Method and device for predicting elastic property of concrete interface transition area

Abstract

The invention discloses a method and device for predicting the elastic property of a concrete interface transition area, and relates to the technical field of material performance prediction and evaluation. The method comprises the following steps: constructing a data set; input parameters of the data set are macroparameters of the concrete, and output parameters of the data set are elastic parameters of the interface transition area; the macroscopic parameters of the concrete comprise a macroscopic elastic parameter of the concrete, a volume fraction of aggregate, an elastic parameter of the aggregate, an elastic parameter of a matrix and a volume fraction of an interface transition area; constructing a machine learning model; training a machine learning model by using the data set to obtain an interface transition area elastic performance prediction model; inputting the macroscopic parameters of the to-be-tested concrete into the interface transition area elastic performance prediction model, and outputting the elastic parameters of the interface transition area of the to-be-tested concrete; the elastic parameter of the to-be-tested concrete interface transition area represents the elastic property of the to-be-tested concrete interface transition area. According to the method, the elastic property of the concrete interface transition area can be efficiently and accurately predicted.

Description

Technical Field

[0001] The present application relates to the technical field of material performance prediction and evaluation, and in particular to a method and device for predicting the elastic performance of a concrete interface transition zone. Background Art

[0002] Concrete is a multiphase composite material widely used in civil engineering. Its overall mechanical properties depend not only on the characteristics of aggregate and cement matrix, but also on the interfacial transition zone (ITZ). ITZ is a microscopic region between aggregate and matrix. Although its volume fraction is small, it has a significant impact on the macroscopic elastic modulus, strength and durability of concrete due to its complex microstructure (such as high porosity and composition inhomogeneity).

[0003] Traditionally, the performance evaluation of ITZ usually relies on experimental methods, such as microstructure analysis, nanoindentation test and mechanical property test. These experiments can directly measure the micromechanical properties of ITZ, but because ITZ is inside the concrete and difficult to separate independently, the testing process is often complicated and expensive. In addition, such experiments can usually only be carried out on small-scale samples and are difficult to promote on a large scale, especially under different environmental conditions and mix ratios, it becomes very difficult to obtain a large amount of ITZ data.

[0004] Therefore, in many practical engineering applications, it is often impractical to directly measure the elastic properties of ITZ. Faced with this problem, inferring the microscopic properties of ITZ from the macroscopic mechanical properties of concrete becomes a feasible alternative. Because the macroscopic properties of concrete reflect the comprehensive performance of its constituent materials, including the joint effects of aggregates, matrix and ITZ. If a relationship model between the macroscopic properties and the properties of each component can be established, the elastic properties of ITZ can be indirectly derived through the known macroscopic elastic modulus and other measurable parameters (such as the elastic parameters and volume fraction of aggregates). This type of problem is called an inversion problem in materials science.

[0005] The difficulty of the inversion problem lies in the fact that the relationship between macroscopic and microscopic properties is often nonlinear and affected by multiple factors. Although traditional analytical methods or numerical simulations can solve this problem to a certain extent, they are often inefficient and have many model assumptions, which makes it impossible to efficiently and accurately obtain the microscopic properties of ITZ. Summary of the invention

[0006] The purpose of this application is to provide a method and device for predicting the elastic properties of a concrete interface transition zone, which can efficiently and accurately predict the elastic properties of a concrete interface transition zone.

[0007] To achieve the above objectives, this application provides the following solutions:

[0008] In a first aspect, the present application provides a method for predicting the elastic properties of a concrete interface transition zone, comprising: constructing a data set; the input parameters of the data set are macroscopic parameters of concrete, and the output parameters of the data set are elastic parameters of the interface transition zone; the macroscopic parameters of concrete include macroscopic elastic parameters of concrete, volume fraction of aggregate, elastic parameters of aggregate, elastic parameters of matrix and volume fraction of interface transition zone; constructing a machine learning model; training the machine learning model using the data set to obtain an interface transition zone elastic properties prediction model; inputting the macroscopic parameters of the concrete to be tested into the interface transition zone elastic properties prediction model, and outputting the elastic parameters of the interface transition zone of the concrete to be tested; the elastic parameters of the interface transition zone of the concrete to be tested characterize the elastic properties of the interface transition zone of the concrete to be tested.

