CN117153297A - Cement concrete compressive strength detection method, system and electronic equipment - Google Patents
Cement concrete compressive strength detection method, system and electronic equipment Download PDFInfo
- Publication number
- CN117153297A CN117153297A CN202310293500.8A CN202310293500A CN117153297A CN 117153297 A CN117153297 A CN 117153297A CN 202310293500 A CN202310293500 A CN 202310293500A CN 117153297 A CN117153297 A CN 117153297A
- Authority
- CN
- China
- Prior art keywords
- sample
- compressive strength
- cement concrete
- feature
- base learner
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000004568 cement Substances 0.000 title claims abstract description 98
- 238000001514 detection method Methods 0.000 title claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 94
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 34
- 238000000034 method Methods 0.000 claims abstract description 31
- 230000002159 abnormal effect Effects 0.000 claims abstract description 17
- 238000005070 sampling Methods 0.000 claims abstract description 17
- 239000011159 matrix material Substances 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 8
- 239000010881 fly ash Substances 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 239000003638 chemical reducing agent Substances 0.000 claims description 4
- 239000002893 slag Substances 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 4
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000003066 decision tree Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 2
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000000053 physical method Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000008030 superplasticizer Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02W—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO WASTEWATER TREATMENT OR WASTE MANAGEMENT
- Y02W30/00—Technologies for solid waste management
- Y02W30/50—Reuse, recycling or recovery technologies
- Y02W30/91—Use of waste materials as fillers for mortars or concrete
Landscapes
- Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Crystallography & Structural Chemistry (AREA)
- Chemical & Material Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)
Abstract
The application provides a method, a system and electronic equipment for detecting compressive strength of cement concrete, comprising the following steps: constructing a compressive strength training sample set of cement concrete, removing abnormal compressive strength samples based on an improved random sampling consistency algorithm, and obtaining an updated compressive strength training sample set; respectively inputting the original features selected based on each sliding window into a first, a second and a third base learners for model training to respectively obtain a first, a second and a third output features; inputting the first, second and third output characteristics and the characteristic set reconstructed by the original characteristic set into a fourth base learner for model training, obtaining model parameters of each base learner, and constructing by the first, second, third and fourth base learners to obtain a cement concrete compressive strength model; and inputting the cement concrete sample to be tested into a cement concrete compressive strength model to obtain the compressive strength value of the sample to be tested. The application improves the accuracy of the detection result of the compressive strength of the cement concrete.
Description
Technical Field
The application relates to the technical field of concrete detection, in particular to a method and a system for detecting compressive strength of cement concrete and electronic equipment.
Background
In the construction industry, cement concrete is one of the most important structural materials, and the compressive strength of cement concrete is a key technical parameter because the compressive strength of cement concrete largely determines the load and safety performance of a building, thereby affecting the quality of the building. For the measurement of the compressive strength of cement concrete, the traditional method is to manufacture a standard component according to the component proportion of the cement concrete, maintain for a period of time (about 28 days) under standard conditions, and finally perform physical measurement of the compressive strength. The method has the advantages of complex process, huge time consumption, strict maintenance and measurement environment, uneconomical, time-consuming and low-efficiency, and tired engineering progress.
With the rise of machine learning technology and its good performance, machine learning technology has been used for cement concrete detection. In the related art, although the method for measuring the compressive strength of the cement concrete can be used for rapidly detecting the compressive strength of the cement concrete, the influence factors existing in the process of detecting the compressive strength of the cement concrete are more, and the requirement of detection precision cannot be met to a certain extent; in addition, model overfitting is also prone to occur.
Disclosure of Invention
In view of the above, the application provides a method, a system and electronic equipment for detecting the compressive strength of cement concrete, which can improve the accuracy of the detection result of the compressive strength of cement concrete and reduce the false alarm rate of the detection result.
In order to achieve the above object, the present application provides a method for detecting compressive strength of cement concrete, comprising the steps of:
s1, setting proportioning parameters and external influence parameters of each cement concrete sample, correspondingly measuring to obtain compressive strength of each cement concrete sample, and constructing a compressive strength training sample set from the proportioning parameters, the external influence parameters and the measured compressive strength data of each cement concrete sample;
s2, based on an improved random sampling consistency algorithm, abnormal compressive strength samples are removed from the compressive strength training sample set, and an updated compressive strength training sample set is obtained;
s3, acquiring an original feature set of the updated compressive strength training sample set, carrying out feature selection on the original feature set in a sliding window mode, and correspondingly inputting the original feature selected based on each sliding into a first base learner, a second base learner and a third base learner to carry out model training to obtain a first output feature, a second output feature and a third output feature respectively;
s4, forming a reconstructed feature set by the first output feature, the second output feature, the third output feature and the original feature set, inputting the reconstructed feature set into a fourth base learner for model training, and integrating the first base learner, the second base learner, the third base learner and the fourth base learner to obtain a trained cement concrete compressive strength model;
s5, inputting the cement concrete sample to be tested into a cement concrete compressive strength model to obtain the compressive strength value of the cement concrete sample to be tested.
