Coarse aggregate ultra-high performance concrete strength prediction method and related device

Abstract

The application provides a coarse aggregate ultra-high performance concrete strength prediction method and related equipment. The method comprises the following steps: acquiring first related information of coarse aggregate and second related information of paste in a first cross-section image of ultra-high performance concrete based on an aggregate identification model, wherein the first related information of the coarse aggregate comprises particle size, area, orientation distribution, average distance between aggregates, grading, density, maximum thickness of the paste layer, crushing value and filling rate, the second related information of the paste comprises water-cement ratio and viscosity, and the ultra-high performance concrete contains coarse aggregate; and obtaining the strength of the ultra-high performance concrete corresponding to the first cross-section image through a performance prediction model based on the first related information and the second related information. Through the application, efficient, accurate and intelligent prediction of the strength of the coarse aggregate ultra-high performance concrete is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of concrete strength prediction, and particularly relates to a coarse aggregate ultra-high performance concrete strength prediction method and device. BACKGROUND

[0002] Ultra-high performance concrete is a kind of ultra-high strength cement-based material with high strength, high durability and low porosity. The coarse aggregate ultra-high performance concrete is prepared by adding coarse aggregate in the ultra-high performance concrete, which can reduce the cement dosage, improve the elastic modulus of the concrete, reduce the early shrinkage and increase the application range of the ultra-high performance concrete under the condition of minimizing the loss of mechanical properties. The compressive strength is the most important index parameter of the coarse aggregate ultra-high performance concrete and is an important basis for structural design and construction. Therefore, detecting the compressive strength of the coarse aggregate ultra-high performance concrete is the core content of ensuring the quality of the concrete.

[0003] At present, when the strength of the coarse aggregate ultra-high performance concrete is detected, it is necessary to first carry out proportioning experiments in the laboratory for 28 days of curing age, and then detect according to the standard, which leads to a long period for obtaining accurate detection results. Moreover, only the "strength detection" can be completed, and the "strength prediction" function in a short period cannot be realized. The strength prediction of the coarse aggregate ultra-high performance concrete can greatly reduce the detection workload, improve the efficiency, save the engineering manpower and time cost, and has a significant application prospect. Therefore, a technical scheme for predicting the strength of the coarse aggregate ultra-high performance concrete is urgently needed. SUMMARY

[0004] The main purpose of the present application is to provide a coarse aggregate ultra-high performance concrete strength prediction method and related device, which aims to realize efficient, accurate and intelligent prediction of the strength of the coarse aggregate ultra-high performance concrete.

[0005] In a first aspect, the present application provides a coarse aggregate ultra-high performance concrete strength prediction method, which comprises:

[0006] obtaining first related information of coarse aggregate and second related information of paste in a first cross-section image of the ultra-high performance concrete based on an aggregate recognition model, wherein the first related information of the coarse aggregate comprises particle size, area, orientation distribution, average distance between aggregates, gradation, density, maximum thickness of paste layer, crushing value and filling rate of the coarse aggregate, and the second related information of the paste comprises water-cement ratio and viscosity of the paste, and the ultra-high performance concrete contains coarse aggregate;

[0007] obtaining the strength of the ultra-high performance concrete corresponding to the first cross-section image through a performance prediction model based on the first related information and the second related information.

[0008] Optionally, before the step of obtaining the first related information of the coarse aggregate and the second related information of the paste of the ultra-high performance concrete section image based on the aggregate identification model, the method comprises the following steps of:

[0009] obtaining a plurality of second section images of the ultra-high performance concrete in the training set and the gradation, density, crushing value of the coarse aggregate in each second section image, the water-cement ratio of the paste, the viscosity of the paste and the actual strength of the ultra-high performance concrete corresponding to each second section image;

[0010] respectively pre-processing each second section image to obtain each pre-processed second section image;

[0011] annotating the coarse aggregate in each pre-processed second section image through the annotation software, training the deep learning model based on each annotated second section image to obtain the aggregate identification model;

[0012] obtaining the third related information of the coarse aggregate in each second section image through the aggregate identification model based on each pre-processed second section image, wherein the third related information of the coarse aggregate comprises the particle size, area, orientation distribution, maximum thickness of the paste layer, average distance between aggregates and filling rate of the coarse aggregate;

[0013] normalizing the fourth related information corresponding to each pre-processed second section image to obtain the normalized fourth related information corresponding to each pre-processed second section image, wherein the fourth related information comprises the water-cement ratio of the paste, the viscosity of the paste, the third related information of the coarse aggregate, the gradation, the density, the crushing value and the actual strength of the ultra-high performance concrete corresponding to each second section image;

[0014] training the deep learning model based on the normalized fourth related information to obtain the performance prediction model.

[0015] Optionally, the step of normalizing the fourth related information corresponding to each pre-processed second section image to obtain the normalized fourth related information corresponding to each pre-processed second section image comprises the following steps of:

[0016] substituting the gradation of the coarse aggregate corresponding to the i-th pre-processed second section image and the average value and variance of the gradation of the coarse aggregate into the normalization processing calculation formula to calculate the normalized gradation of the coarse aggregate corresponding to each i-th pre-processed second section image, wherein the normalization processing calculation formula is as follows:

[0017]

[0018] wherein x irepresents the normalized gradation of coarse aggregate corresponding to the i-th preprocessed second cross-section image, x i represents the gradation of coarse aggregate corresponding to the i-th preprocessed second cross-section image, x represents the average value corresponding to the gradation of coarse aggregate, s represents the variance corresponding to the gradation of coarse aggregate, and so on. The normalized fourth related information corresponding to each preprocessed second cross-section image is calculated in this way.

[0019] Optionally, after the step of training the deep learning model based on the normalized fourth related information to obtain the performance prediction model, the following steps are included:

[0020] Based on the third cross-section image of each ultra-high performance concrete in the test set, the predicted strength of each ultra-high performance concrete corresponding to each third cross-section image is obtained through the performance prediction model;

[0021] The predicted strength of each ultra-high performance concrete is respectively corrected based on the correction coefficient to obtain the corrected predicted strength of each ultra-high performance concrete;

[0022] Based on the corrected predicted strength of each ultra-high performance concrete and the actual strength of the corresponding ultra-high performance concrete, the mean square error and the correlation coefficient of the performance prediction model are calculated.