[0009] In a second aspect, the present application provides a device for predicting the elastic properties of a concrete interface transition zone, comprising: a data set construction module, used to construct a data set; the input parameters of the data set are macroscopic parameters of concrete, and the output parameters of the data set are elastic parameters of the interface transition zone; the macroscopic parameters of concrete include macroscopic elastic parameters of concrete, volume fraction of aggregate, elastic parameters of aggregate, elastic parameters of matrix and volume fraction of interface transition zone; the elastic parameters of the interface transition zone characterize the elastic properties of the interface transition zone; a model construction module, used to construct a machine learning model; a training module, used to train the machine learning model using the data set to obtain an interface transition zone elastic properties prediction model; an application module, used to input the macroscopic parameters of the concrete to be tested into the interface transition zone elastic properties prediction model, and output the elastic parameters of the interface transition zone of the concrete to be tested.

[0010] According to the specific embodiments provided in this application, this application has the following technical effects:

[0011] The present application provides a method and device for predicting the elastic properties of a concrete interface transition zone. By performing inverse calculations using a machine learning model, the efficiency of evaluating the elastic properties of the interface transition zone is greatly improved, and tedious experimental tests are avoided. At the same time, the machine learning model trained with a large-scale data set can capture complex nonlinear relationships and achieve high-precision prediction of the elastic properties of the interface transition zone. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0013] Figure 1 A schematic flow chart of a method for predicting elastic properties of a concrete interface transition zone provided in one embodiment of the present application;

[0014] Figure 2 A more specific flow chart of a method for predicting elastic properties of a concrete interface transition zone provided in one embodiment of the present application;

[0015] Figure 3 A schematic diagram of the structure of a trained artificial neural network model provided in another embodiment of the present application;

[0016] Figure 4 A schematic diagram comparing actual values ​​and predicted values ​​of a model provided in another embodiment of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0019] In an exemplary embodiment, Figure 1 As shown, the present application provides a method for predicting the elastic properties of a concrete interface transition zone, comprising the following steps 101 to 104. Among them:

[0020] Step 101: construct a data set; the input parameters of the data set are macroscopic parameters of concrete, and the output parameters of the data set are elastic parameters of the interface transition zone; the macroscopic parameters of concrete include macroscopic elastic parameters of concrete, volume fraction of aggregate, elastic parameters of aggregate, elastic parameters of matrix and volume fraction of interface transition zone.

[0021] Step 102: Build a machine learning model.

[0022] Step 103: Use the data set to train the machine learning model to obtain an interface transition zone elastic performance prediction model.

[0023] Step 104: inputting the macroscopic parameters of the concrete to be tested into the interface transition zone elastic performance prediction model, and outputting the elastic parameters of the interface transition zone of the concrete to be tested; the elastic parameters of the interface transition zone of the concrete to be tested represent the elastic performance of the interface transition zone of the concrete to be tested.

[0024] Implementing the above steps 101 to 104 and performing inversion calculations through a machine learning model can greatly improve the efficiency of evaluating the elastic properties of the interface transition zone and avoid tedious experimental tests. At the same time, the machine learning model trained with a large-scale data set can capture complex nonlinear relationships and achieve high-precision prediction of the elastic properties of the interface transition zone.

[0025] There are many ways to construct a data set in step 101. The data set comes from numerical simulation and is not limited to Fast Fourier Transform (FFT). The data set can also come from experimental data and open source data sets to increase the diversity and wide applicability of the model.

[0026] In another exemplary embodiment of the present application, the above step 101 may be replaced by the following steps 201 to 203:

[0027] Step 201: Collect the original elastic parameters of the macroscopic concrete and the original elastic parameters of the interface transition zone.

[0028] Step 202: cleaning and normalizing the original elastic parameters of the concrete macroscopically and the original elastic parameters of the interface transition zone to obtain the elastic parameters of the concrete macroscopically and the elastic parameters of the interface transition zone; the cleaning includes removing outliers and removing incomplete data.

[0029] Step 203: The macroscopic elastic parameters of concrete and the elastic parameters of the interface transition zone are combined into a data set.

[0030] In an exemplary embodiment, the above step 201 may be replaced by the following steps 301 to 302:

[0031] Step 301: The original volume fraction of aggregate, the original elastic parameters of aggregate, the original elastic parameters of matrix, the original volume fraction of interface transition zone and the original elastic parameters of interface transition zone are taken as input data.

