Method for mix proportion optimization and performance prediction of canal lining concrete

Abstract

The invention discloses a method for mix proportion optimization and performance prediction of canal lining concrete, which comprises the following steps: 1) collecting compressive strength and durability data of concrete under different mix proportion conditions through laboratory testing, and establishing an early-stage data set; 2) constructing a distribution model, and simulating and expanding the data set by adopting a Markov chain-Monte Carlo method; 3) training a Kriging agent model by using the expanded data set to obtain a high-precision virtual sample set; 4) standardizing a virtual sample output by the KS model, inputting the standardized virtual sample into the artificial neural network model as a training set, and determining an ANN model structure; 5) creating different types of test data sets, and evaluating the precision and robustness of the KS-ANN model; and 6) proposing an optimization method for concrete. According to the invention, by expanding and training limited test data, a high-precision concrete performance index prediction model is constructed, an optimization method is established, and the trial and error cost of the test is effectively reduced.

Description

Technical Field

[0001] The invention relates to the technical field of building materials, and in particular to a method for optimizing the mix ratio and predicting the performance of channel lining concrete. Background Art

[0002] In the field of water conservancy project construction, channel lining concrete is a vital building material. Its core role is to enhance the channel's resistance to erosion and effectively extend its service life by pouring concrete on the channel surface. However, during the construction and operation of the channel lining, it faces the combined effects of multiple factors such as material properties, extreme climate, complex working conditions and human factors, which can easily cause problems such as cracking and spalling. These problems will not only damage the overall performance of the channel lining, but also, in severe cases, may cause the risk of water loss. Therefore, ensuring the durability of concrete is of vital importance to maintaining the integrity of the channel structure. Concrete lining with high durability can effectively prevent the occurrence of damage such as cracks and spalling, thereby reducing the maintenance cost of the channel, ensuring the stability of the channel structure and the safety of the water delivery process.

[0003] Considering that the proportion of concrete raw materials plays a decisive role in its service performance and production cost, the proportion of channel lining concrete must be scientifically and reasonably designed in engineering practice. This design process needs to comprehensively consider multiple factors, such as the mechanical properties, durability and economy of concrete, to ensure that the concrete can maintain its design performance under the influence of long-term water scouring, immersion and extreme climatic conditions, extend the service life of the project, and improve its economic and social benefits.

[0004] In recent years, the rapid progress of artificial intelligence technology has promoted the widespread application of data-driven machine learning technology in the field of civil engineering, especially in the optimization of concrete mix ratio. In the traditional channel lining concrete design process, engineers need to carry out a large number of experiments to explore the impact of different mix ratios on concrete performance. This process not only covers multiple trial mixes and test links, but also requires detailed performance evaluation for each mix ratio, with evaluation indicators including strength, durability and workability. However, this method not only consumes a lot of manpower and material resources, but also leads to an extension of the design cycle, which in turn has an adverse impact on the progress of the project. Therefore, exploring the use of machine learning models to predict concrete performance and optimize mix design has become a vital research direction. With the help of machine learning technology, the accuracy of prediction can be improved while reducing the number of experiments, providing scientific support for the optimization of concrete mix ratio.

[0005] In previous studies, although traditional artificial neural network (ANN) models have been widely used, they have significant limitations. For example, ANN models usually require sufficient training data to effectively learn the complex nonlinear relationship between input and output, which may lead to overfitting or underfitting problems when data is scarce. In addition, when processing spatial data, traditional ANN models often find it difficult to fully consider the spatial correlation and non-stationarity of data, which limits their application in dealing with complex engineering problems. In contrast, Kriging interpolation, as a powerful statistical tool, can not only accurately estimate spatial correlation data, but also quantify the uncertainty of prediction, thereby improving the accuracy and robustness of concrete mix optimization and performance prediction models. However, in the field of mix optimization and performance prediction of channel lining concrete, there is still a lack of effective methods to combine Kriging interpolation with ANN. Summary of the invention

[0006] In view of the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a method for optimizing the mix ratio and predicting the performance of channel lining concrete, so as to provide a fast and economical solution to the performance evaluation and mix ratio optimization design of channel lining concrete and similar materials.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for optimizing the mix ratio and predicting the performance of channel lining concrete, comprising the following steps:

[0008] S1. In a laboratory environment, the compressive strength and durability experimental data of channel lining concrete under various mix ratios were collected through testing to construct a data set;

[0009] S2. Construct a statistical distribution model of the experimental data and use the Markov Chain Monte Carlo (MCMC) method to simulate and expand the data set. For each independent data set, 40,000 discrete data points are generated by simulation.