[0023] Optionally, the step of obtaining the first related information of coarse aggregate and the second related information of paste in the first cross-section image of ultra-high performance concrete based on the aggregate recognition model includes:

[0024] The gradation, density, crushing value, packing density of coarse aggregate, water-cement ratio of paste, viscosity of paste, and volume fraction of coarse aggregate in unit volume of ultra-high performance concrete in the first cross-section image of ultra-high performance concrete containing coarse aggregate are obtained;

[0025] Based on the aggregate recognition model, the maximum coarse aggregate particle size, coarse aggregate particle size, orientation distribution, and average distance between aggregates in the first cross-section image are calculated through a polygon fitting algorithm, and the actual coarse aggregate area and the cross-section area of the concrete in the first cross-section image are calculated through a pixel algorithm;

[0026] The packing density of coarse aggregate in the first cross-section image, the volume fraction of coarse aggregate in unit volume of ultra-high performance concrete, and the maximum coarse aggregate particle size in the first cross-section image are substituted into the first preset formula to calculate the maximum thickness of the paste layer in the first cross-section image, wherein the first preset formula is as follows:

[0027]

[0028] wherein MPT is the maximum paste thickness of the paste layer in the first cross-section image of ultra-high performance concrete, φmax is the packing density of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, φ is the volume fraction of the coarse aggregate in the unit volume of the concrete in the first cross-sectional image of the ultra-high performance concrete, D max is the maximum coarse aggregate particle size in the first cross-sectional image of the ultra-high performance concrete.

[0029] The area actually occupied by the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete and the cross-sectional area of the ultra-high performance concrete are substituted into the second preset formula to calculate the filling rate of the coarse aggregate, wherein the second preset formula is as follows:

[0030]

[0031] wherein F represents the filling rate of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, M1 represents the area actually occupied by the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, and M2 represents the cross-sectional area of the ultra-high performance concrete.

[0032] In a second aspect, the present application further provides a coarse aggregate ultra-high performance concrete strength prediction device, the concrete strength prediction device comprising:

[0033] an information acquisition module configured to acquire a first cross-sectional image of an ultra-high performance concrete in a test set, and acquire first related information of coarse aggregate and second related information of paste in the first cross-sectional image based on an aggregate recognition model, wherein the first related information of the coarse aggregate comprises particle size, area, orientation distribution, average distance between aggregates, grading, density, maximum thickness of paste layer, crushing value and filling rate of the coarse aggregate, and the second related information of the paste comprises water-cement ratio and viscosity of the paste, and the ultra-high performance concrete comprises the coarse aggregate;

[0034] a strength prediction module configured to obtain the strength of the ultra-high performance concrete corresponding to the first cross-sectional image by a performance prediction model based on the first related information and the second related information.

[0035] Optionally, the information acquisition module is specifically configured to:

[0036] acquire the grading, density, crushing value, packing density of the coarse aggregate, water-cement ratio of the paste, viscosity of the paste and volume fraction of the coarse aggregate in the unit volume of the ultra-high performance concrete in the first cross-sectional image of the ultra-high performance concrete containing the coarse aggregate;

[0037] the maximum coarse aggregate particle size, the particle size of the coarse aggregate, the orientation distribution and the average distance between aggregates in the first cross-sectional image are calculated by a polygon fitting algorithm based on the aggregate recognition model, and the area actually occupied by the coarse aggregate in the first cross-sectional image and the cross-sectional area of the concrete are calculated by a pixel unit algorithm;

[0038] The packing density of the coarse aggregate in the first cross-sectional image, the volume fraction of the coarse aggregate in the unit volume of the ultra-high performance concrete, and the maximum coarse aggregate particle size in the first cross-sectional image are substituted into a first preset formula to calculate the maximum thickness of the paste layer in the first cross-sectional image, wherein the first preset formula is as follows:

[0039]

[0040] wherein MPT is the maximum paste thickness of the paste layer in the first cross-sectional image of the ultra-high performance concrete, φ is the packing density of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, φ is the volume fraction of the coarse aggregate in the unit volume of the ultra-high performance concrete, and D is the maximum coarse aggregate particle size in the first cross-sectional image of the ultra-high performance concrete. max max

[0041] The area actually occupied by the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete and the cross-sectional area of the ultra-high performance concrete are substituted into a second preset formula to calculate the filling rate of the coarse aggregate, wherein the second preset formula is as follows:

[0042]

[0043] wherein F represents the filling rate of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, M1 represents the area actually occupied by the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, and M2 represents the cross-sectional area of the ultra-high performance concrete.

[0044] Optionally, the coarse aggregate ultra-high performance concrete strength prediction device further comprises a calculation module configured to:

[0045] Based on the third cross-sectional image of the ultra-high performance concrete in each test set, the performance prediction model is used to obtain the predicted strength of each ultra-high performance concrete corresponding to each third cross-sectional image.

[0046] Based on the correction coefficient, the predicted strength of each ultra-high performance concrete is corrected to obtain the predicted strength of each corrected ultra-high performance concrete.

[0047] Based on the predicted strength of each corrected ultra-high performance concrete and the actual strength of the corresponding ultra-high performance concrete, the mean square error and the correlation coefficient of the performance prediction model are calculated.

[0048] ​​In a third aspect, the present application further provides a coarse aggregate ultra-high performance concrete strength prediction device, comprising a processor, a memory, and a coarse aggregate ultra-high performance concrete strength prediction program stored in the memory and executable by the processor, wherein the coarse aggregate ultra-high performance concrete strength prediction program, when executed by the processor, implements the steps of the coarse aggregate ultra-high performance concrete strength prediction method as described above.

[0049] In a fourth aspect, the present application further provides a readable storage medium having a coarse aggregate ultra-high performance concrete strength prediction program stored thereon, wherein the coarse aggregate ultra-high performance concrete strength prediction program, when executed by a processor, implements the steps of the coarse aggregate ultra-high performance concrete strength prediction method as described above.

[0050] In the present application, the first related information of the coarse aggregate and the second related information of the paste in the first cross-sectional image of the ultra-high performance concrete are obtained based on the aggregate recognition model, wherein the first related information of the coarse aggregate includes the particle size, area, orientation distribution, average distance between aggregates, gradation, density, maximum thickness of the paste layer, crushing value, and filling rate of the coarse aggregate, and the second related information of the paste includes the water-cement ratio and viscosity of the paste, and the ultra-high performance concrete contains coarse aggregate; and the strength of the ultra-high performance concrete corresponding to the first cross-sectional image is obtained through the performance prediction model based on the first related information and the second related information. Through the present application, the strength of the concrete to be detected can be obtained through the aggregate recognition model and the performance prediction model based on the key characteristic parameters of the coarse aggregate and the key characteristic parameters of the paste in the coarse aggregate ultra-high performance concrete, without the need for curing and then detecting according to the standard, thereby realizing efficient, accurate, and intelligent prediction of the strength of the coarse aggregate ultra-high performance concrete, reducing the workload of the current detection, improving the efficiency, and reducing the labor and time costs. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 It is a flowchart of the first embodiment of the coarse aggregate ultra-high performance concrete strength prediction method of the present application;

[0052] Figure 2a It is a cross-sectional image of the coarse aggregate ultra-high performance concrete of the present application;

[0053] Figure 2b It is a cross-sectional image of the coarse aggregate ultra-high performance concrete after preprocessing of the present application;

[0054] Figure 3 It is a deep learning network diagram of the performance prediction model of the present application;

[0055] Figure 4 It is a flowchart of the second embodiment of the coarse aggregate ultra-high performance concrete strength prediction method of the present application;

[0056] Figure 5 This is a schematic diagram of the functional modules of an embodiment of the coarse aggregate ultra-high performance concrete strength prediction device of the present invention.