[0032] Step 302: According to the input data, a numerical simulation method based on fast Fourier transform is used to simulate and obtain the macroscopic original elastic parameters of concrete.

[0033] In another exemplary embodiment, the above step 201 may also be: directly collecting the macroscopic original elastic parameters of concrete and the original elastic parameters of the interface transition zone from experimental data or open source data sets.

[0034] In another exemplary embodiment of the present application, the elastic parameters include: shear modulus and bulk modulus. That is, the output parameters of the data set are the shear modulus of the interface transition zone and the bulk modulus of the interface transition zone, and the input parameters of the data set include: macroscopic shear modulus of concrete, macroscopic bulk modulus of concrete, volume fraction of aggregate, shear modulus of aggregate, bulk modulus of aggregate, shear modulus of matrix, bulk modulus of matrix and volume fraction of interface transition zone.

[0035] In another exemplary embodiment of the present application, the generated data set is randomly divided into a training set and a test set. The training set is used to build a machine learning model. The trained model is tested by the test set to evaluate the robustness and prediction accuracy of the model to ensure that it can accurately infer the elastic parameters of the ITZ on the new data set. The model input is the macro-elastic properties of concrete, the volume fraction and elastic parameters of aggregates, the elastic parameters of the matrix, and the volume fraction of the ITZ. The model output is the elastic parameters of the ITZ. The hyperparameters are adjusted through an optimization algorithm (such as a grid search or a random search) to ensure the optimal performance of the model. Then the above step 103 can be replaced by the following steps 401 to 405:

[0036] Step 401: Divide the data set into a training set and a test set.

[0037] Step 402: Use the training set to train the machine learning model to obtain a trained machine learning model.

[0038] The training set is used to adjust the hyperparameters of the neural network model, including but not limited to the learning rate, number of neurons, and regularization coefficient, to optimize the training effect of the model.

[0039] Step 403: Use the test set to verify the prediction performance of the trained machine learning model.

[0040] Step 404: If the prediction performance does not meet the preset performance requirements, adjust the hyperparameters of the trained machine learning model and return to the step "Use the test set to verify the prediction performance of the trained machine learning model".

[0041] Step 405: If the predicted performance meets the preset performance requirements, the trained machine learning model is determined as the interface transition zone elastic performance prediction model.

[0042] In another exemplary embodiment of the present application, the machine learning model is not limited to the artificial neural network model, and other regression models such as random forest, support vector machine or gradient boosting tree can also be used.

[0043] Exemplarily, when the machine learning model is an artificial neural network model, the interface transition zone elastic performance prediction model includes: an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer connected in sequence; the first hidden layer, the second hidden layer and the third hidden layer each have 90 neurons.

[0044] The trained and verified artificial neural network model (interface transition zone elastic performance prediction model) was applied to the new data set, and the new concrete macro-elastic properties, aggregate volume fraction and elastic parameters, matrix elastic parameters, and ITZ volume fraction were input. The model was used to predict and evaluate the elastic parameters of the ITZ.

[0045] The following artificial neural network model is used as the machine learning model. Figure 2 The method shown is a more specific process, which explains the method of this application in detail.

[0046] Figure 2 The entire process shown includes dataset generation, model building, training and testing, as well as application in actual engineering. Figure 2 The whole process from generating data sets by FFT numerical simulation to predicting the elastic properties of ITZ by machine learning techniques is described in detail.

[0047] Step 1: Dataset Generation

[0048] The FFT numerical simulation method is used to generate a data set of macro-elastic properties of concrete. The input parameters include: the shear modulus μ of the aggregate agg (3.72-98.61GPa) and bulk modulus κ agg (6.27-99.94GPa); ITZ shear modulus μ itz (0.37-56.17 GPa) and bulk modulus κ itz (0.29-62.59GPa); shear modulus of the matrix μ m (1.05-89.48GPa) and bulk modulus κ m (0.91-93.65GPa). In this example, for simplicity, the volume fraction of aggregate is taken as f agg =30%; volume fraction of ITZ f itz =5%.

[0049] Through FFT numerical simulation, the corresponding macro elastic performance data are output, including the macro shear modulus μ of concrete. hom and bulk modulus κ hom . These outputs are used to construct the dataset.