[0010] S3. Use the expanded data set to train the Kriging proxy (KS) model, and use this model to predict the compressive strength and durability performance indicators under more mix ratio combinations, so as to obtain high-precision virtual samples;

[0011] S4, using the virtual samples output by the KS model as a training set to train an artificial neural network (ANN) model;

[0012] S5. Create a test data set to evaluate the accuracy and robustness of the prediction model;

[0013] S6. Based on the prediction results of the KS-ANN model, a mix ratio optimization method for channel lining concrete is constructed.

[0014] Furthermore, the mix ratio described in step S1 includes cement content and water-cement ratio.

[0015] Furthermore, in step S3, the predicted performance indicators include compressive strength and two durability indicators, namely, impermeability and frost resistance.

[0016] Furthermore, step S1 specifically includes the following steps:

[0017] S11. Nine concrete specimens with different mix designs were prepared. These specimens consisted of three different water-cement ratios (0.3, 0.4, and 0.5) and three different cement content levels (300, 400, and 500 kg / m 3 ), each mix ratio combination contains 3 types of repeated samples, which are used to test the compressive strength and two durability performance indicators respectively;

[0018] S12. Conduct indoor tests to obtain data sets of compressive strength and durability performance indicators of channel lining concrete specimens under different mix designs through uniaxial compression tests, water penetration resistance tests and rapid freeze-thaw tests. Each group of experiments tested 3 repeated specimens, and each performance test was conducted 9 groups in total. The final data set contained 81 data points.

[0019] Furthermore, step S2 specifically includes the following steps:

[0020] S21. Assuming that the data distribution under individual conditions follows the data distribution law under group conditions, the statistical distribution model of all data is statistically analyzed to determine that the experimental data of compressive strength, impermeability and frost resistance all obey normal distribution;

[0021] S22. Use the normal distribution model to simulate each set of data, and calculate its mean and variance values. Use the numerical results obtained from the simulation as the input of the MCMN model. For each condition, simulate 40,000 discrete data points. For 9 different combinations of conditions, a total of 360,000 data points are simulated. Each condition includes compressive strength and four durability performance indicators. The final total number of simulated data reaches 1.35 million data points, and these simulated data points have the same distribution characteristics.

[0022] Further, in step S4, the input layer is the water-cement ratio (W / C) and cement content (C), the output layer is the compressive strength (σ), impermeability grade (F), and frost resistance grade (P), the trial and error method is used to adjust the hidden layer parameters, and the model is optimized by minimizing the mean square error (MSE). The learning rate of the ANN model is set to 0.002, the maximum number of iterations is 2000, the convergence error is 0.001, the ratio of the training set to the test set is 9:1, and the final ANN structure contains 3 hidden layers, each hidden layer has 128 neurons, and the specific calculation formula of MSE is as follows:

[0023]

[0024] In the formula, N represents the amount of data in the test data set. and are the actual and predicted values ​​in the test dataset, respectively.

[0025] Furthermore, in step S5, the coefficient of determination (R 2 ) to evaluate the accuracy and robustness of the model. In the model evaluation process, three different test sets were used, namely: (1) the average value of the compressive strength and durability performance indicators output by the KS model; (2) the discrete data of the compressive strength and durability performance indicators output by the KS model; (3) 30% randomly selected laboratory test data with a determination coefficient R 2 is defined as follows:

[0026]

[0027] In the formula, represents the actual data of the test dataset, Represents the predicted value of the test data set and They represent the average of the predicted and actual values ​​of the test data set respectively.

[0028] Furthermore, in step S6, compressive strength (σ), impermeability grade (F) and frost resistance grade (P) are selected as performance evaluation indicators of laboratory or field sample preparation results, and the expression of the performance function is as follows:

[0029]

[0030] In the formula, σ min 、F min and P min The minimum values ​​specified in the relevant specifications or regulations are pre-selected in the present invention as follows: the compressive strength is 40 MPa, the frost resistance grade is F300, and the impermeability grade is P8.