[0057] Figure 6 This is a schematic diagram of the hardware structure of the coarse aggregate ultra-high performance concrete strength prediction device involved in the embodiment of the present invention.

[0058] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0059] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0060] In a first aspect, embodiments of the present invention provide a method for predicting the strength of coarse aggregate ultra-high performance concrete.

[0061] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the method for predicting the strength of ultra-high performance concrete with coarse aggregate according to the present invention. Figure 1 As shown, the method for predicting the strength of ultra-high performance concrete with coarse aggregate includes:

[0062] Step S10: Based on the aggregate identification model, obtain the first relevant information of coarse aggregate and the second relevant information of paste in the first cross-sectional image of ultra-high performance concrete. The first relevant information of coarse aggregate includes the particle size, area, orientation distribution, average distance between aggregates, gradation, density, maximum thickness of paste layer, crushing value and filling rate of coarse aggregate. The second relevant information of paste includes the water-cement ratio and viscosity of paste. Ultra-high performance concrete contains coarse aggregate.

[0063] In this embodiment, refer to Figure 2a , Figure 2a This is a cross-sectional image of the coarse aggregate ultra-high performance concrete of this invention. Figure 2a As shown, a cross-sectional image of the ultra-high performance concrete to be predicted containing coarse aggregate is obtained and denoted as the first cross-sectional image. Based on the aggregate recognition model, the particle size, area, orientation distribution, average distance between aggregates, gradation, density, maximum thickness of the paste layer, crushing value, and filling rate of the coarse aggregate in the first cross-sectional image are obtained, as well as the water-cement ratio and viscosity of the paste.

[0064] Further, in one embodiment, step S10 includes:

[0065] obtaining gradation, density, crushing value, packing density of coarse aggregate, water-cement ratio of paste, viscosity of paste and volume fraction of coarse aggregate in unit volume of the ultra-high performance concrete in the first cross-section image of the ultra-high performance concrete containing coarse aggregate;

[0066] calculating maximum coarse aggregate particle size, particle size, orientation distribution and average distance between aggregates of coarse aggregate in the first cross-section image based on the aggregate recognition model through a polygon fitting algorithm, and calculating area occupied by actual coarse aggregate and cross-sectional area of the concrete in the first cross-section image through a pixel statistical algorithm;

[0067] substituting packing density of coarse aggregate, volume fraction of coarse aggregate in unit volume of the ultra-high performance concrete and maximum coarse aggregate particle size in the first cross-section image into a first preset formula to calculate maximum thickness of the paste layer in the first cross-section image, wherein the first preset formula is as follows:

[0068]

[0069] wherein MPT is maximum paste thickness of the paste layer in the first cross-section image of the ultra-high performance concrete, φ is packing density of coarse aggregate in the first cross-section image of the ultra-high performance concrete, φ is volume fraction of coarse aggregate in unit volume of the concrete in the first cross-section image of the ultra-high performance concrete, and D is maximum coarse aggregate particle size in the first cross-section image of the ultra-high performance concrete. max max

[0070] substituting area occupied by actual coarse aggregate in the first cross-section image of the ultra-high performance concrete and cross-sectional area of the ultra-high performance concrete into a second preset formula to calculate filling rate of coarse aggregate, wherein the second preset formula is as follows:

[0071]

[0072] wherein F represents filling rate of coarse aggregate in the first cross-section image of the ultra-high performance concrete, M1 represents area occupied by actual coarse aggregate in the first cross-section image of the ultra-high performance concrete, and M2 represents cross-sectional area of the ultra-high performance concrete.

[0073] In the embodiment, gradation, density, crushing value, packing density of coarse aggregate, water-cement ratio of paste, viscosity of paste and volume fraction of coarse aggregate in unit volume of the ultra-high performance concrete in the first cross-section image of the ultra-high performance concrete input by the user are obtained.

[0074] ​​The maximum coarse aggregate particle size, the particle size, the orientation distribution and the average distance between aggregates of the coarse aggregate in the first cross-sectional image are calculated by calling a polygon fitting algorithm based on the aggregate identification model, and the actual coarse aggregate area and the cross-sectional area of the concrete in the first cross-sectional image are calculated by calling a pixel counting method.

[0075] The maximum thickness of the paste layer in the first cross-sectional image of the ultra-high performance concrete is calculated by substituting the packing density of the coarse aggregate in the first cross-sectional image, the volume fraction of the coarse aggregate in the unit volume of the ultra-high performance concrete and the maximum coarse aggregate particle size in the first cross-sectional image into the first preset formula, wherein the first preset formula is as follows:

[0076]

[0077] Wherein, MPT is the maximum paste thickness of the paste layer in the first cross-sectional image of the ultra-high performance concrete, φ max is the packing density of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, φ is the volume fraction of the coarse aggregate in the unit volume of the ultra-high performance concrete, D max is the maximum coarse aggregate particle size in the first cross-sectional image of the ultra-high performance concrete.

[0078] The filling rate of the coarse aggregate is calculated by substituting the actual coarse aggregate area in the first cross-sectional image of the ultra-high performance concrete and the cross-sectional area of the ultra-high performance concrete into the second preset formula, wherein the second preset formula is as follows:

[0079]

[0080] Wherein, F represents the filling rate of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, M1 represents the actual coarse aggregate area in the first cross-sectional image of the ultra-high performance concrete, and M2 represents the cross-sectional area of the ultra-high performance concrete.

[0081] In step S20, the strength of the ultra-high performance concrete corresponding to the first cross-sectional image is obtained by the performance prediction model based on the first related information and the second related information.

[0082] In this embodiment, based on the particle size, area, orientation distribution, average distance between aggregates, grading, density, maximum thickness of the paste layer, crushing value and filling rate of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, the water-cement ratio and the viscosity of the paste, the strength of the ultra-high performance concrete corresponding to the first cross-sectional image can be obtained by the performance prediction model.

[0083] It is proved that the prediction of the strength of concrete based on the characteristics of coarse aggregate is technically feasible and reliable. However, the matrix strength of ordinary concrete is generally not high (30-60 MPa), and the matrix is often damaged during the stress failure process, so the prediction method based on the characteristics of coarse aggregate is often less accurate for ordinary concrete. The matrix strength of coarse aggregate ultra-high performance concrete is high (≥100 MPa), and the stress failure usually occurs near the coarse aggregate, so the strength prediction method based on the characteristics of coarse aggregate is reasonable and has high accuracy guarantee for the coarse aggregate ultra-high performance concrete system.