[0050] By changing the input parameters (such as aggregate volume fraction, elastic modulus, etc.) and performing simulations multiple times, 2222 sets of data were generated. These data sets represent the macro-elastic properties of concrete under different mix proportions and conditions, providing the diversity required for training the model.

[0051] Step 2: Construction of artificial neural network model

[0052] After generating the data set, the data needs to be cleaned first to remove outliers and incomplete data. Incomplete data can be removed by deletion, filling or interpolation, and outliers can be processed by setting empirical thresholds. For example, for ITZ, 0.3μ m <κ itz <0.7μ m and 0.3κ m <κ itz <0.7κ m ; For aggregate, take μ m <μ agg and k m <k agg ; In addition, in order to ensure the rationality of the numerical value, the value of each elastic parameter is controlled in the range of 1-100 GPa.

[0053] In order to ensure the stability of model training, the input parameters of aggregate, matrix and ITZ need to be normalized. Common normalization methods include minimum-maximum normalization, which scales the data to the range of [0,1]. The normalized data enables the model to have better processing capabilities for data of different dimensions and avoids certain input variables from having too much impact on the model. The cleaned and normalized dataset is randomly divided into a training set (80%) and a test set (20%). The training set is used to train the model, while the test set is used to verify the performance of the model.

[0054] The input layer of the artificial neural network model includes 6 key parameters: the bulk modulus μ of the aggregate agg and shear modulus k agg ; Cutting edge modulus μ of the matrix m and bulk modulus k m ; Macroscopic shear modulus μ hom and bulk modulus κ hom The output layer is the elastic properties of ITZ, including the shear modulus μ itz and bulk modulus κ itz .

[0055] The model structure initially adopted a feedforward neural network with 90 neurons in each layer. The activation function of the hidden layer selected the ReLU (rectified linear unit) function to handle complex nonlinear relationships, and the output layer used a linear activation function to generate continuous output values. The entire model was trained using the back-propagation algorithm, with the goal of minimizing the error between the model output and the actual value.

[0056] During the hyperparameter adjustment process, the learning rate of the model is initially set to 0.001. Through experiments, it is found that if the model converges slowly or the loss function fluctuates greatly, the learning rate can be gradually adjusted to 0.0005 or 0.01 to speed up the convergence or stabilize the training process. The regularization term is initially set to 0.0001 to prevent the model from overfitting. When the model shows a large error on the test set, the regularization coefficient can be appropriately increased to further constrain the complexity of the model. If the model shows underfitting, the weight of the regularization term can be reduced to increase the flexibility of the model.

[0057] The number of hidden layers and neurons is another hyperparameter that needs to be adjusted. The initial model uses one hidden layer with 90 neurons in each layer. In subsequent experiments, if the model shows underfitting, the nonlinear expression ability of the model can be improved by increasing the number of hidden layers (such as 3 or 5 layers) or increasing the number of neurons (such as 100 or 120 neurons). If the model is overfitting, you can consider reducing the number of neurons or hidden layers.

[0058] After multiple hyperparameter adjustments and verifications, an artificial neural network structure with 3 hidden layers and 90 neurons in each layer was finally selected. The input layer of the model receives 6 normalized key parameters, namely the shear modulus and bulk modulus of aggregate and matrix, the macroscopic shear modulus and bulk modulus, and the output layer is used to predict the shear modulus and bulk modulus of ITZ. Each hidden layer uses the ReLU activation function to handle the nonlinear relationship in the data, while the output layer uses a linear activation function to generate continuous output values. The training process of the model uses the stochastic gradient descent (SGD) algorithm, with the initial learning rate set to 0.001 and the regularization coefficient set to 0.0001 to prevent overfitting. Under this structure, the model shows good prediction accuracy and stability, and can effectively capture the complex relationship between the macroscopic properties of concrete and the elastic properties of ITZ. The trained artificial neural network model is as follows: Figure 3 shown.

