A method for detecting and analyzing concrete mechanical parameters

Abstract

A concrete mechanical parameter detection and analysis method relates to the field of concrete parameter detection, including: obtaining multiple original images of concrete samples, using image enhancement algorithms to preprocess the original images to obtain multiple super-resolution original images; using convolutional neural networks to obtain original structured data corresponding to the super-resolution original images; mathematically modeling the original structured data and the super-resolution original images to obtain multiple pre-trained models of concrete samples; model training, selecting the optimal model as a concrete mechanical parameter detection model; using the model to characterize the mechanical structure characteristics of concrete; obtaining mechanical parameter data of concrete samples, using the mechanical parameter data and the concrete mechanical parameter detection model to perform assignment operations to obtain concrete mechanical parameters. The present invention improves the detection accuracy of concrete mechanical parameters, simplifies the detection steps, and improves the detection efficiency.

Description

Technical Field

[0001] The invention relates to the technical field of concrete parameter detection, and in particular to a concrete mechanical parameter detection and analysis method. Background Art

[0002] In today's society, concrete is widely used, including in the fields of construction, civil engineering, geothermal engineering, mechanical engineering, manufacturing, etc. Concrete is made of binder, aggregate and water in proportion. Therefore, concrete is not a single material, but a mixture of multiple ingredients. Concrete is used more and more because of its rich raw materials, simple production process and low production cost. Concrete has a wide range of strength grades, good durability and high compressive strength. Therefore, the scope of use of concrete is becoming more and more extensive.

[0003] At present, the detection of concrete mechanical parameters is mainly carried out by detection equipment, for example, commercial concrete strength tester is used to detect concrete strength; some indicators are also directly measured by measurement methods, for example, when detecting the standard value of concrete cube compressive strength, GB50010 "Concrete Structure Design Code" stipulates: The standard value of concrete cube compressive strength refers to the cube compressive strength with a 95% guarantee rate measured by the standard test method at the age of 28 days for a standard cube specimen with a side length of 150mm made and maintained according to the standard method. In addition, there is also a method of using a spherical random distribution model to detect concrete mechanical parameters, but this method has the following problems, that is, the spherical random distribution model used can only represent the structural characteristics of the concrete appearance, and its internal spatial structure cannot be characterized. Therefore, the use of the spherical random distribution model for the detection of concrete mechanical parameters will result in certain detection errors and cannot achieve high-precision measurement. Summary of the invention

[0004] In order to solve the technical problem of low precision of existing concrete mechanical parameter detection methods, the present invention provides a concrete mechanical parameter detection and analysis method. The present invention uses multiple original images to obtain super-resolution original images and corresponding original structured data, and obtains a corresponding concrete mechanical parameter detection model by mathematically modeling and training the original structured data and the super-resolution original images. Then, an assignment operation is performed according to the mechanical parameter data of the concrete sample to obtain the concrete mechanical parameters, thereby improving the detection accuracy of the concrete mechanical parameters. In addition, only one set of concrete mechanical parameter detection models can be used to realize the detection of concrete mechanical parameters, namely, strength parameters and grading parameters, thereby simplifying the measurement steps and improving the measurement efficiency.

[0005] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0006] A concrete mechanical parameter detection and analysis method of the present invention mainly comprises the following steps:

[0007] Step S1: obtaining multiple original images of concrete samples, and preprocessing the original images using an image enhancement algorithm to obtain multiple super-resolution original images;

[0008] Step S2: using a convolutional neural network to obtain original structured data corresponding to the super-resolution original image;

[0009] Step S3: performing mathematical modeling on the original structured data and the super-resolution original image to obtain multiple pre-trained models of the concrete sample;

[0010] Step S4: training multiple pre-trained models, sorting the training results, and selecting the model with the best training effect as the concrete mechanical parameter detection model;

[0011] Step S5: using a concrete mechanical parameter detection model to characterize the mechanical structure characteristics of the concrete;

[0012] Step S6: Acquire mechanical parameter data of the concrete sample, and use the mechanical parameter data and the concrete mechanical parameter detection model to perform assignment operations to obtain concrete mechanical parameters, and analyze the obtained concrete mechanical parameters to complete concrete mechanical parameter detection and analysis.