[0084] In the embodiment, the first related information of the coarse aggregate and the second related information of the paste in the first cross-section image of the ultra-high performance concrete are obtained based on the aggregate recognition model, wherein the first related information of the coarse aggregate includes the particle size, area, orientation distribution, average distance between aggregates, gradation, density, maximum thickness of the paste layer, crushing value and filling rate of the coarse aggregate, and the second related information of the paste includes the water-cement ratio and viscosity of the paste, and the ultra-high performance concrete contains coarse aggregate; the strength of the ultra-high performance concrete corresponding to the first cross-section image is obtained by the performance prediction model based on the first related information and the second related information. Through the embodiment, the strength of the concrete to be detected can be obtained by the aggregate recognition model and the performance prediction model based on the key characteristic parameters of the coarse aggregate and the key characteristic parameters of the paste in the coarse aggregate ultra-high performance concrete, without the need for curing and then detecting according to the standard, realizing efficient, accurate and intelligent prediction of the strength of the coarse aggregate ultra-high performance concrete, reducing the workload of the current detection, improving the efficiency, and reducing the labor and time cost.

[0085] Further, in an embodiment, referring to Figure 4 , Figure 4 The flowchart of the second embodiment of the strength prediction method of the coarse aggregate ultra-high performance concrete of the present application is shown in FIG. 2. As shown in FIG. 2, before step S10, the following steps are included: Figure 4

[0086] Step S001, a plurality of second cross-section images of ultra-high performance concrete and the gradation, density, crushing value of the coarse aggregate in each second cross-section image, the water-cement ratio of the paste, the viscosity of the paste and the actual strength of the ultra-high performance concrete corresponding to each second cross-section image in the training set are obtained;

[0087] Step S002, each second cross-section image is preprocessed respectively to obtain each preprocessed second cross-section image;

[0088] Step S003, the coarse aggregate in each preprocessed second cross-section image is labeled by a labeling software, and a deep learning model is trained based on each labeled second cross-section image to obtain an aggregate recognition model; ​

[0089] Step S004, based on each pre-processed second cross-section image, the third related information of coarse aggregate in each second cross-section image is obtained through the aggregate recognition model, wherein the third related information of coarse aggregate includes the particle size, area, orientation distribution, maximum thickness of paste layer, average distance between aggregates and filling rate of coarse aggregate;

[0090] Step S005, the fourth related information corresponding to each pre-processed second cross-section image is normalized to obtain the fourth related information after normalization corresponding to each pre-processed second cross-section image, wherein the fourth related information includes the water-cement ratio of paste, the viscosity of paste, the third related information of coarse aggregate, the gradation, the density, the crushing value and the actual strength of ultra-high performance concrete corresponding to each second cross-section image;

[0091] Step S006, the deep learning model is trained based on the fourth related information after normalization to obtain the performance prediction model.

[0092] In this embodiment, the cross-section images of a plurality of ultra-high performance concrete test blocks are obtained, and the cross-section images of the plurality of ultra-high performance concrete test blocks are respectively stored in the training set and the test set according to a preset ratio, such as a ratio of 8:2. A plurality of ultra-high performance concrete cross-section images are obtained from the training set, denoted as second cross-section images. Then, the gradation, density, crushing value of coarse aggregate in each second cross-section image, the water-cement ratio of paste, the viscosity of paste and the actual strength of ultra-high performance concrete corresponding to each second cross-section image are obtained based on user input.

[0093] Reference Figure 2b , Figure 2b is the pre-processed cross-section image of coarse aggregate ultra-high performance concrete of the present application. As shown in Figure 2b , each second cross-section image is pre-processed to obtain each pre-processed second cross-section image, thereby simplifying the cross-section image information, wherein the pre-processing includes cutting, denoising, binarization and selection.

[0094] Each coarse aggregate in the pre-processed second cross-section image is labeled by a labeling software, and a deep learning model is trained based on each labeled second cross-section image to obtain an aggregate recognition model. The deep learning training model is a convolutional neural network training model. The aggregate recognition model includes 1 input layer, 13 convolutional layers, 3 fully connected layers and 1 output layer, each neural network is constructed on multiple planes combined by independently distributed neurons, the connection mode between layers is non-full connection convolution calculation, each neuron is the weighted sum of part of the dimension values of the input unit, and the convolutional layer and the input acceptance area need to perform the following related operations:

[0095]

[0096] wherein x is a two-dimensional vector of the receptive field (M, N), ω is a convolution kernel with length j and width i, b is a bias term added to each output feature map, y conv is the output of the convolution, M is the length of the two-dimensional vector, N is the width of the two-dimensional vector, and f is an activation function.

[0097] The loss function of the aggregate recognition model is a perception loss function, and the formula is as follows:

[0098] wherein, y respectively represent the generated image and the source image, represents a neural network used, j represents the jth layer of the neural network, C j H j W j is the shape of the jth layer feature map.

[0099] The activation unit of all hidden layers of the aggregate recognition model adopts a ReLU function, and the formula is as follows:

[0100] f(x) = max(0, x).

[0101] Based on each preprocessed second cross-section image, the aggregate recognition model calls a polygon fitting algorithm to calculate the particle size, orientation distribution and average distance between aggregates in the second cross-section image, calculates the maximum thickness of the paste layer through a preset formula, and calculates the filling rate through a second preset formula.

[0102] The fourth relevant information corresponding to each preprocessed second cross-section image is normalized to obtain normalized fourth relevant information corresponding to each preprocessed second cross-section image, wherein the fourth relevant information includes the water-cement ratio of the paste, the viscosity of the paste, the third relevant information of the coarse aggregate, the gradation, the density, the crushing value, and the actual strength of the high-performance concrete corresponding to each second cross-section image.

[0103] Based on the normalized fourth relevant information, a deep learning model is trained to obtain a performance prediction model. Figure 3 , Figure 3 is a deep learning network diagram of the performance prediction model of the application. As Figure 3As shown, the performance prediction model includes an input layer, a convolutional layer, two fully connected layers, and a traditional output layer. The input layer is a 3*3 matrix composed of nine selected parameters (water-cement ratio of the paste, viscosity of the paste, orientation distribution of the coarse aggregate, average distance between coarse aggregates, grading of the coarse aggregate, crushing value of the coarse aggregate, density of the coarse aggregate, maximum thickness of the paste layer, and coarse aggregate filling rate). In the convolutional layer, 128 convolution kernels (k) and 128 biases (b) are set, and the activation function is selected as the Sigmoid function, as shown in the following formula:

[0104] After the nine selected parameters pass through the 3*3*128 convolutional layer, they become a 1*1*128 matrix, then pass through two fully connected layers with 64 neurons, and then pass through the activation function Sigmoid function, and enter the output layer with only one neuron. The output layer outputs the strength of the concrete.