[0031] Compared with the prior art, the present invention brings the following significant advantages:

[0032] 1) By applying Kriging interpolation and Markov Chain Monte Carlo (MCMC) algorithms, the present invention significantly expands the size of the sample data set. This more comprehensive data set is used to train the artificial neural network model, which not only improves the robustness, generalization ability and prediction accuracy of the concrete performance prediction model, but also ensures the high credibility of the mix design based on the predicted performance results. This provides a solid technical foundation for the scientific design and engineering application of channel lining concrete.

[0033] 2) The present invention can use the optimization function to quickly determine the mix ratio value that meets the durability performance standard according to the specific requirements of concrete mix ratio design in actual engineering construction. This method effectively reduces the resource, time and manpower consumption caused by the traditional trial and error method, significantly improves R&D efficiency and reduces overall costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The figure is a schematic diagram of the steps of the method of the present invention. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of the present invention.

[0036] Figure 1 A schematic flow chart of the steps of the method for optimizing the mix ratio and predicting the performance of channel lining concrete developed by the present invention is given. The method includes 6 steps:

[0037] S1. Collect experimental data on compressive strength and durability of channel lining concrete under different mix ratios through laboratory tests and establish a data set.

[0038] The parameters considered in the mix proportion include water-cement ratio and cement content, and the performance indicators measured in the laboratory include compressive strength, impermeability, frost resistance, resistance to chloride ion erosion and resistance to sulfate erosion.

[0039] Concrete specimens under nine different mix designs were prepared with three different water-cement ratio levels (0.3, 0.4, and 0.5) and three different cement content levels (300, 400, and 500 kg / m3). Each mix combination included three types of repeated specimens corresponding to the compressive strength and two durability performance indicators.

[0040] Indoor tests were carried out to obtain the durability performance index data sets of channel lining concrete specimens under different mix ratio designs through uniaxial compression test, water penetration test and rapid freeze-thaw test, in which three repeated specimens were tested in each group of experiments. The final data set includes 81 data points in total.

[0041] S2. Construct a statistical distribution model of the experimental data and use the Markov Chain Monte Carlo (MCMC) method to simulate and expand the data set. For each separate set of data, simulate 40,000 discrete data points.

[0042] Assuming that the data distribution under individual conditions follows the data distribution under group conditions, by analyzing the statistical distribution model of all data, it is determined that the data distributions of compressive strength, impermeability, and frost resistance all obey the normal distribution.

[0043] A normal distribution model is used to fit each set of individual data and obtain its mean and variance values. The simulated numerical results are used as the input of the MCMN model. For each individual condition, 40,000 discrete data points are simulated. In the present invention, 9 different condition combinations are considered, so a total of 360,000 data points are simulated. Each condition includes four durability performance indicators, resulting in a total of 1.35 million data points in the final simulation data, and these simulated data points have the same distribution characteristics.

[0044] S3. Use the expanded data to train the Kriging proxy (KS) model, and use the model to predict the compressive strength and durability performance indicators under more mix ratio combinations to obtain a virtual sample set.

[0045] The KS model is a semi-parametric proxy model based on statistics, which can accurately predict data samples in a domain, thereby providing additional prediction samples with the same physical meaning as real samples in the domain. A large number of data samples expanded in step S2 are used to train a reliable KS model, and a large number of virtual water-cement ratios and cement content combinations are interpolated through the trained KS model to predict the durability performance parameters under different mix designs, thereby obtaining a large number of high-precision virtual samples distributed in the domain.

[0046] S4. Use the high-precision virtual samples output by the KS model as training sets to train the artificial neural network (ANN) model.

[0047] S41. Define the input and output of information. In the present invention, the input layer includes cement content (C) and water-cement ratio (W / C), and the output layer includes compressive strength (σ), impermeability grade (F), and frost resistance grade (P).

[0048] S42. The hidden layer parameters are adjusted by trial and error, and the model is optimized by minimizing the mean square error (MSE). The learning rate of the ANN model is set to 0.002, the maximum number of iterations is 2000, the convergence error is 0.001, and the ratio of the training set to the test set is 9:1. The number of hidden layers finally determined in the present invention is 3, and each hidden layer contains 128 neurons. Before starting the training process, the input data needs to be standardized, and the mean of the standardized data set x* is 0 and the standard deviation is 1. The specific formulas for standardization and MSE are expressed as follows:

[0049]

[0050] Where x* and x represent the standardized and original input data, respectively. In addition, μ and η are the mean and variance of the data x, respectively.