[0013] Furthermore, in step S1, the specific operation process of the image enhancement algorithm is as follows:

[0014] Step S1.1: construct the model using weighted least squares method;

[0015] Step S1.2: Generate an affine matrix using a soft clustering algorithm;

[0016] Step S1.3: After inputting the RGB standard image, the generated affine matrix is ​​used to optimize the constructed model;

[0017] Step S1.4: Set the corresponding confidence parameters and input the original image into the soft clustering solver for solving to obtain the processed super-resolution original image.

[0018] Furthermore, the confidence parameters include: depth map hole filling confidence, depth map super-resolution confidence and depth map correction confidence.

[0019] Furthermore, the method for solving the confidence of the depth map hole filling is:

[0020] First, the depth value of the depth map is obtained, and then the confidence of the missing area in the depth map is set to 0, and the confidence of the area in the depth map that is not missing and has a depth value is set to 1. All areas in the depth map are marked in the above way to obtain the depth map hole filling confidence matrix.

[0021] Furthermore, the method for solving the super-resolution confidence of the depth map is:

[0022] First, an RGB standard image is input, and the depth map is upsampled under the guidance of the RGB standard image. Then, the depth map is interpolated using the bicubic interpolation algorithm to make its resolution the same as that of the RGB standard image. Finally, the Gaussian function envelope is used to obtain the super-resolution confidence of the depth map.

[0023] Furthermore, the method for solving the depth map correction confidence is:

[0024] First, the original depth map is processed by a transform filter to obtain a processed depth map. Then, the difference between the original depth map and the processed depth map is divided by 2 times the square of the bandwidth to obtain the depth map correction confidence.

[0025] Furthermore, in step S3, the software Mathpix is ​​used to perform mathematical modeling on the original structured data and the super-resolution original image.

[0026] Furthermore, the specific operation steps of step S4 are as follows:

[0027] S4.1: Build a neural network classification model in model.py;

[0028] The neural network classification model includes: convolution layer 1, maximum pooling layer 1, convolution layer 2, maximum pooling layer 2, convolution layer 3, maximum pooling layer 3, flattening layer and linear layer;

[0029] Convolutional layer 1: The number of input channels is 3, the number of output channels is 32, the convolution kernel size is 3×3, the stride is 1, and the padding is 2;

[0030] The convolution kernel size of the max pooling layer 1, max pooling layer 2, and max pooling layer 3 is 2;

[0031] Convolutional layer 2: The number of input channels is 32, the number of output channels is 32, the convolution kernel size is 3×3, the stride is 1, and the padding is 2;

[0032] Convolutional layer 3: The number of input channels is 32, the number of output channels is 64, the convolution kernel size is 5×5, the stride is 1, and the padding is 2;

[0033] Linear layer: The size of the input sample is 64×4×4, and the size of the output sample is 64;

[0034] S4.2: Divide the dataset into training dataset and test dataset, and use DataLoader to load the dataset;

[0035] S4.3: Set the network model, loss function, optimizer and network training parameters, and start training and testing;

[0036] S4.4: Sort the training results and select the model with the best training effect as the concrete mechanical parameter detection model.

[0037] Furthermore, the loss function is a cross entropy loss function; the network training parameters include the number of training times, the number of tests and the number of training rounds; and the optimizer is an SGD stochastic gradient descent optimizer.

[0038] Furthermore, the concrete mechanical parameter detection model includes: a strength parameter detection model and a grading parameter detection model.

[0039] Furthermore, the intensity parameter detection model is constructed using the nonlinear least squares method as follows:

[0040] f c =af 1 (1 / Ab)(-cB 2 -dB+e)

[0041] Among them, f c Indicates the strength of concrete, a, b, c, d, e are coefficients, f 1 It represents cement strength, A represents water-cement ratio, and B represents fly ash content.

[0042] Furthermore, the value range of the coefficient a is: 0.3-0.4; the value range of the coefficient b is: 0.3-0.4; the value range of the coefficient c is: 0.7-0.8; the value range of the coefficient d is: 0.3-0.4; the value range of the coefficient e is: 1.3-1.5.

[0043] Furthermore, the grading parameter detection model specifically adopts a cube compressive strength standard value calculation model and a grade classification judgment model. The cube compressive strength standard value calculation model is:

[0044] f cu,k =0.348a k (f 1 ) 0.55 (1-1.645δ) 0.45

[0045] Among them, a k Denotes the size factor of the concrete test block, f 1represents cement strength, δ represents the standard deviation of concrete cube compressive strength;

[0046] The grade classification judgment model consists of a data input module, a data standardization processing module, a judgment module and a data output module;

[0047] Wherein, the data input module is used to input the standard value of cube compressive strength;

[0048] The data standardization processing module is used to perform format conversion processing on the input cube compressive strength standard value;

[0049] The determination module is used to determine the standard value data of the cube compressive strength after the format conversion process, and obtain the corresponding concrete grade.