[0105] The loss function of the performance prediction model uses the squared error loss function, as shown in the following formula:

[0106] where N is the number of preprocessed second cross-sectional images, c is the number of parameters, is the kth dimension of the nth preprocessed second cross-sectional image associated with the label k, is the value of the kth output layer unit responding to the nth preprocessed concrete cross-sectional image.

[0107] Further, in an embodiment, the step of normalizing the fourth related information corresponding to each preprocessed second cross-sectional image to obtain normalized fourth related information corresponding to each preprocessed second cross-sectional image comprises:

[0108] The grading of the coarse aggregate corresponding to the ith preprocessed second cross-sectional image and the average value and variance of the grading of the coarse aggregate are substituted into the normalization calculation formula to obtain the normalized grading of the coarse aggregate corresponding to the ith preprocessed second cross-sectional image, wherein the normalization calculation formula is as follows:

[0109]

[0110] where x i represents the normalized grading of the coarse aggregate corresponding to the ith preprocessed second cross-sectional image, x i represents the grading of the coarse aggregate corresponding to the ith preprocessed second cross-sectional image, The average value corresponding to the gradation of coarse aggregate is represented by s, the variance corresponding to the gradation of coarse aggregate is represented by s, and so on, so as to calculate the normalized fourth related information corresponding to each preprocessed second cross-sectional image.

[0111] In the embodiment, when the gradation of coarse aggregate corresponding to the i-th preprocessed second cross-sectional image is normalized, the average value and the variance corresponding to the gradation of coarse aggregate are calculated based on the gradation of coarse aggregate in each second cross-sectional image;

[0112] Then, the gradation of coarse aggregate corresponding to the i-th preprocessed second cross-sectional image and the average value and the variance corresponding to the gradation of coarse aggregate are substituted into the normalization calculation formula to calculate the normalized gradation of coarse aggregate corresponding to each i-th preprocessed second cross-sectional image, wherein the normalization calculation formula is as follows:

[0113]

[0114] wherein x i ′ represents the normalized gradation of coarse aggregate corresponding to the i-th preprocessed second cross-sectional image, x i represents the gradation of coarse aggregate corresponding to the i-th preprocessed second cross-sectional image, The average value corresponding to the gradation of coarse aggregate is represented by s, the variance corresponding to the gradation of coarse aggregate is represented by s, and so on, so as to calculate the normalized fourth related information corresponding to each preprocessed second cross-sectional image.

[0115] By analogy, the water-cement ratio of the paste, the viscosity of the paste, the third related information of the coarse aggregate, the density, the crushing value of the i-th preprocessed second cross-sectional image and the actual strength of the ultra-high performance concrete corresponding to each second cross-sectional image are normalized by the normalization calculation formula, so as to calculate the normalized fourth related information corresponding to the i-th preprocessed second cross-sectional image. By analogy again, the normalized fourth related information corresponding to each preprocessed second cross-sectional image can be obtained.

[0116] Further, in an embodiment, after the step of training the deep learning model based on the normalized fourth related information to obtain the performance prediction model, the method further comprises:

[0117] Based on the third cross-sectional image of the ultra-high performance concrete in each test set, the predicted strength of each ultra-high performance concrete corresponding to each third cross-sectional image is obtained by the performance prediction model;

[0118] The predicted strength of each ultra-high performance concrete is corrected based on the correction coefficient to obtain the corrected predicted strength of each ultra-high performance concrete.

[0119] The mean square error and the correlation coefficient of the performance prediction model are calculated based on the predicted strength of each modified ultra-high performance concrete and the actual strength of the corresponding ultra-high performance concrete.

[0120] In this embodiment, a plurality of cross-sectional images of the ultra-high performance concrete are obtained from the test set, denoted as third cross-sectional images. The predicted strength of each ultra-high performance concrete corresponding to each third cross-sectional image is obtained based on the performance prediction model.

[0121] The predicted strength of each ultra-high performance concrete corresponding to each third cross-sectional image is modified by the modification coefficient respectively to obtain the predicted strength of each modified ultra-high performance concrete, wherein the modification processing formula is: x′ 修正 = τ * x 原始 , τ represents the modification coefficient after considering the instrument and other reasons, and the value range is 0.93-0.98, x 原始 represents the predicted strength of the ultra-high performance concrete before modification, and x′ 修正 represents the predicted strength of the ultra-high performance concrete after modification.

[0122] The mean square error and the correlation coefficient of the predicted strength of each modified ultra-high performance concrete and the actual strength of the ultra-high performance concrete are calculated, wherein the calculation formula of the mean square error (MSE) is as follows:

[0123]

[0124] The calculation formula of the correlation coefficient (R 2 ) is as follows:

[0125] Wherein, y_test i represents the actual strength of the i-th ultra-high performance concrete, y_predict i represents the predicted strength of the modified i-th ultra-high performance concrete, and n represents the number of each ultra-high performance concrete corresponding to each third cross-sectional image.

[0126] Specifically, if the number n of each ultra-high performance concrete corresponding to each third cross-sectional image is 40, 40 groups of predicted strength of the ultra-high performance concrete output by the performance prediction model and the actual strength of the ultra-high performance concrete are obtained as shown in Table 1 below:

[0127] Table 1

[0128]

[0129] Based on the predicted strength of the ultra-high performance concrete output by the 40 groups of performance prediction models and the actual strength of the ultra-high performance concrete, the mean square error (MSE) is calculated to be 15.31, and the correlation coefficient (R 2 ) is 0.955.

[0130] 200 groups of sample data are collected, and the traditional neural network, the ordinary deep learning model and the computer vision and deep learning model are used for simulation training, and then the strength of the coarse aggregate ultra-high performance concrete is predicted, and finally the mean square error (MSE) and the correlation coefficient (R 2 ) are shown in Table 2 as follows:

[0131] Table 2

[0132]

[0133] From Table 2, it can be seen that the model obtained by using the computer vision and deep learning model for simulation training has the highest prediction accuracy for the strength of the coarse aggregate ultra-high performance concrete. Since the aggregate recognition model and the performance prediction model are obtained by using the computer vision and deep learning model for simulation training, the efficient, accurate and intelligent prediction of the strength of the coarse aggregate ultra-high performance concrete is realized.

[0134] In a second aspect, the embodiments of the present application also provide a concrete strength prediction device.