[0051]

[0052] In the formula, N represents the amount of data in the test dataset. and are the actual and predicted values ​​in the test dataset, respectively.

[0053] S5. Create a test dataset to evaluate the accuracy and robustness of the prediction model.

[0054] (1) Using the coefficient of determination (R 2 ) to evaluate the model accuracy and robustness to prevent underfitting or overfitting of the ANN model. Three different test sets were used in the model evaluation process, namely: (1) the average value of the compressive strength and durability performance indicators output by the KS model; (2) the discrete data of the compressive strength and durability performance indicators output by the KS model; (3) 30% randomly selected laboratory test data.

[0055] Coefficient of determination R 2 is defined as follows:

[0056]

[0057] In the formula, represents the actual data of the test dataset, Represents the predicted value of the test data set and They represent the average of the predicted and actual values ​​of the test data set respectively.

[0058] The water-cement ratio levels (0.35 and 0.45) and two different cement content levels (350 kg / m 3 and 450kg / m 3 ), formed multiple verification groups of concrete mix proportions, conducted experiments and predictions on the verification groups, and compared the predicted results with the actual experimental results to evaluate the performance characteristics of the model.

[0059]

[0060]

[0061] S6. Based on the prediction results of the KS-ANN model, a mix ratio optimization method for channel lining concrete is established.

[0062] The compressive strength (σ), impermeability grade (F) and frost resistance grade (P) are selected as the performance evaluation indicators of laboratory or on-site sample preparation results, and the expression of the performance function is as follows:

[0063]

[0064] In the formula, σ min 、F min and P min The minimum values ​​specified in the relevant specifications or regulations are pre-selected in the present invention as follows: the compressive strength is 40 MPa, the frost resistance grade is F300, and the impermeability grade is P8.

[0065] The matrix diagrams of the prediction results of compressive strength, impermeability grade and frost resistance grade based on the KS-ANN model were drawn respectively and transformed into the Cartesian coordinate system, with the abscissa being the cement content (C) and the ordinate being the water-cement ratio (W / C).

[0066] The present invention uses Python program to obtain data, and the specific code is as follows:

[0067] (1) Maldives Chain Monte Carlo Algorithm:

[0068]

[0069]

[0070]

[0071] (2) Kriging proxy algorithm:

[0072]

[0073]

[0074]

[0075]

[0076]

[0077] (3) Artificial Neural Networks:

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084] Then the prediction result matrix is ​​used to draw the points that satisfy the performance function, and these equivalent points are projected onto the corresponding matrix to obtain the contour that satisfies the performance function. The ideal optimal area that satisfies the compressive strength and durability can be obtained with the contour and the coordinate axis as the boundary. With the help of this area, the range of the best cement content can be determined by the water-cement ratio, or the optimal water-cement ratio range can be determined according to the cement content.

[0085] The technical solution of the present invention effectively integrates real laboratory tests, mathematical modeling and machine learning. By combining Markov chain Monte Carlo (MCMN) simulation with Kriging interpolation, a statistical distribution model is established from limited test data to enrich the data set. Further, an artificial neural network (ANN) model is trained using the enhanced data set. The model evaluation results show that the well-trained KS-ANN model in the present invention can effectively capture the distribution characteristics of actual data, and even in the case of limited data, it can have high prediction accuracy and robustness, thereby accurately predicting the performance of channel lining concrete. In actual engineering construction, the KS-ANN model proposed in the present invention can accurately predict the key performance indicators of concrete such as compressive strength, impermeability, frost resistance, chloride ion corrosion resistance and sulfate corrosion resistance according to the specific values ​​of water-cement ratio and cement content in the preparation process of channel lining concrete. In addition, the model can also quickly determine the optimal concrete mix ratio range that meets the performance standards based on the specific requirements of the project for cement content or water-cement ratio. This method significantly overcomes the limitations of traditional laboratory testing and improves R&D efficiency. While providing an innovative technical path for performance optimization and mix design of channel lining concrete and similar materials, it also makes the application of machine learning in small databases possible.