Further, step S1 includes:
the proportioning parameters at least comprise one of the following: cement content, blast furnace slag content, fly ash content, water reducing agent, coarse aggregate and fine aggregate;
the external influence parameters at least comprise one of age, temperature and humidity.
Further, step S2 includes:
s201, acquiring a first point cloud image corresponding to an updated compressive strength training sample set, randomly selecting a first sample and a second sample from the first point cloud image, calculating the Euclidean distance between the first sample and the second sample to obtain a midpoint coordinate between the first sample and the second sample, marking all samples in the first point cloud image, from which the first sample and the second sample are removed, as second point cloud images, and respectively calculating the first Euclidean distance between the midpoint and each sample in the second point cloud image;
s202, determining a sample with the largest first Euclidean distance as a third sample, calculating to obtain the inner coordinates of a triangle formed by the first sample, the second sample and the third sample, marking all samples of the second point cloud picture except the third sample as third point cloud pictures, respectively calculating the second Euclidean distance between the inner and each sample of the third point cloud pictures,
s203, determining a sample with the largest second Euclidean distance as a fourth sample, and marking all samples in the third point cloud image except the fourth sample as fourth point cloud images;
s204, calculating transformation matrixes of the four samples according to the determined first sample, second sample, third sample and fourth sample, and obtaining a rejected sample model M.
Further, the construction of the point cloud image in the step S2 includes:
taking the proportioning parameter, the external influence parameter and the compressive strength of each cement concrete sample as input to form n-dimensional data corresponding to each sample, wherein the number of samples is m;
forming n rows and m columns of matrix X from multi-dimensional data of m samples, carrying out zero-mean on each row of the matrix X, and calculating to obtain a covariance matrix of the matrix X;
calculating eigenvalues and corresponding eigenvectors of the covariance matrix, sequentially sequencing the eigenvectors according to the order of the eigenvalues from large to small, and selecting the eigenvectors of the first 3 rows to form a matrix P of 3 rows and m columns;
and constructing a matrix P to obtain a first point cloud image.
Further, step S2 further includes:
s205, defining a training sample set I best Training sample set I best Initial value of the set I is an empty set best The number of samples in (a) is N 0 Setting the maximum iteration number K of the algorithm max ;
S206, setting an error threshold lambda of the sample model M to be removed, respectively calculating a projection error d between each sample in the fourth point cloud picture and the sample model M to be removed, if the projection error d is smaller than the error threshold lambda, the sample is a qualified sample, otherwise, the sample is an abnormal sample, and the sample point is removed;
s207, forming a set I by all qualified samples 1 Statistics of the set I 1 Number of samples N 1 If N 1 Greater than N0, set I best Update to set I 1 Set I best The number of samples is updated to N1;
s208, repeatedly executing the steps S201-S204, and updating the iteration number k of each execution step;
s209, when the iteration number reaches K max And when the training sample set obtained by the iteration is an updated compressive strength training sample set.
Further, step S208 includes:
the calculation formula of the iteration number k is as follows:
p is the confidence, 0.995 is taken, w is the ratio of the number of samples in the set acquired at each iteration to the total number of samples, and m is taken to be 4.
Further, step S3 includes:
selecting features from the original feature set by using a first window size, wherein the selected features form first features, inputting the first features into a first base learner for model training, and obtaining first output features;
selecting features from the original feature set by using a second window size, forming second features by the selected features, inputting the second features into a second base learner for model training, and obtaining second output features;
selecting features from the original feature set by a third window size, forming a third feature by the selected features, inputting the third feature into a third base learner for model training, and obtaining a third output feature;
wherein the first feature, the second feature, and the third feature are all different.
Further, the method further comprises: the first base learner, the second base learner, the third base learner, and the fourth base learner use the same or different algorithms.
The application also provides a system for detecting the compressive strength of cement concrete, which comprises:
the sample acquisition module is used for setting different proportioning parameters and external influence parameters of each compressive strength training sample, correspondingly measuring the compressive strength of each compressive strength training sample, acquiring the compressive strength of each compressive strength training sample, and constructing a compressive strength training sample set of the cement concrete;
the sample removing module is used for removing abnormal samples in the compressive strength training sample set to obtain an updated compressive strength training sample set;
the basic learner module acquires an original feature set of the updated compressive strength training sample set, performs feature selection on the original feature set in a sliding window mode, and correspondingly inputs the original feature selected based on each sliding into a first basic learner, a second basic learner and a third basic learner for model training to obtain a first output feature, a second output feature and a third output feature respectively;
the compressive strength model module is used for forming a reconstructed feature set by the first output feature, the second output feature, the third output feature and the original feature set, inputting the reconstructed feature set into a fourth base learner for model training, and integrating the first base learner, the second base learner, the third base learner and the fourth base learner to obtain a trained cement concrete compressive strength model;
and the detection module inputs the cement concrete sample to be detected into the cement concrete compressive strength model to obtain the compressive strength value of the cement concrete sample to be detected.