[0135] In an embodiment, referring to Figure 5 , Figure 5 is a functional module schematic diagram of an embodiment of the coarse aggregate ultra-high performance concrete strength prediction device of the present application. As shown in Figure 5 , the concrete strength prediction device comprises:

[0136] The information acquisition module 10 is configured to acquire a first cross-sectional image of the ultra-high performance concrete in the test set, and acquire first related information of coarse aggregate and second related information of paste in the first cross-sectional image based on the aggregate recognition model, wherein the first related information of the coarse aggregate includes particle size, area, orientation distribution, average distance between aggregates, gradation, density, maximum thickness of paste layer, crushing value and filling rate of the coarse aggregate, and the second related information of the paste includes water-cement ratio and viscosity of the paste, and the ultra-high performance concrete contains coarse aggregate.

[0137] The strength prediction module 20 is configured to obtain the strength of the ultra-high performance concrete corresponding to the first cross-sectional image by the performance prediction model based on the first related information and the second related information.

[0138] Further, in an embodiment, the information acquisition module 10 is further configured to:

[0139] obtain a plurality of second cross-section images of the ultra-high performance concrete in the training set, and gradation, density, crushing value of coarse aggregates in each second cross-section image, water-cement ratio of paste, viscosity of the paste, and actual strength of the ultra-high performance concrete corresponding to each second cross-section image;

[0140] The concrete strength prediction device further comprises an image processing module, configured to:

[0141] Preprocess each second cross-section image to obtain a preprocessed second cross-section image;

[0142] The concrete strength prediction device further comprises a training module, configured to:

[0143] Label coarse aggregates in each preprocessed second cross-section image by using a labeling software, train the deep learning model based on each labeled second cross-section image to obtain an aggregate recognition model;

[0144] The information acquisition module 10 is further configured to: based on each preprocessed second cross-section image, obtain third related information of coarse aggregates in each second cross-section image by using the aggregate recognition model, wherein the third related information of the coarse aggregates includes particle size, area, orientation distribution, maximum thickness of paste layer, average distance between aggregates, and filling rate of the coarse aggregates;

[0145] The image processing module is further configured to: normalize fourth related information corresponding to each preprocessed second cross-section image to obtain normalized fourth related information corresponding to each preprocessed second cross-section image, wherein the fourth related information includes water-cement ratio of paste, viscosity of the paste, third related information of coarse aggregates, gradation, density, crushing value of the coarse aggregates, and actual strength of the ultra-high performance concrete corresponding to each second cross-section image;

[0146] The training module is further configured to: train the deep learning model based on the normalized fourth related information to obtain a performance prediction model.

[0147] Further, in an embodiment, the image processing module is specifically configured to:

[0148] Substitute the gradation of coarse aggregates corresponding to the i-th preprocessed second cross-section image and the average value and variance of the gradation of coarse aggregates into the normalization calculation formula to calculate the normalized gradation of coarse aggregates corresponding to each i-th preprocessed second cross-section image, wherein the normalization calculation formula is as follows:

[0149]

[0150] wherein x irepresents the gradation of the coarse aggregate corresponding to the i-th preprocessed second cross-section image, x i represents the gradation of the coarse aggregate corresponding to the i-th preprocessed second cross-section image, represents the average value corresponding to the gradation of the coarse aggregate, s represents the variance corresponding to the gradation of the coarse aggregate, and so on, so as to calculate the normalized fourth related information corresponding to each preprocessed second cross-section image.

[0151] Further, in an embodiment, the coarse aggregate ultra-high performance concrete strength prediction device further comprises a calculation module configured to:

[0152] obtain the predicted strength of each ultra-high performance concrete corresponding to each third cross-section image of the test set based on the third cross-section image of each test set through the performance prediction model;

[0153] correct the predicted strength of each ultra-high performance concrete based on the correction coefficient respectively, to obtain the predicted strength of each corrected ultra-high performance concrete;

[0154] calculate the mean square error and the correlation coefficient of the performance prediction model based on the predicted strength of each corrected ultra-high performance concrete and the actual strength of the corresponding ultra-high performance concrete.

[0155] Further, in an embodiment, the information acquisition module 10 is specifically configured to:

[0156] obtain the gradation, density, crushing value, packing density of the coarse aggregate in the first cross-section image of the ultra-high performance concrete containing the coarse aggregate, the water-cement ratio of the paste, the viscosity of the paste, and the volume fraction of the coarse aggregate in the unit volume of the ultra-high performance concrete;

[0157] obtain the maximum coarse aggregate particle size, the particle size, the orientation distribution, and the average distance between aggregates in the first cross-section image based on the aggregate recognition model through the polygon fitting algorithm, and obtain the area occupied by the actual coarse aggregate and the cross-sectional area of the concrete in the first cross-section image through the pixel algorithm;

[0158] substitute the packing density of the coarse aggregate in the first cross-section image, the volume fraction of the coarse aggregate in the unit volume of the ultra-high performance concrete, and the maximum coarse aggregate particle size in the first cross-section image into the first preset formula to calculate the maximum thickness of the paste layer in the first cross-section image, wherein the first preset formula is as follows:

[0159]

[0160] wherein MPT is the maximum paste thickness of the paste layer in the first cross-section image of the ultra-high performance concrete, φ maxis the packing density of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, is the volume fraction of the coarse aggregate in the unit volume of the concrete in the first cross-sectional image of the ultra-high performance concrete, D max is the maximum coarse aggregate particle size in the first cross-sectional image of the ultra-high performance concrete;

[0161] The area actually occupied by the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete and the cross-sectional area of the ultra-high performance concrete are substituted into the second preset formula to calculate the filling rate of the coarse aggregate, wherein the second preset formula is as follows:

[0162]

[0163] wherein F represents the filling rate of the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, M1 represents the area actually occupied by the coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, and M2 represents the cross-sectional area of the ultra-high performance concrete.

[0164] The functions of each module in the coarse aggregate ultra-high performance concrete strength prediction device correspond to the steps in the coarse aggregate ultra-high performance concrete strength prediction method, and the functions and implementation processes will not be repeated here.

[0165] In a third aspect, an embodiment of the present application provides a coarse aggregate ultra-high performance concrete strength prediction device.

[0166] Referring to Figure 6 , Figure 6 is a hardware structure schematic diagram of the coarse aggregate ultra-high performance concrete strength prediction device involved in the embodiment of the present application. In the embodiment of the present application, the coarse aggregate ultra-high performance concrete strength prediction device can include a processor 1001 (for example, a central processing unit, CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to realize the connection and communication between these components; the user interface 1003 can include a display screen and an input unit such as a keyboard; the network interface 1004 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface); the memory 1005 can be a high-speed random access memory (RAM) and can also be a stable memory (non-volatile memory) such as a disk memory; the memory 1005 can optionally be a storage device independent of the aforementioned processor 1001. Those skilled in the art can understand that Figure 6The hardware structure shown in the foregoing figures does not constitute a limitation on the present application, and can include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0167] With reference to the foregoing description of the embodiments, those skilled in the art can clearly understand that the foregoing embodiment methods can be implemented by means of software plus a general-purpose hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium such as a ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a terminal device to execute the methods described in the various embodiments of the present application. Figure 6 Figure 6 With reference to the foregoing description of the embodiments, those skilled in the art can clearly understand that the foregoing embodiment methods can be implemented by means of software plus a general-purpose hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium such as a ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a terminal device to execute the methods described in the various embodiments of the present application.