[0086] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for optimizing mix ratio and predicting performance of channel lining concrete, characterized in that: The following steps are involved: S1. In a laboratory environment, the compressive strength and durability experimental data of channel lining concrete under various mix ratios were collected through testing to construct a data set; S2. Construct a statistical distribution model of the experimental data and use the Markov chain-Monte Carlo method to simulate and expand the data set. For each independent data set, simulate and generate 40,000 discrete data points; S3. Use the expanded data set to train the Kriging proxy KS model, and use this model to predict the compressive strength and durability performance indicators under more mix ratio combinations, so as to obtain high-precision virtual samples; S4, using the virtual samples output by the KS model as a training set to train the artificial neural network ANN model; S5. Create a test data set to evaluate the accuracy and robustness of the prediction model; S6. Based on the prediction results of the KS-ANN model, a mix ratio optimization method for channel lining concrete is constructed.

2. A method for optimizing mix ratio and predicting performance of channel lining concrete according to claim 1, characterized in that: The mix ratio described in step S1 includes cement content and water-cement ratio.

3. The method for optimizing mix ratio and predicting performance of channel lining concrete according to claim 1, characterized in that: In step S3, the predicted performance indicators include compressive strength and four durability indicators of impermeability, frost resistance, resistance to chloride ion corrosion, and resistance to sulfate corrosion.

4. The method for optimizing mix ratio and predicting performance of channel lining concrete according to claim 1, characterized in that: Step S1 specifically includes the following steps: S11. Prepare nine concrete specimens with different mix designs, which are composed of three different water-cement ratio levels and three different cement content levels. Each mix combination contains three types of repeated specimens, which are used to test compressive strength and two durability performance indicators respectively; S12. Conduct indoor tests to obtain data sets of compressive strength and durability performance indicators of channel lining concrete specimens under different mix designs through uniaxial compression tests, water penetration resistance tests and rapid freeze-thaw tests. Each group of experiments tested 3 repeated specimens, and each performance test was conducted 9 groups in total. The final data set contained 81 data points.

5. The method for optimizing mix ratio and predicting performance of channel lining concrete according to claim 1, characterized in that: Step S2 specifically includes the following steps: S21. Assuming that the data distribution under individual conditions follows the data distribution law under group conditions, the statistical distribution model of all data is statistically analyzed to determine that the experimental data of compressive strength, impermeability and frost resistance all obey normal distribution; S22. Use the normal distribution model to simulate each set of data, and calculate its mean and variance. Use the numerical results obtained from the simulation as the input of the MCMN model. For each condition, simulate 40,000 discrete data points. Nine different combinations of conditions, a total of 360,000 data points, each condition includes compressive strength and two durability performance indicators. The final total number of simulated data reaches 1.08 million data points, and these simulated data points have the same distribution characteristics.

6. The method for optimizing mix ratio and predicting performance of channel lining concrete according to claim 1, characterized in that: In step S4, the input layer is the water-cement ratio and cement content, the output layer is the compressive strength, impermeability grade, and frost resistance grade. The trial and error method is used to adjust the hidden layer parameters, and the model is optimized by minimizing the mean square error. The learning rate of the ANN model is set to 0.002, the maximum number of iterations is 2000, the convergence error is 0.001, the ratio of the training set to the test set is 9:1, and the final ANN structure contains 3 hidden layers, each hidden layer has 128 neurons, and the specific calculation formula of MSE is as follows: In the formula, MSE represents mean square error, N represents the amount of data in the test data set, and are the actual and predicted values ​​in the test dataset, respectively.

7. The method for optimizing mix ratio and predicting performance of channel lining concrete according to claim 1, characterized in that: In step S5, the coefficient of determination R 2 To evaluate the accuracy and robustness of the model. In the model evaluation process, three different test sets were used, namely: (1) the average value of the compressive strength and durability performance indicators output by the KS model; (2) the discrete data of the compressive strength and durability performance indicators output by the KS model; (3) 30% randomly selected laboratory test data with a determination coefficient R 2 is defined as follows: In the formula, represents the actual data of the test dataset, Represents the predicted value of the test data set and They represent the average of the predicted and actual values ​​of the test data set respectively.

8. The method for optimizing mix ratio and predicting performance of channel lining concrete according to claim 1, characterized in that: In step S6, compressive strength σ, impermeability grade F and frost resistance grade P are selected as performance evaluation indicators of laboratory or on-site sample preparation results, and the expression of the performance function is as follows: In the formula, σ min 、F min and P min The minimum values ​​specified in the relevant specifications or regulations are pre-selected in the present invention as follows: the compressive strength is 40 MPa, the frost resistance grade is F300, and the impermeability grade is P8.

Images