The present application also provides an electronic device including:
a processor;
a memory having executable code stored therein that, when executed, causes the processor to perform one or more of the methods described above.
According to the application, abnormal samples are removed through an improved random sampling consistency algorithm, so that the problem of poor training effect of a later model caused by the abnormal samples when the samples are collected is solved, the problem that a large number of iterations are required by the random sampling consistency algorithm is also solved, and the detection accuracy of the compressive strength of cement concrete is improved; combining the original features with the first-stage output of the integrated learning, and simultaneously taking the original features as the second-stage input of the integrated learning, so as to avoid the phenomenon of fitting after the original features directly pass through the two layers of integrated learning structures; the original characteristics of the compressive strength samples are selected by a sliding sampling method, so that the input of each algorithm is ensured to have specificity, the formed output characteristics have diversity, and the overfitting phenomenon of integrated learning is further avoided; the detection of the compressive strength of the cement concrete is simpler, and the measurement time of the compressive strength of the cement concrete is effectively shortened.
Drawings
FIG. 1 is a flow chart of a method for detecting compressive strength of cement concrete according to an embodiment of the application;
FIG. 2 is a flow chart of a method for testing the compressive strength of cement concrete according to an embodiment of the application;
FIG. 3 is a schematic view of culling abnormal samples according to an embodiment of the present application;
FIG. 4 is a diagram of a base learner training a slide selection feature, according to an embodiment of the present application;
fig. 5 is a system schematic diagram of a cement concrete compressive strength detection system according to one embodiment of the application.
Detailed Description
The present application will be described in detail below with reference to the specific embodiments shown in the drawings, but these embodiments are not limited to the present application, and structural, method, or functional modifications made by those skilled in the art based on these embodiments are included in the scope of the present application.
In one embodiment of the present application as shown in fig. 1, the present application provides a method for detecting compressive strength of cement concrete, the method comprising;
s1, setting proportioning parameters and external influences of each cement concrete sample, correspondingly measuring compressive strength of each cement concrete sample, and constructing a compressive strength training sample set from the proportioning parameters, the external influence parameters and the measured compressive strength data of each cement concrete sample;
s2, based on an improved random sampling consistency algorithm, abnormal compressive strength samples are removed from the compressive strength training sample set, and an updated compressive strength training sample set is obtained;
s3, acquiring an original feature set of the updated compressive strength training sample set, carrying out feature selection on the original feature set in a sliding window mode, and correspondingly inputting the original feature selected based on each sliding into a first base learner, a second base learner and a third base learner to carry out model training to obtain a first output feature, a second output feature and a third output feature respectively;
s4, forming a reconstructed feature set by the first output feature, the second output feature, the third output feature and the original feature set, inputting the reconstructed feature set into a fourth base learner for model training, and integrating the first base learner, the second base learner, the third base learner and the fourth base learner to obtain a trained cement concrete compressive strength model;
s5, inputting the cement concrete sample to be tested into a cement concrete compressive strength model to obtain the compressive strength value of the cement concrete sample to be tested.
According to the application, through an improved random sampling consistency algorithm, abnormal compressive strength samples are removed from the compressive strength training sample set, so that an updated compressive strength training sample set is obtained, and the balance of sample data is ensured. The method comprises the steps of obtaining an original feature set of each sample in an updated compressive strength training sample set, carrying out feature selection in a sliding window mode, inputting the sliding selected original feature into each corresponding base learner to carry out model training, integrating each trained base learner to obtain a cement concrete compressive strength model, carrying out original feature selection in the sliding window mode to ensure that the input of each base learner has specificity, and obtaining output features with diversity to avoid overfitting phenomenon of integrated learning.
The proportioning parameters and external influence data of each cement concrete sample are set. According to factors possibly influencing the compressive strength of the cement concrete, setting proportioning data and external influence data of different components of the cement concrete. The proportioning parameters at least comprise one of the following: cement content, blast furnace slag content, fly ash content, water reducing agent, coarse aggregate and fine aggregate. The external influence parameters include one of the following: age, temperature, humidity. It can be understood that the ratio parameters of the samples and the external influence parameters are set as long as a parameter is ensured to be different among the samples. Under the condition that different proportioning parameters and external influence parameters are set on the basis of each cement concrete sample, the compressive strength of each training sample is measured, and the proportioning parameters, the external influence parameters and the measured compressive strength of the cement concrete are constructed into a compressive strength training sample set. The compressive strength is measured by a national standard measuring method, so long as the proportioning parameters and the external influence parameters of each cement concrete sample are set differently.