[0168] In a fourth aspect, the embodiments of the present application further provide a readable storage medium.

[0169] The readable storage medium of the present application stores a coarse aggregate ultra-high performance concrete strength prediction program therein, wherein the coarse aggregate ultra-high performance concrete strength prediction program, when executed by a processor, implements the steps of the coarse aggregate ultra-high performance concrete strength prediction method described above.

[0170] The method implemented when the coarse aggregate ultra-high performance concrete strength prediction program is executed can refer to the various embodiments of the coarse aggregate ultra-high performance concrete strength prediction method of the present application, and will not be described here again.

[0171] It should be noted that, in this document, the terms "comprising", "including", or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that includes a list of elements does not only include those elements, but can also include other elements not expressly listed, or inherent to such process, method, article, or system. Without more limitations, an element defined by the statement "comprising a" does not exclude the existence of additional identical elements in the process, method, article, or system that includes the element.

[0172] The above-mentioned embodiment numbers of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0173] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a general-purpose hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium (such as a ROM / RAM, a magnetic disk, or an optical disk) as described above, and includes a number of instructions for causing a terminal device to execute the methods described in the various embodiments of the present application.​

[0174] The above merely provides the preferred embodiments of the present application, and is not intended to limit the patent scope of the present application. Any equivalent structure or equivalent flowchart transformation, or direct or indirect application in other related technical fields, which is made based on the contents of the present application specification and drawings, shall be included in the patent protection scope of the present application.

Claims

1. A method for predicting the strength of ultra-high performance concrete with coarse aggregate, characterized in that, The concrete strength prediction method includes: Based on the aggregate identification model, the first relevant information of coarse aggregate and the second relevant information of paste in the first cross-sectional image of ultra-high performance concrete are obtained. The first relevant information of coarse aggregate includes the particle size, area, orientation distribution, average distance between aggregates, gradation, density, maximum thickness of paste layer, crushing value and filling rate of coarse aggregate. The second relevant information of paste includes the water-cement ratio and viscosity of paste. Ultra-high performance concrete contains coarse aggregate. Based on the first and second relevant information, the strength of the ultra-high performance concrete corresponding to the first cross-sectional image is obtained through the performance prediction model; Before the step of obtaining the first relevant information of coarse aggregate and the second relevant information of paste in the cross-sectional image of ultra-high performance concrete based on the aggregate recognition model, the following steps are included: Acquire multiple second-section images of ultra-high performance concrete from the training set, as well as the gradation, density, crushing value of coarse aggregate, water-cement ratio of paste, viscosity of paste, and actual strength of ultra-high performance concrete corresponding to each second-section image in each second-section image; Each second cross-sectional image is preprocessed to obtain a preprocessed second cross-sectional image; The coarse aggregate in each preprocessed second cross-sectional image is labeled using annotation software. A deep learning model is then trained based on each labeled second cross-sectional image to obtain an aggregate recognition model. Based on each preprocessed second cross-sectional image, the third relevant information of coarse aggregate in each second cross-sectional image is obtained through the aggregate recognition model. The third relevant information of coarse aggregate includes the particle size, area, orientation distribution, maximum thickness of the slurry layer, average distance between aggregates, and filling rate of coarse aggregate. The fourth relevant information corresponding to each preprocessed second cross-sectional image is normalized to obtain the normalized fourth relevant information corresponding to each preprocessed second cross-sectional image. The fourth relevant information includes the water-cement ratio of the slurry, the viscosity of the slurry, the third relevant information of the coarse aggregate, the gradation, density, crushing value, and the actual strength of the ultra-high performance concrete corresponding to each second cross-sectional image. The deep learning model is trained based on the normalized fourth relevant information to obtain the performance prediction model.

2. The method for predicting the strength of ultra-high performance concrete with coarse aggregate as described in claim 1, characterized in that, The step of normalizing the fourth relevant information corresponding to each preprocessed second cross-sectional image to obtain the normalized fourth relevant information corresponding to each preprocessed second cross-sectional image includes: Substituting the coarse aggregate gradation corresponding to the i-th preprocessed second cross-sectional image, along with the mean and variance of the coarse aggregate gradation, into the normalization calculation formula, the normalized coarse aggregate gradation corresponding to each i-th preprocessed second cross-sectional image is calculated. The normalization calculation formula is as follows: in, This represents the normalized coarse aggregate gradation corresponding to the i-th preprocessed second cross-sectional image. This represents the coarse aggregate gradation corresponding to the i-th preprocessed second cross-sectional image. s represents the average value corresponding to the gradation of coarse aggregate, s represents the variance corresponding to the gradation of coarse aggregate, and so on, to calculate the normalized fourth relevant information corresponding to each preprocessed second cross-sectional image.

3. The method for predicting the strength of ultra-high performance concrete with coarse aggregate as described in claim 1, characterized in that, After the step of training the deep learning model based on the normalized fourth relevant information to obtain the performance prediction model, the following steps are included: Based on the third cross-sectional image of ultra-high performance concrete in each test set, the predicted strength of each ultra-high performance concrete corresponding to each third cross-sectional image is obtained through the performance prediction model. The predicted strength of each ultra-high performance concrete is corrected based on the correction factor to obtain the corrected predicted strength of each ultra-high performance concrete. The mean square error and correlation coefficient of the performance prediction model are calculated based on the predicted strength of each corrected ultra-high performance concrete and the corresponding actual strength of the ultra-high performance concrete.

4. The method for predicting the strength of ultra-high performance concrete with coarse aggregate as described in claim 1, characterized in that, The steps for obtaining the first relevant information of coarse aggregate and the second relevant information of paste in the first cross-sectional image of ultra-high performance concrete based on the aggregate recognition model include: The gradation, density, crushing value, bulk density, water-cement ratio, viscosity of the paste, and volume fraction of coarse aggregate in a unit volume of ultra-high performance concrete containing coarse aggregate are obtained from the first cross-sectional image of the ultra-high performance concrete containing coarse aggregate. Based on the aggregate identification model, the maximum coarse aggregate particle size, coarse aggregate particle size, orientation distribution and average distance between aggregates in the first cross-sectional image are calculated by polygon fitting algorithm. The actual area occupied by coarse aggregate and the cross-sectional area of ​​concrete in the first cross-sectional image are calculated by pixel statistics algorithm. The bulk density of coarse aggregate in the first cross-sectional image, the volume fraction of coarse aggregate in the ultra-high performance concrete per unit volume, and the maximum coarse aggregate particle size in the first cross-sectional image are substituted into the first preset formula to calculate the maximum thickness of the slurry layer in the first cross-sectional image. The first preset formula is as follows: Where MPT is the maximum grout thickness in the first cross-sectional image of the ultra-high performance concrete. The packing density of coarse aggregate in the first cross-sectional image of ultra-high performance concrete. This represents the volume fraction of coarse aggregate per unit volume of concrete in the first cross-sectional image of ultra-high performance concrete. The maximum coarse aggregate particle size in the first cross-sectional image of ultra-high performance concrete; Substituting the actual area occupied by coarse aggregate in the first cross-sectional image of ultra-high performance concrete and the cross-sectional area of ​​ultra-high performance concrete into the second preset formula, the coarse aggregate filling rate is calculated. The second preset formula is as follows: Where F represents the coarse aggregate filling rate in the first cross-sectional image of the ultra-high performance concrete, M1 represents the actual area occupied by coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, and M2 represents the cross-sectional area of ​​the ultra-high performance concrete.

