model. The trained support vector regression model is used as the second prediction criterion; Step 4: Concrete strength prediction; Several concrete test blocks were prepared from concrete used to predict compressive strength. Obtain basic data for the aforementioned concrete test blocks, including: rebound value, carbonation depth, age, and design strength grade; The rebound value and carbonation depth of the aforementioned concrete test blocks are input into the first prediction criterion to obtain the output result of the first prediction criterion. The rebound value, carbonation depth, age, and design strength grade of the aforementioned concrete test blocks are input into the second prediction criterion to obtain the output result of the second prediction criterion. Step 5: Predictive data fusion; The outputs of the first prediction criterion and the second prediction criterion are fused using a Kalman filter to obtain the final concrete strength prediction result.

Images