In the constructed compressive strength training sample set, normal compressive strength samples and abnormal compressive strength samples are included, and abnormal samples need to be removed. In the embodiment of the application, an abnormal compressive strength sample in a compressive strength training sample set is removed by adopting an improved random sampling consistency algorithm, so that an updated compressive strength training sample set is obtained. The difference between the improved random sampling consistency algorithm in the embodiment of the application and the distance sampling consistency algorithm in the prior art is that: the random sampling consistency algorithm randomly selects 4 sample points, and can not ensure that the sample point distances are sufficiently dispersed; the improved random sampling consistency algorithm randomly selects 2 sample points firstly, then determines a third sample point according to the midpoint of the 2 sample points, and finally determines a 4 th sample point according to the inner centers of triangles formed by the first 3 sample points, thereby ensuring that the sample points are sufficiently dispersed.
As an alternative implementation, as shown in fig. 2, step S2 includes:
s201, acquiring a first point cloud image corresponding to a compressive strength training sample set, randomly selecting a first sample and a second sample from the first point cloud image, calculating Euclidean distance between the first sample and the second sample to obtain coordinates of a midpoint between the first sample and the second sample, marking all samples in the first point cloud image except the first sample and the second sample as second point cloud images, and respectively calculating first Euclidean distance between the midpoint and each sample in the second point cloud image;
s202, determining a sample with the largest first Euclidean distance as a third sample, calculating to obtain coordinates of the inner center of a triangle formed by the first sample, the second sample and the third sample, marking all samples in the second point cloud image except the third sample as third point cloud images, and respectively calculating second Euclidean distances between the inner center and each sample in the third point cloud image;
s203, determining a sample with the largest second Euclidean distance as a fourth sample, and marking all samples in the third point cloud image except the fourth sample as fourth point cloud images;
s204, calculating transformation matrixes of the four samples according to the determined first sample, second sample, third sample and fourth sample, and obtaining a rejected sample model M.
As shown in fig. 4, the point cloud image corresponding to the compressive strength training sample set is marked as a first point cloud image, and a first sample and a second sample are selected from the first point cloud image in a random sampling manner, and the random sampling manner cannot ensure that the first sample and the second sample are sufficiently dispersed, so that multiple iterative selection needs to be performed when the first sample and the second sample are selected, and two samples with the largest euclidean distance are selected from the selected samples as the first sample and the second sample, so that the first sample and the second sample are sufficiently dispersed. And selecting a third sample and a fourth sample in an improved random sampling consistency mode, and ensuring that sample points are sufficiently dispersed so that the acquired compressive strength samples are sufficiently in line with the conditions. Specifically, point clouds corresponding to all samples except the first sample and the second sample in the first point cloud image are marked as second point cloud images. And calculating the midpoint coordinates of the first sample and the second sample according to Euclidean distance between the first sample and the second sample, calculating the first Euclidean distance between the midpoint and all samples in the second point cloud image, and selecting the sample with the largest first Euclidean distance as a third sample. Marking point cloud images corresponding to all samples except a third sample in the second point cloud image as a third point cloud image, constructing the first sample, the second sample and the third sample as a triangle, calculating the inner coordinates of the triangle, calculating the second Euclidean distance between all samples in the third point cloud image and the inner, and taking the sample with the largest second Euclidean distance as a fourth sample. And marking the point cloud images corresponding to all samples except the fourth sample in the third point cloud image as a fourth point cloud image.
As an alternative implementation manner, the point cloud image constructing step of step 2 includes:
taking the proportioning parameter, the external influence parameter and the compressive strength of each cement concrete sample as input to form n-dimensional data corresponding to each sample, wherein the number of samples is m;
forming a matrix X of n rows and m columns from multidimensional data of m samples, carrying out zero-mean on each row of the matrix X, and calculating to obtain a covariance matrix of the matrix X;
calculating eigenvalues and corresponding eigenvectors of the covariance matrix, sequentially sequencing the eigenvectors according to the order of the eigenvalues from large to small, and selecting the eigenvectors of the first 3 rows to form a matrix P of 3 rows and m columns; and constructing a matrix P to obtain a first point cloud image.
In the embodiment of the application, the n is 11, specifically, the cement content, the blast furnace slag content, the fly ash content, the water reducing agent, the coarse aggregate, the fine aggregate, the age, the temperature, the humidity and the compressive strength of the cement concrete sample are taken as inputs to form 11-dimensional data, the number of samples in the cement concrete sample set is m, the multi-dimensional data of m samples form an 11-row m-column matrix X, zero-averaging is carried out on each row of the matrix X, namely the average value of the row is subtracted, and the covariance matrix with the matrix X is calculated. Because the sample in the high-dimensional space has sparsity, the model is difficult to find data features, so that dimension processing needs to be performed on the matrix X, 11-dimensional data is reduced to 3-dimensional data, calculation is simpler, feature vectors of the first 3 rows in the matrix X are selected to form a matrix P of 3 rows and m columns, and the matrix P is constructed as a first point cloud picture.