5. A device for predicting the strength of ultra-high performance concrete with coarse aggregate, characterized in that, The concrete strength prediction device includes: The information acquisition module is used to acquire a first cross-sectional image of ultra-high performance concrete in the test set, and to acquire first relevant information of coarse aggregate and second relevant information of paste in the first cross-sectional image based on the aggregate recognition model. The first relevant information of coarse aggregate includes the particle size, area, orientation distribution, average distance between aggregates, gradation, density, maximum thickness of paste layer, crushing value and filling rate of coarse aggregate. The second relevant information of paste includes the water-cement ratio and viscosity of paste. The ultra-high performance concrete contains coarse aggregate. The strength prediction module is used to obtain the strength of the ultra-high performance concrete corresponding to the first cross-sectional image based on the first relevant information and the second relevant information through a performance prediction model. Before the step of obtaining the first relevant information of coarse aggregate and the second relevant information of paste in the cross-sectional image of ultra-high performance concrete based on the aggregate recognition model, the following steps are included: Acquire multiple second-section images of ultra-high performance concrete from the training set, as well as the gradation, density, crushing value of coarse aggregate, water-cement ratio of paste, viscosity of paste, and actual strength of ultra-high performance concrete corresponding to each second-section image in each second-section image; Each second cross-sectional image is preprocessed to obtain a preprocessed second cross-sectional image; The coarse aggregate in each preprocessed second cross-sectional image is labeled using annotation software. A deep learning model is then trained based on each labeled second cross-sectional image to obtain an aggregate recognition model. Based on each preprocessed second cross-sectional image, the third relevant information of coarse aggregate in each second cross-sectional image is obtained through the aggregate recognition model. The third relevant information of coarse aggregate includes the particle size, area, orientation distribution, maximum thickness of the slurry layer, average distance between aggregates, and filling rate of coarse aggregate. The fourth relevant information corresponding to each preprocessed second cross-sectional image is normalized to obtain the normalized fourth relevant information corresponding to each preprocessed second cross-sectional image. The fourth relevant information includes the water-cement ratio of the slurry, the viscosity of the slurry, the third relevant information of the coarse aggregate, the gradation, density, crushing value, and the actual strength of the ultra-high performance concrete corresponding to each second cross-sectional image. The deep learning model is trained based on the normalized fourth relevant information to obtain the performance prediction model.

6. The coarse aggregate ultra-high performance concrete strength prediction device as described in claim 5, characterized in that, The information acquisition module is specifically used for: The gradation, density, crushing value, bulk density, water-cement ratio, viscosity of the paste, and volume fraction of coarse aggregate in a unit volume of ultra-high performance concrete containing coarse aggregate are obtained from the first cross-sectional image of the ultra-high performance concrete containing coarse aggregate. Based on the aggregate identification model, the maximum coarse aggregate particle size, coarse aggregate particle size, orientation distribution and average distance between aggregates in the first cross-sectional image are calculated by polygon fitting algorithm. The actual area occupied by coarse aggregate and the cross-sectional area of ​​concrete in the first cross-sectional image are calculated by pixel statistics algorithm. The bulk density of coarse aggregate in the first cross-sectional image, the volume fraction of coarse aggregate in the ultra-high performance concrete per unit volume, and the maximum coarse aggregate particle size in the first cross-sectional image are substituted into the first preset formula to calculate the maximum thickness of the slurry layer in the first cross-sectional image. The first preset formula is as follows: Where MPT is the maximum grout thickness in the first cross-sectional image of the ultra-high performance concrete. The packing density of coarse aggregate in the first cross-sectional image of ultra-high performance concrete. This represents the volume fraction of coarse aggregate per unit volume of concrete in the first cross-sectional image of ultra-high performance concrete. The maximum coarse aggregate particle size in the first cross-sectional image of ultra-high performance concrete; Substituting the actual area occupied by coarse aggregate in the first cross-sectional image of ultra-high performance concrete and the cross-sectional area of ​​ultra-high performance concrete into the second preset formula, the coarse aggregate filling rate is calculated. The second preset formula is as follows: Where F represents the coarse aggregate filling rate in the first cross-sectional image of the ultra-high performance concrete, M1 represents the actual area occupied by coarse aggregate in the first cross-sectional image of the ultra-high performance concrete, and M2 represents the cross-sectional area of ​​the ultra-high performance concrete.

7. The coarse aggregate ultra-high performance concrete strength prediction device as described in claim 5, characterized in that, The coarse aggregate ultra-high performance concrete strength prediction device also includes a calculation module for: Based on the third cross-sectional image of ultra-high performance concrete in each test set, the predicted strength of each ultra-high performance concrete corresponding to each third cross-sectional image is obtained through the performance prediction model. The predicted strength of each ultra-high performance concrete is corrected based on the correction factor to obtain the corrected predicted strength of each ultra-high performance concrete. The mean square error and correlation coefficient of the performance prediction model are calculated based on the predicted strength of each corrected ultra-high performance concrete and the corresponding actual strength of the ultra-high performance concrete.

8. A device for predicting the strength of ultra-high performance concrete with coarse aggregate, characterized in that, The coarse aggregate ultra-high performance concrete strength prediction device includes a processor, a memory, and a coarse aggregate ultra-high performance concrete strength prediction program stored in the memory and executable by the processor, wherein when the coarse aggregate ultra-high performance concrete strength prediction program is executed by the processor, it implements the steps of the coarse aggregate ultra-high performance concrete strength prediction method as described in any one of claims 1 to 4.

9. A readable storage medium, characterized in that, The readable storage medium stores a coarse aggregate ultra-high performance concrete strength prediction program, wherein when the coarse aggregate ultra-high performance concrete strength prediction program is executed by a processor, it implements the steps of the coarse aggregate ultra-high performance concrete strength prediction method as described in any one of claims 1 to 4.

Images