[0059] Step 3: Model testing and evaluation

[0060] After completing the model training, the test set (20%) preprocessed in step 2 is used as input to the trained artificial neural network model. The input parameters of the test set include the shear modulus and bulk modulus of aggregate, matrix and macro. The model outputs the elastic properties (shear modulus and bulk modulus) of ITZ based on the input test set data. The error is calculated by mathematical statistical methods based on the output of the model and the elastic properties of ITZ corresponding to the test set. The relevant formula is as follows:

[0061] R 2 (Determination coefficient): It is used to measure the ability of the model to explain the variables. The closer the value is to 1, the better the model fit is. The calculation formula is:

[0062]

[0063] Among them, y i is the actual value, is the predicted value, is the average of the actual values.

[0064] RMSE (root mean square error): used to measure the deviation between the model prediction value and the actual value. The smaller the value, the better the prediction effect of the model. The calculation formula is:

[0065]

[0066] MAE (mean absolute error): used to calculate the average absolute deviation between the predicted value and the actual value. The smaller the value, the higher the prediction accuracy of the model. The calculation formula is:

[0067]

[0068] MARE (mean absolute relative error): used to measure the percentage error of the predicted error relative to the actual value. The calculation formula is

[0069]

[0070] R 2 The value is close to 1, and the RMSE, MAE, and MARE values ​​are low, indicating that the model has a strong predictive ability and can accurately infer the elastic performance of the ITZ on the test data.

[0071] The actual value of the model is compared with the predicted value. Figure 4 This shows a high degree of agreement. During the model validation, error evaluation was performed on the shear modulus and bulk modulus, respectively. For the shear modulus, the R 2 The value is 0.9982, the RMSE is 0.3919 GPa, the MAE is 0.3048 GPa, and the MARE is 3.0783%, showing a high degree of fit and prediction accuracy. For the bulk modulus, R2 The value is 0.9974, the RMSE is 0.5057 GPa, the MAE is 0.3995 GPa, and the MARE is 3.0971%, which also shows that the prediction deviation of the model is small, the model has excellent prediction performance on the test data, and has good generalization ability. Overall, the prediction performance of the model on shear modulus and bulk modulus is relatively ideal, and the elastic properties of ITZ can be accurately inferred from the test data.

[0072] If the evaluation result is not ideal, the model can be optimized by adjusting the hyperparameters of the model (such as learning rate, number of neurons, regularization coefficient). In this embodiment, the test results show that the error control of the model is within a reasonable range and no further adjustment is required.

[0073] Step 4: Practical Application

[0074] In actual engineering, the macro elastic performance parameters of concrete are known, including the specific elastic parameters of aggregate and matrix. The specific parameters in a certain engineering project were measured by the following method: the macro shear modulus of concrete was measured to be 21.73GPa and the bulk modulus was 28.25GPa using a uniaxial compression test; the shear modulus of aggregate was estimated to be 52.63GPa and the bulk modulus was 83.64GPa based on reference data and material manuals combined with the actual engineering situation, and the aggregate volume fraction was 30%; the shear modulus of the matrix was 16.11GPa and the bulk modulus was 19.95GPa, which were obtained through compression tests; the volume fraction of ITZ was 5%, which was determined by X-ray tomography (X-ray CT) technology. These parameters were used as inputs into the trained artificial neural network model to predict the elastic properties of ITZ.

[0075] Based on the input macro parameters, the model outputs the predicted ITZ performance results: shear modulus 9.73GPa, bulk modulus 11.37GPa. These predictions can help engineers understand the actual performance of ITZ and adjust the mix and composition of concrete accordingly. By evaluating the elastic properties of ITZ, the overall mechanical properties of concrete can be optimized, especially the durability and compressive resistance in long-term use.

[0076] The method of this application can be applied to concrete material design and structural health monitoring, providing an efficient technical means for evaluating ITZ performance.

[0077] Based on the prediction results of ITZ elastic properties, engineers can optimize the composition of concrete to improve its overall mechanical properties. Through the trained artificial neural network model, the elastic properties of ITZ (such as shear modulus and bulk modulus) are predicted, thus providing a basis for adjusting the proportion of concrete. For example, if the model prediction shows that the elastic properties of ITZ are weak, it may cause the concrete to perform insufficiently under certain stress conditions. Engineers can increase the proportion of aggregate or matrix accordingly, or use reinforcing agents to improve the microstructure of the material and enhance the durability and crack resistance of concrete. This method of material optimization based on elastic property prediction not only improves the design efficiency of concrete materials, but also effectively improves the long-term performance of concrete.