As an alternative implementation, as shown in fig. 2, step S2 further includes:
s205, defining a training sample set I best The training sample set I best Initial value is empty set, training sample set I best The number of samples in (a) is N 0 Setting the maximum iteration number K of the algorithm max ;
S206, setting an error threshold lambda of the sample model M to be removed, respectively calculating a projection error d between each sample in the fourth point cloud picture and the sample model M to be removed, if the projection error d is smaller than the error threshold lambda, the sample is a qualified sample, otherwise, the sample is an abnormal sample, and the sample point is removed;
and forming a transformation matrix by the first sample, the second sample, the third sample and the fourth sample to obtain a rejected sample model M.
S207, forming a set I by all qualified samples 1 Statistics of the set I 1 Number of samples N 1 If N 1 Greater than N 0 Set I best Update to set I 1 Set I best The number of samples of (a) is updated to N 1 ;
S208, repeatedly executing the steps S201-S204, and updating the iteration number k of each execution step;
s209, when the iteration number reaches K max And when the training sample set obtained by the iteration is an updated compressive strength training sample set.
As an alternative implementation, the calculation formula of the iteration number k is:
p is the confidence, 0.995 is taken, w is the ratio of the number of samples in the set acquired at each iteration to the total number of samples, and m is taken to be 4. Where the value of k is not simply increased by 1, m in the embodiment of the present application takes a first sample, a second sample, a third sample, and a fourth sample.
Extracting original features of each sample in the compressive strength training sample set, constructing a corresponding original feature set, selecting the original feature set in a sliding window mode, and inputting the original features selected based on sliding each time into each corresponding basic learner for training. Specifically, the original features collected by the first window are input into a first base learner, the original features collected by the second window are input into a second base learner, and the original features collected by the third window are input into a third base learner for training, so that corresponding first output features, second output features and third output features are obtained. The original characteristics are selected in a sliding window mode, so that the input of each base learner is ensured to have specificity, the acquired output characteristics have diversity, and the overfitting phenomenon of integrated learning is avoided.
As an alternative implementation, step S3 includes:
selecting features from the original feature set by using a first window size, wherein the selected features form first features, and inputting the first features into a first base learner for model training to obtain first output features;
selecting features from the original feature set by using a second window size, forming second features by the selected features, inputting the second features into a second base learner for model training to obtain second output features;
selecting features from the original feature set by a third window size, forming a third feature by the selected features, inputting the third feature into a third base learner for model training to obtain a third output feature;
wherein the first, second and third features are all different.
As shown in fig. 4, when samples are selected in a sliding window manner, the sliding window may have the same or different sizes, i.e., the number of features collected in each window may be the same or different. For example, the number of original features collected by the first window is 2, the number of original features input into the first base learner is 2, and the number of original features collected by the second window and the third window may be 2 or not, so long as the fact that the features selected each time are different is ensured. In the embodiment of the application, the size of each sliding window is not specified, and only the mapping relation between each sliding window and each basic learner is required to be determined, namely, the original features acquired by the first window are input into the first basic learner; the original features collected by the second window are input into a second base learner; the original features collected by the third window are input into the third base learner, so that the input of each base learner is ensured to have specificity, the formed first output features, second output features and third output features have diversity, and the overfitting phenomenon of integrated learning is avoided.
And (3) forming a reconstructed feature set by the first output feature, the second output feature, the third output feature and the original feature set, inputting the reconstructed feature set into a fourth base learner for model training, obtaining each parameter of the first base learner, the second base learner, the third base learner and the fourth base learner, and integrating the first base learner, the second base learner, the third base learner and the fourth base learner to obtain a trained cement concrete compressive strength model.
As an alternative implementation, the first base learner, the second base learner, the third base learner, and the fourth base learner use the same or different algorithms. That is, the first base learner, the second base learner, and the third base learner may all select the same algorithm, or may select different algorithms. For example, the first base learner, the second base learner, and the third base learner all select an AdaBoost algorithm; or the first base learner selects the AdaBoost algorithm, and the second base learner and the third learner select the GBDT algorithm or the GBDT algorithm and the CatBoost algorithm respectively. Wherein the first base learner, the second base learner, and the third base learner may select the following algorithm: the bagging class algorithm (Random Forest, etc.), the boosting class algorithm (AdaBoost, GBDT, XGBoost, lightGBM, catboost, etc.), the Deep Forest, SVR, DT, etc. The fourth base learner may select the following algorithm: GBDT, XGBoost, lightGBM, catboost, etc.