[0050] The beneficial effects of the present invention are:

[0051] A concrete mechanical parameter detection and analysis method of the present invention obtains a super-resolution original image and corresponding original structured data by utilizing multiple original images, and obtains a corresponding concrete mechanical parameter detection model by mathematically modeling and training the original structured data and the super-resolution original image, and then performs an assignment operation according to the mechanical parameter data of the concrete sample to obtain the concrete mechanical parameters.

[0052] The concrete mechanical parameter detection and analysis method of the present invention conforms to the actual situation of concrete mechanical parameter detection, improves the concrete mechanical parameter detection accuracy, simplifies the detection steps, and improves the detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The present invention is a flow chart of a concrete mechanical parameter detection and analysis method. DETAILED DESCRIPTION

[0054] like Figure 1 As shown, a concrete mechanical parameter detection and analysis method of the present invention specifically includes the following steps:

[0055] Step S1: obtaining multiple original images of concrete samples, and preprocessing the original images using an image enhancement algorithm to obtain multiple super-resolution original images;

[0056] Among them, the specific operation process of the image enhancement algorithm used is as follows:

[0057] Step S1.1: construct the model using weighted least squares method;

[0058] Step S1.2: Generate an affine matrix using a soft clustering algorithm;

[0059] Step S1.3: After inputting the RGB standard image, the generated affine matrix is ​​used to optimize the constructed model;

[0060] Step S1.4: Set the corresponding confidence parameters and input the original image into the soft clustering solver for solving to obtain the processed super-resolution original image.

[0061] The confidence parameters include: depth map hole filling confidence, depth map super-resolution confidence and depth map correction confidence.

[0062] The method for solving the depth map hole filling confidence is as follows:

[0063] First, the depth value of the depth map is obtained, and then the confidence of the missing area in the depth map is set to 0, and the confidence of the area in the depth map that is not missing and has a depth value is set to 1. All areas in the depth map are marked in the above way to obtain the depth map hole filling confidence matrix.

[0064] The method for solving the super-resolution confidence of the depth map is as follows:

[0065] First, an RGB standard image is input, and the depth map is upsampled under the guidance of the RGB standard image. Then, the depth map is interpolated using the bicubic interpolation algorithm to make its resolution the same as that of the RGB standard image. Finally, the Gaussian function envelope is used to obtain the super-resolution confidence of the depth map.

[0066] The method for solving the depth map correction confidence is as follows:

[0067] First, the original depth map is processed by a transform filter to obtain a processed depth map. Then, the difference between the original depth map and the processed depth map is divided by 2 times the square of the bandwidth to obtain the depth map correction confidence.

[0068] Step S2: using Convolutional Neural Networks (CNN) to obtain original structured data corresponding to the super-resolution original image;

[0069] Step S3: performing mathematical modeling on the original structured data and the super-resolution original image to obtain multiple pre-trained models of the concrete sample;

[0070] Specifically, software Mathpix may be used to perform mathematical modeling on the original structured data and the super-resolution original image, but the invention is not limited thereto.

[0071] Step S4: training multiple pre-trained models, sorting the training results, and selecting the model with the best training effect as the concrete mechanical parameter detection model;

[0072] The specific operation steps of step S4 are as follows:

[0073] S4.1: Build a neural network classification model in model.py;

[0074] The neural network classification model includes: convolution layer 1, maximum pooling layer 1, convolution layer 2, maximum pooling layer 2, convolution layer 3, maximum pooling layer 3, flattening layer and linear layer;

[0075] Convolutional layer 1: The number of input channels is 3, the number of output channels is 32, the convolution kernel size is 3×3, the stride is 1, and the padding is 2;

[0076] The convolution kernel size of the max pooling layer 1, max pooling layer 2, and max pooling layer 3 is 2;

[0077] Convolutional layer 2: The number of input channels is 32, the number of output channels is 32, the convolution kernel size is 3×3, the stride is 1, and the padding is 2;

[0078] Convolutional layer 3: The number of input channels is 32, the number of output channels is 64, the convolution kernel size is 5×5, the stride is 1, and the padding is 2;

[0079] Linear layer: The size of the input sample is 64×4×4, and the size of the output sample is 64.