[0078] In terms of structural health monitoring, although elastic performance prediction is used for the optimization design of preliminary materials, the long-term health status of the structure needs to rely on the monitoring of the sensor network. Specifically, sensors (such as strain gauges, accelerometers, etc.) will be deployed in key parts of the concrete structure to collect data such as stress, strain, and crack extension in real time. Through the information fed back by the sensors, combined with the preliminary prediction of ITZ performance, engineers can dynamically analyze the stress response of concrete under different environmental conditions and during use to ensure that the structure remains stable in actual work. In addition, the monitoring data of the sensors is not only used for real-time evaluation, but also can be used to analyze the long-term damage trend of the structure through the long-term accumulated data combined with the initial prediction of the artificial neural network model. If the structural strain is monitored to exceed expectations, or the sensor captures stress conditions inconsistent with the model prediction, the engineer can adjust the composition of the concrete in time, or take appropriate repair measures to avoid further deterioration of the structure. Regular input of monitoring data can also help engineers evaluate the degradation of materials in long-term use and ensure the safety and stability of the structure throughout its life cycle. Through this comprehensive solution combining ITZ elastic performance prediction and structural health monitoring, engineers can achieve closed-loop control of material design and structural maintenance. The prediction of ITZ elastic properties provides the basis for optimizing the initial concrete mix, while sensor monitoring ensures the long-term health of the structure. This intelligent structural assessment and adjustment method is highly innovative and can not only improve the safety of concrete structures, but also significantly extend their service life.

[0079] This application solves the problem that the ITZ elastic properties of the concrete interface transition zone are difficult to measure directly. By combining the inversion idea with machine learning technology, it realizes the efficient evaluation of the microscopic properties of concrete, providing a new technical approach for concrete material design and structural health monitoring.

[0080] Compared with the prior art, this application has the following advantages:

[0081] 1) High efficiency: The inversion calculation through machine learning model greatly improves the efficiency of ITZ elastic performance evaluation and avoids tedious experimental testing;

[0082] 2) High accuracy: The neural network model trained with large-scale data sets can capture complex nonlinear relationships and achieve high-precision prediction of ITZ performance;

[0083] 3) Wide applicability: This application is applicable to concrete materials with different mix proportions and environmental conditions, has strong generalization ability, and can be widely used in practical engineering.

[0084] Based on the same inventive concept, the embodiment of the present application also provides a concrete interface transition zone elastic performance prediction device for implementing the above-mentioned concrete interface transition zone elastic performance prediction method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the concrete interface transition zone elastic performance prediction device provided below can refer to the limitations of the concrete interface transition zone elastic performance prediction method above, and will not be repeated here.

[0085] In an exemplary embodiment, a device for predicting elastic properties of a transition zone of a concrete interface is provided, comprising: a data set construction module, a model construction module, a training module, and an application module.

[0086] A data set construction module is used to construct a data set; the input parameters of the data set are macroscopic parameters of concrete, and the output parameters of the data set are elastic parameters of the interface transition zone; the macroscopic parameters of concrete include macroscopic elastic parameters of concrete, volume fraction of aggregate, elastic parameters of aggregate, elastic parameters of matrix and volume fraction of interface transition zone; the elastic parameters of interface transition zone characterize the elastic properties of interface transition zone.

[0087] Model building module for building machine learning models.

[0088] A training module is used to train the machine learning model using the data set to obtain a prediction model for the elastic properties of the interface transition zone.

[0089] The application module is used to input the macroscopic parameters of the concrete to be tested into the elastic performance prediction model of the interface transition zone and output the elastic parameters of the interface transition zone of the concrete to be tested.

[0090] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0091] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting elastic properties of a concrete interface transition zone, characterized in that: include: Constructing a data set; the input parameters of the data set are macroscopic parameters of concrete, and the output parameters of the data set are elastic parameters of the interface transition zone; the macroscopic parameters of concrete include macroscopic elastic parameters of concrete, volume fraction of aggregate, elastic parameters of aggregate, elastic parameters of matrix and volume fraction of interface transition zone; Build machine learning models; Using the data set to train the machine learning model to obtain an interface transition zone elastic performance prediction model; The macroscopic parameters of the concrete to be tested are input into the interface transition zone elastic performance prediction model, and the elastic parameters of the interface transition zone of the concrete to be tested are output; the elastic parameters of the interface transition zone of the concrete to be tested represent the elastic performance of the interface transition zone of the concrete to be tested.