In some embodiments, to evaluate the performance index of the trained cement concrete compressive strength model and the original Stacking, the method comprises the following stepsApplication examples training was performed by sampling a public dataset, which was literature { Yeh I C.modeling slump of concrete with fly ash and superplasticizer [ J ]]Set 1030 of cement concrete compressive strength data sets acquired 559-572 (6), computers and Concrete, an International Journal,2008,5. The selected 1030 sets of data sets were divided into training and test sets at a ratio of 8:2. The algorithm of the first base learner is set as a Support Vector Regression (SVR), the algorithm of the second base learner is a Decision Tree (DT), and the algorithms of the fourth base learner and the third base learner are GBDT. And respectively inputting the training sets into each basic learner to train so as to obtain the cement concrete compressive strength model after training. When the first base learner is trained, the kernel function is set as RBF kernel function, C is 1000, gamma is 0.0001, epsilon is 0.04. When training the second base learner, n_optimizers are set to 500. When training the fourth and third base learners, n_rest is set to 1000, learning_rate is set to 0.1, max_depth is set to 5, and loss is set to huber. Conventional and improved algorithms mean absolute error (Mean Absolute Error, MAE), root mean square error (Root Mean Square Error, RMSE), mean absolute percent error (Mean Absolute Percentage Error, MAPE), decision coefficients (Coefficient of Determination, R) for test and training sets 2 ) See table below.
Evaluation index table for traditional and improved integrated learning
As can be seen from the above table, for the disclosed dataset, the various indices of the predictive model on the test set after improvement are all superior to those of the conventional predictive model. Wherein MAE is reduced by approximately 20% and RMSE is reduced by approximately 25% and MAPE is reduced by approximately 24% on the test set, R 2 The improvement by 1.8 percent is within acceptable limits, although the consumption increases somewhat. This shows that the improved ensemble learning model has a greater performance improvement, which can more effectively mitigate the risk of overfitting.
In one embodiment of the present application as shown in fig. 5, a cement concrete compressive strength testing system 100 is provided, the system 100 comprising a sample acquisition module 11, a sample rejection module 12, a base learner module 13, a compressive strength model module 14, and a testing module 15.
The sample collection module 11 sets the proportioning parameter and the external influence of each cement concrete sample, correspondingly measures the compressive strength of each cement concrete sample, and constructs the proportioning parameter, the external influence parameter and the measured compressive strength data of each cement concrete sample into a compressive strength training sample set.
The sample rejection module 12 is configured to reject abnormal samples in the compressive strength training sample set, and obtain an updated compressive strength training sample set.
The base learner module 13 acquires an original feature set of the updated compressive strength training sample set, performs feature selection on the original feature set in a sliding window mode, and correspondingly inputs the original feature selected based on each sliding into the first base learner, the second base learner and the third base learner to perform model training to obtain a first output feature, a second output feature and a third output feature respectively.
The compressive strength model module 14 constructs the first output feature, the second output feature, the third output feature and the original feature set to obtain a reconstructed feature set, inputs the reconstructed feature set to a fourth base learner for model training, and integrates the trained first base learner, second base learner, third base learner and fourth base learner to obtain a trained cement concrete compressive strength model.
The detection module 15 inputs the cement concrete sample to be detected into a cement concrete compressive strength model, and can measure and obtain the compressive strength value of the cement concrete sample to be detected.
An embodiment of the present application also provides an electronic device including a processor and a memory, in which executable code is stored, which when executed causes the processor to perform the above-described method of measuring compressive strength of cement concrete. From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of a necessary general hardware platform, or may be implemented by means of a combination of hardware and software. Based on such understanding, the foregoing aspects, in essence and portions contributing to the art, may be embodied in the form of a computer program product, which may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
The above disclosure is illustrative of the preferred embodiments of the present application, but it should not be construed as limiting the scope of the application as will be understood by those skilled in the art: changes, modifications, substitutions, combinations, and simplifications may be made without departing from the spirit and scope of the application and the appended claims, and equivalents may be substituted and still fall within the scope of the application.
Claims (10)
1. A method for detecting compressive strength of cement concrete, the method comprising the steps of:
s1, setting proportioning parameters and external influence parameters of each cement concrete sample, correspondingly measuring to obtain compressive strength of each cement concrete sample, and constructing a compressive strength training sample set from the proportioning parameters, the external influence parameters and the measured compressive strength data of each cement concrete sample;
s2, based on an improved random sampling consistency algorithm, abnormal compressive strength samples are removed from the compressive strength training sample set, and an updated compressive strength training sample set is obtained;
s3, acquiring an original feature set of the updated compressive strength training sample set, carrying out feature selection on the original feature set in a sliding window mode, and correspondingly inputting the original feature selected based on each sliding into a first base learner, a second base learner and a third base learner to carry out model training to obtain a first output feature, a second output feature and a third output feature respectively;
s4, the first output feature, the second output feature, the third output feature and the original feature set form a reconstructed feature set, the reconstructed feature set is input into a fourth base learner to perform model training, and the first base learner, the second base learner, the third base learner and the fourth base learner are integrated to obtain a trained cement concrete compressive strength model;
s5, inputting the cement concrete sample to be tested into the cement concrete compressive strength model to obtain the compressive strength value of the cement concrete sample to be tested.