[0080] S4.2: Divide the dataset into training dataset and test dataset, and use DataLoader to load the dataset.

[0081] S4.3: Set the network model, loss function, optimizer and network training parameters, and start training and testing;

[0082] The training steps include: optimizing the network model using an optimizer, clearing the gradient, back-propagating to obtain the gradient value of each parameter node, and finally optimizing the network training parameters.

[0083] The test steps include: when the gradient is zero, no tuning operation is performed, and the network loss at this time is the loss of part of the data on the network model, and the data type of the loss is the tensor data type; the number of correct predictions and the loss on the test data set are calculated, so as to calculate the accuracy on the test data set, and the accuracy is the number of correct predictions divided by the length of the test data set, and the corresponding results are saved for each round of testing.

[0084] The loss function is preferably a cross entropy loss function.

[0085] The network training parameters mainly include the number of training times, the number of tests and the number of training rounds.

[0086] The optimizer is preferably a SGD stochastic gradient descent optimizer.

[0087] S4.4: sort the training results and select the model with the best training effect as the concrete mechanical parameter detection model;

[0088] The concrete mechanical parameter detection model includes: a strength parameter detection model and a grading parameter detection model.

[0089] Specifically, the intensity parameter detection model can be constructed using the nonlinear least squares method as follows:

[0090] f c =af 1 (1 / Ab)(-cB 2 -dB+e)

[0091] Among them, f c Indicates the strength of concrete, a, b, c, d, e are coefficients, f 1 It represents cement strength, A represents water-cement ratio, and B represents fly ash content.

[0092] The value ranges of each coefficient are as follows:

[0093] The value range of coefficient a is: 0.3-0.4;

[0094] The value range of coefficient b is: 0.3-0.4;

[0095] The value range of coefficient c is: 0.7-0.8;

[0096] The value range of coefficient d is: 0.3-0.4;

[0097] The value range of coefficient e is: 1.3-1.5.

[0098] The grading parameter detection model may specifically adopt a cube compressive strength standard value calculation model and a grade classification judgment model, wherein the cube compressive strength standard value calculation model is:

[0099] f cu,k =0.348a k (f 1 ) 0.55 (1-1.645δ) 0.45

[0100] Among them, a k Denotes the size factor of the concrete test block, f 1 represents cement strength and δ represents the standard deviation of concrete cube compressive strength.

[0101] The said grade division judgment model consists of a data input module, a data standardization processing module, a judgment module and a data output module;

[0102] Wherein, the data input module is used to input the standard value of cube compressive strength;

[0103] The data standardization processing module is used to perform format conversion processing on the input cube compressive strength standard value;

[0104] The determination module is used to determine the standard value data of the cube compressive strength after the format conversion process, and obtain the corresponding concrete grade.

[0105] At present, China's ordinary concrete strength grades are divided into 14 levels: C15, C20, C25, C30, C35, C40, C45, C50, C55, C60, C65, C70, C75 and C80.

[0106] Step S5: using a concrete mechanical parameter detection model to characterize the mechanical structure characteristics of the concrete;