2. The method for predicting elastic properties of concrete interface transition zone according to claim 1, characterized in that: Construct a dataset, including: Collect the original elastic parameters of the macroscopic concrete and the original elastic parameters of the interface transition zone; Cleaning and normalizing the original elastic parameters of the concrete macroscopically and the original elastic parameters of the interface transition zone to obtain the elastic parameters of the concrete macroscopically and the elastic parameters of the interface transition zone; the cleaning includes removing abnormal values ​​and removing incomplete data; The macroscopic elastic parameters of concrete and the elastic parameters of the interface transition zone constitute the data set.

3. The method for predicting elastic properties of concrete interface transition zone according to claim 2, characterized in that: Collect the original elastic parameters of the concrete macroscopic and the original elastic parameters of the interface transition zone, including: The original volume fraction of aggregate, the original elastic parameters of aggregate, the original elastic parameters of matrix, the original volume fraction of interface transition zone and the original elastic parameters of interface transition zone are used as input data; According to the input data, a numerical simulation method based on fast Fourier transform is used to simulate and obtain the macroscopic original elastic parameters of concrete.

4. The method for predicting elastic properties of concrete interface transition zone according to claim 2, characterized in that: Collect the original elastic parameters of the concrete macroscopic and the original elastic parameters of the interface transition zone, including: The original elastic parameters of the macroscopic concrete and the original elastic parameters of the interface transition zone are directly collected from experimental data or open source data sets.

5. The method for predicting elastic properties of concrete interface transition zone according to claim 1, characterized in that: The elastic parameters include: shear modulus and bulk modulus.

6. The method for predicting elastic properties of concrete interface transition zone according to claim 1, characterized in that: The machine learning model is trained using the data set to obtain an interface transition zone elastic performance prediction model, specifically including: Dividing the data set into a training set and a test set; Using the training set to train the machine learning model to obtain a trained machine learning model; Use the test set to verify the predictive performance of the trained machine learning model; If the prediction performance does not meet the preset performance requirements, adjust the hyperparameters of the trained machine learning model and return to step "Use the test set to verify the prediction performance of the trained machine learning model"; If the predicted performance meets the preset performance requirements, the trained machine learning model is determined as the interface transition zone elastic performance prediction model.

7. The method for predicting elastic properties of concrete interface transition zone according to claim 1, characterized in that: The machine learning model includes: an artificial neural network model or a regression model.

8. The method for predicting elastic properties of concrete interface transition zone according to claim 7, characterized in that: When the machine learning model is an artificial neural network model, the interface transition zone elastic performance prediction model includes: an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer connected in sequence; The first hidden layer, the second hidden layer, and the third hidden layer each have 90 neurons.

9. The method for predicting elastic properties of concrete interface transition zone according to claim 1, characterized in that: The macro parameters of the concrete to be tested are input into the elastic performance prediction model of the interface transition zone, and the elastic parameters of the interface transition zone of the concrete to be tested are output, and then the following are included: If the elastic parameter of the interface transition zone of the concrete to be tested is less than the preset elastic parameter threshold, the mix ratio of the concrete to be tested is adjusted; the adjustment of the mix ratio of the concrete to be tested includes: increasing the proportion of aggregates, increasing the proportion of matrix, using reinforcing agents, or improving the microstructure of the materials used for the concrete to be tested.

10. A device for predicting elastic properties of a concrete interface transition zone, characterized in that: The concrete interface transition zone elastic performance prediction device comprises: A data set construction module is used to construct a data set; the input parameters of the data set are macroscopic parameters of concrete, and the output parameters of the data set are elastic parameters of the interface transition zone; the macroscopic parameters of concrete include macroscopic elastic parameters of concrete, volume fraction of aggregate, elastic parameters of aggregate, elastic parameters of matrix and volume fraction of interface transition zone; the elastic parameters of interface transition zone characterize the elastic properties of interface transition zone; Model building module, used to build machine learning models; A training module, used to train the machine learning model using the data set to obtain an interface transition zone elastic performance prediction model; The application module is used to input the macroscopic parameters of the concrete to be tested into the elastic performance prediction model of the interface transition zone and output the elastic parameters of the interface transition zone of the concrete to be tested.

Images