2. The method for detecting compressive strength of cement concrete according to claim 1, wherein the step S1 comprises:
the proportioning parameters at least comprise one of the following: cement content, blast furnace slag content, fly ash content, water reducing agent, coarse aggregate and fine aggregate;
the external influence parameters at least comprise one of age, temperature and humidity.
3. The method for detecting compressive strength of cement concrete according to claim 1, wherein the step S2 comprises:
s201, acquiring a first point cloud image corresponding to the updated compressive strength training sample set, randomly selecting a first sample and a second sample from the first point cloud image, calculating the Euclidean distance between the first sample and the second sample to obtain a midpoint coordinate between the first sample and the second sample, marking all samples in the first point cloud image, from which the first sample and the second sample are removed, as second point cloud images, and respectively calculating the first Euclidean distance between the midpoint and each sample in the second point cloud image;
s202, determining a sample with the largest first Euclidean distance as a third sample, calculating to obtain the inner coordinates of a triangle formed by the first sample, the second sample and the third sample, marking all samples of the second point cloud image except the third sample as third point cloud images, and respectively calculating the second Euclidean distance between the inner and each sample in the third point cloud images;
s203, determining a sample with the largest second Euclidean distance as a fourth sample, and marking all samples of the third point cloud image from which the fourth sample is removed as a fourth point cloud image;
s204, calculating transformation matrixes of the four samples according to the determined first sample, second sample, third sample and fourth sample, and obtaining a rejected sample model M.
4. The method for detecting compressive strength of cement concrete according to claim 3, wherein the constructing of the point cloud image in the step 2 comprises:
taking the proportioning parameter, the external influence parameter and the compressive strength of each cement concrete sample as input to form n-dimensional data corresponding to each sample, wherein the number of samples is m;
forming n rows and m columns of matrix X from multi-dimensional data of m samples, carrying out zero-mean on each row of the matrix X, and calculating to obtain a covariance matrix of the matrix X;
calculating eigenvalues and corresponding eigenvectors of the covariance matrix, sequentially sequencing the eigenvectors according to the order of the eigenvalues from large to small, and selecting the eigenvectors of the first 3 rows to form a matrix P of 3 rows and m columns;
and constructing a matrix P to obtain a first point cloud image.
5. A method for detecting the compressive strength of cement concrete according to claim 3, wherein the step S2 further comprises:
s205, defining a training sample set I best The training sample set I best Initial value of the set I is an empty set best The number of samples in (a) is N 0 Setting the maximum iteration number K of the algorithm max ;
S206, setting an error threshold lambda of the sample model M to be removed, respectively calculating a projection error d between each sample in the fourth point cloud picture and the sample model M to be removed, if the projection error d is smaller than the error threshold lambda, the sample is a qualified sample, otherwise, the sample is an abnormal sample, and the sample point is removed;
s207, forming a set I by all qualified samples 1 Statistics of the set I 1 Number of samples N 1 If N 1 Greater than N 0 Set I best Update to set I 1 Set I best The number of samples of (a) is updated to N 1 ;
S208, repeatedly executing the steps S201-S204, and updating the iteration number k of each execution step;
s209, when the iteration number reaches K max And when the training sample set obtained by the iteration is an updated compressive strength training sample set.
6. The method for detecting compressive strength of cement concrete according to claim 5, wherein the step S208 comprises:
the calculation formula of the iteration number k is as follows:
p is the confidence, 0.995 is taken, w is the ratio of the number of samples in the set acquired at each iteration to the total number of samples, and m is taken to be 4.
7. The method for detecting compressive strength of cement concrete according to claim 1, wherein the step S3 comprises:
feature selection is carried out from the original feature set by a first window size, the selected features form first features, the first features are input into the first base learner for model training, and first output features are obtained;
feature selection is carried out from the original feature set according to a second window size, the selected features form second features, the second features are input into the second base learner for model training, and second output features are obtained;
feature selection is carried out from the original feature set by a third window size, the selected features form a third feature, the third feature is input into the third base learner for model training, and a third output feature is obtained;
wherein the first, second and third features are all different.
8. The method for detecting compressive strength of cement concrete according to claim 7, wherein the step S3 comprises:
the first base learner, the second base learner, the third base learner, and the fourth base learner use the same or different algorithms.