[0107] Step S6: Acquire mechanical parameter data of the concrete sample, and use the mechanical parameter data and the concrete mechanical parameter detection model to perform assignment operations to obtain concrete mechanical parameters, and analyze the obtained concrete mechanical parameters to complete concrete mechanical parameter detection and analysis.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein, but these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting and analyzing concrete mechanical parameters, characterized in that: The following steps are involved: Step S1: obtaining multiple original images of concrete samples, and preprocessing the original images using an image enhancement algorithm to obtain multiple super-resolution original images; Step S1.1: construct the model using weighted least squares method; Step S1.2: Generate an affine matrix using a soft clustering algorithm; Step S1.3: After inputting the RGB standard image, the generated affine matrix is ​​used to optimize the constructed model; Step S1.4: setting corresponding confidence parameters, inputting the original image into the soft clustering solver for solving, so as to obtain the processed super-resolution original image; Step S2: using a convolutional neural network to obtain original structured data corresponding to the super-resolution original image; Step S3: performing mathematical modeling on the original structured data and the super-resolution original image to obtain multiple pre-trained models of the concrete sample; Step S4: training multiple pre-trained models, sorting the training results, and selecting the model with the best training effect as the concrete mechanical parameter detection model; S4.1: Build a neural network classification model in model.py; The neural network classification model includes: convolution layer 1, maximum pooling layer 1, convolution layer 2, maximum pooling layer 2, convolution layer 3, maximum pooling layer 3, flattening layer and linear layer; Convolutional layer 1: The number of input channels is 3, the number of output channels is 32, the convolution kernel size is 3×3, the stride is 1, and the padding is 2; The convolution kernel size of the max pooling layer 1, max pooling layer 2, and max pooling layer 3 is 2; Convolutional layer 2: The number of input channels is 32, the number of output channels is 32, the convolution kernel size is 3×3, the stride is 1, and the padding is 2; Convolutional layer 3: The number of input channels is 32, the number of output channels is 64, the convolution kernel size is 5×5, the stride is 1, and the padding is 2; Linear layer: The size of the input sample is 64×4×4, and the size of the output sample is 64; S4.2: Divide the dataset into training dataset and test dataset, and use DataLoader to load the dataset; S4.3: Set the network model, loss function, optimizer and network training parameters, and start training and testing; S4.4: sort the training results and select the model with the best training effect as the concrete mechanical parameter detection model; Step S5: using a concrete mechanical parameter detection model to characterize the mechanical structure characteristics of the concrete; The concrete mechanical parameter detection model includes: a strength parameter detection model and a graded parameter detection model; the strength parameter detection model is constructed using a nonlinear least squares method, as follows: f c =af1(1 / A-b)(-cB 2 -dB+e) Among them, f c represents the strength of concrete, a, b, c, d, e all represent coefficients, f1 represents the strength of cement, A represents the water-binder ratio, and B represents the fly ash content; The grading parameter detection model specifically adopts a cube compressive strength standard value calculation model and a grade classification judgment model. The cube compressive strength standard value calculation model is: f cu,k =0.348a k (f1) 0.55 (1-1.645δ) 0.45 Among them, a k represents the size factor of the concrete test block, f1 represents the cement strength, and δ represents the standard deviation of the compressive strength of the concrete cube; The grading judgment model is composed of a data input module, a data standardization processing module, a judgment module and a data output module; Wherein, the data input module is used to input the standard value of cube compressive strength; The data standardization processing module is used to perform format conversion processing on the input cube compressive strength standard value; The determination module is used to determine the standard value data of the cube compressive strength after the format conversion process, and obtain the corresponding concrete grade; Step S6: Acquire mechanical parameter data of the concrete sample, and use the mechanical parameter data and the concrete mechanical parameter detection model to perform assignment operations to obtain concrete mechanical parameters, and analyze the obtained concrete mechanical parameters to complete concrete mechanical parameter detection and analysis.

2. A method for detecting and analyzing concrete mechanical parameters according to claim 1, characterized in that: The confidence parameters include: depth map hole filling confidence, depth map super-resolution confidence and depth map correction confidence; The method for solving the depth map hole filling confidence is: First, the depth value of the depth map is obtained, and then the confidence of the missing area in the depth map is set to 0, and the confidence of the area in the depth map that is not missing and has a depth value is set to 1. All areas in the depth map are marked in the above manner, so as to obtain the confidence matrix of the depth map hole filling; The method for solving the super-resolution confidence of the depth map is: First, the RGB standard image is input, and the depth map is upsampled under the guidance of the RGB standard image. Then, the depth map is interpolated using the bicubic interpolation algorithm to make its resolution the same as that of the RGB standard image. Finally, the Gaussian function envelope is used to obtain the super-resolution confidence of the depth map. The method for solving the depth map correction confidence is: First, the original depth map is processed by a transform filter to obtain a processed depth map. Then, the difference between the original depth map and the processed depth map is divided by 2 times the square of the bandwidth to obtain the depth map correction confidence.

3. A method for detecting and analyzing concrete mechanical parameters according to claim 1, characterized in that: In step S3, the software Mathpix is ​​used to perform mathematical modeling on the original structured data and the super-resolution original image.

4. A method for detecting and analyzing concrete mechanical parameters according to claim 1, characterized in that: The loss function is a cross entropy loss function; the network training parameters include the number of training times, the number of tests and the number of training rounds; and the optimizer is an SGD stochastic gradient descent optimizer.

5. A method for detecting and analyzing concrete mechanical parameters according to claim 1, characterized in that: The value range of the coefficient a is: 0.3-0.4; the value range of the coefficient b is: 0.3-0.4; the value range of the coefficient c is: 0.7-0.8; the value range of the coefficient d is: 0.3-0.4; the value range of the coefficient e is: 1.3-1.5.

Images