9. A system for testing the compressive strength of cement concrete, said system comprising:
the sample acquisition module is used for setting different proportioning parameters and external influence parameters of each compressive strength training sample, correspondingly measuring the compressive strength of each compressive strength training sample, acquiring the compressive strength of each compressive strength training sample, and constructing a compressive strength training sample set of the cement concrete;
the sample removing module is used for removing abnormal samples in the compressive strength training sample set to obtain an updated compressive strength training sample set;
the basic learner module acquires an original feature set of the updated compressive strength training sample set, performs feature selection on the original feature set in a sliding window mode, and correspondingly inputs the original feature selected based on each sliding into a first basic learner, a second basic learner and a third basic learner for model training to obtain a first output feature, a second output feature and a third output feature respectively;
the compressive strength model module is used for forming the first output characteristic, the second output characteristic, the third output characteristic and the original characteristic set into a reconstructed characteristic set, inputting the reconstructed characteristic set into a fourth base learner for model training, integrating the first base learner, the second base learner, the third base learner and the fourth base learner, and constructing a trained cement concrete compressive strength model;
and the detection module is used for inputting the cement concrete sample to be detected into the cement concrete compressive strength model to obtain the compressive strength value of the cement concrete sample to be detected.
10. An electronic device, comprising:
a processor;
a memory having executable code stored therein that, when executed, causes the processor to perform the method of one or more of claims 1-8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310293500.8A CN117153297A (en) | 2023-03-17 | 2023-03-17 | Cement concrete compressive strength detection method, system and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310293500.8A CN117153297A (en) | 2023-03-17 | 2023-03-17 | Cement concrete compressive strength detection method, system and electronic equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117153297A true CN117153297A (en) | 2023-12-01 |
Family
ID=88885557
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310293500.8A Pending CN117153297A (en) | 2023-03-17 | 2023-03-17 | Cement concrete compressive strength detection method, system and electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117153297A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117763701A (en) * | 2024-02-22 | 2024-03-26 | 四川省交通勘察设计研究院有限公司 | method for predicting strength of steel-concrete connection transition surface of steel arch bridge and related products |
-
2023
- 2023-03-17 CN CN202310293500.8A patent/CN117153297A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117763701A (en) * | 2024-02-22 | 2024-03-26 | 四川省交通勘察设计研究院有限公司 | method for predicting strength of steel-concrete connection transition surface of steel arch bridge and related products |
CN117763701B (en) * | 2024-02-22 | 2024-05-07 | 四川省交通勘察设计研究院有限公司 | Method for predicting strength of steel-concrete connection transition surface of steel arch bridge and related products |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105354595B (en) | A kind of robust visual pattern classification method and system | |
CN104869126B (en) | A kind of network intrusions method for detecting abnormality | |
CN108777873A (en) | The wireless sensor network abnormal deviation data examination method of forest is isolated based on weighted blend | |
CN110121724A (en) | Abnormal detector, method for detecting abnormality and storage medium | |
JP7063389B2 (en) | Processing equipment, processing methods, and programs | |
CN112884721B (en) | Abnormality detection method, abnormality detection system and computer-readable storage medium | |
CN108764276A (en) | A kind of robust weights multi-characters clusterl method automatically | |
CN117153297A (en) | Cement concrete compressive strength detection method, system and electronic equipment | |
CN112348808A (en) | Screen perspective detection method and device | |
CN113624694A (en) | Inversion method and device for atmospheric methane concentration | |
CN110472659B (en) | Data processing method, device, computer readable storage medium and computer equipment | |
CN111915022A (en) | Gaussian process method and device for rapidly identifying stability coefficient of sliding karst dangerous rock | |
Adde et al. | Too many candidates: Embedded covariate selection procedure for species distribution modelling with the covsel R package | |
CN113361194B (en) | Sensor drift calibration method based on deep learning, electronic equipment and storage medium | |
CN111091194B (en) | Operation system identification method based on CAVWBB _ KL algorithm | |
CN113011086A (en) | Estimation method of forest biomass based on GA-SVR algorithm | |
CN114943674A (en) | Defect detection method, electronic device and storage medium | |
CN112465821A (en) | Multi-scale pest image detection method based on boundary key point perception | |
CN110045691B (en) | Multi-task processing fault monitoring method for multi-source heterogeneous big data | |
CN116702005A (en) | Neural network-based data anomaly diagnosis method and electronic equipment | |
JP2020173674A (en) | Data analysis device, data analysis method and analysis program | |
CN113538239B (en) | Interpolation method based on space-time autoregressive neural network model | |
CN112819079A (en) | Model sampling algorithm matching method and device and electronic equipment | |
JP2020161044A (en) | System, method, and program for managing data | |
Cai et al. | Automatic detection of landscape painting elements based on machine learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |