A method for predicting the strength of geopolymer concrete

Abstract

The present invention discloses a method for predicting the strength of geopolymer concrete, which comprises the following steps: step 1: collecting geopolymer concrete strength experimental sample data to train a convolutional neural network, and constructing an initial model for predicting the strength of geopolymer concrete; step 2: optimizing the initial meta-parameters in the initial model for predicting the strength of geopolymer concrete by using a bat algorithm to obtain the optimal values ​​of the meta-parameters; step 3: updating the initial prediction model according to the optimal values ​​of the meta-parameters, and constructing an optimized geopolymer concrete strength prediction model; step 4: outputting a predicted value of the compressive strength of the predicted concrete according to the geopolymer concrete strength prediction model; the present invention is conducive to accurately predicting and evaluating the compressive strength value of geopolymer concrete in industrial production, thereby promoting production.

Description

Technical field:

[0001] The present invention relates to the field of concrete mechanical property judgment, and in particular to a method for predicting the strength of geopolymer concrete. Background technology:

[0002] The compressive strength (CS) of geopolymer concrete is an important indicator of concrete's mechanical properties and a crucial parameter that requires accurate measurement during its production and use. Taking fly ash-slag-based geopolymer concrete as an example, its compressive strength is influenced by multiple parameters, including the type of raw materials, the type and content of the alkali activator, and the liquid-to-solid ratio of the mixture. The compressive strength of geopolymer concrete is typically measured by pressure testing concrete specimens after standard curing. This experimental approach involves steps such as specimen preparation, curing, and pressure testing, resulting in long waiting times and requiring significant human, material, and financial resources, resulting in low efficiency. Therefore, developing a geopolymer concrete strength prediction method based on a deep learning algorithm to replace traditional experimental methods can reduce costs and increase efficiency, and has significant theoretical value and engineering economic benefits for ensuring the construction progress and quality of geopolymer concrete.

[0003] In recent years, the application of artificial intelligence (AI) methods in concrete mix design and performance prediction has gained increasing attention. Some researchers have used machine learning (ML) methods, such as support vector machines (SVMs), artificial neural networks (ANNs), and extreme learning machines (ELMs), to predict the 28-day strength of geopolymer concrete. However, for the prediction of the compressive strength of fly ash-slag-based geopolymer concrete, the accuracy of the prediction results from these ML methods is significantly affected by the selection of model input variables and the setting of model element parameters. Furthermore, traditional machine learning models have simple and limited architectures, and their application scope and effectiveness are significantly limited compared to deep learning.

[0004] Deep learning (DL) methods, through a combination of feature extraction and pattern recognition, are able to create sophisticated learning models that more effectively manage unstructured data through higher-level learning. This enables DL models to perform complex tasks with high precision. Consequently, they are able to better decipher the complex relationship between the mixing parameters of concrete raw materials and their final mechanical properties than traditional ML models. Among the many deep learning methods, those based on convolutional neural networks (CNNs) are widely used in the field of concrete materials research. Results have shown that optimizing the meta-parameters of CNN models is crucial.

[0005] To ensure the accuracy of CNN models in predicting the compressive strength of geopolymer concrete, it is necessary to optimize the model's meta-parameters. The goal of meta-parameter optimization is to find the global optimal solution by minimizing the root mean square error (RMS) and maximizing the coefficient of determination between the model's predicted and true values. The Bat Algorithm (BA) has been successfully applied to many optimization problems. It boasts rapid convergence and can optimize the model's meta-parameters during CNN model training. However, relevant research has shown that BA, like other swarm intelligence algorithms, suffers from shortcomings such as insufficient optimization accuracy, low efficiency, and a tendency to fall into local optima. Summary of the invention:

[0006] In response to the technical problems existing in the prior art, the present invention proposes a method for predicting the strength of geopolymer concrete; this method introduces a new adaptive inertia weight coefficient and random perturbation into the bat velocity iterative update formula through an improved bat algorithm, thereby improving the local search capability and the global search capability; thereby, it can more accurately extract and identify the nonlinear characteristics between multiple parameters of geopolymer concrete and its strength, and improve the accuracy of the prediction results of the compressive strength of geopolymer concrete; at the same time, the improved bat algorithm proposed in the present invention can solve the problem that the traditional bat algorithm has the risk of failure when facing highly nonlinear problems.

[0007] In order to solve the problems of the prior art, the present invention adopts the following technical solutions:

[0008] A method for predicting the strength of geopolymer concrete, comprising the following steps:

[0009] Step 1: Collect training sample data of geopolymer concrete strength experiment to train the convolutional neural network and build the initial model for geopolymer concrete strength prediction;

[0010] Step 2: Optimizing the initial meta-parameters in the initial model for predicting the strength of geopolymer concrete by using the bat algorithm to obtain the optimal values ​​of the meta-parameters;

[0011] Step 3: Update the initial prediction model according to the optimal values ​​of the meta-parameters to construct an optimized geopolymer concrete strength prediction model;

[0012] Step 4: Output the predicted compressive strength value of the predicted concrete according to the geopolymer concrete strength prediction model; wherein:

[0013] The process of optimizing the meta-parameters in the geopolymer concrete model by using the bat algorithm to obtain the optimal values ​​of the meta-parameters includes the following steps:

[0014] Initial values ​​of bat parameters for constructing geopolymer concrete model;

[0015] The meta-parameters in the geopolymer concrete model are located by using a single bat in a randomly initialized population;

[0016] The optimal meta-parameters are obtained by performing fitness calculation on individual bats based on the meta-parameters in the geopolymer concrete model after positioning.

[0017] The speed and position of the optimal element parameters in the geopolymer concrete model are updated according to the following formula;

[0018]

[0019] in: represents the speed of the i-th bat at time t; represents the position of the i-th bat at time t minus the optimal solution of all individuals in the current iteration; x * Indicates the optimal solution for all individuals in the current iteration; f i represents the pulse frequency of the i-th bat; represents a random item; x r represents the position of a randomly selected bat; lc is the learning factor; τ is the adaptive inertia weight coefficient; that is:

[0020]

[0021] Among them: it represents the current iteration number; τ min and τ max Represent the minimum and maximum inertia weights respectively; ω is a constant that adjusts the diffusion area of ​​the curve and is set to 10; N ts is the total number of iterations;

[0022] The random walk model is used to conduct local search on the geopolymer concrete model to update the position of each optimal element parameter;

[0023] An iterative search is performed on the geopolymer concrete model to update each optimal meta-parameter loudness and pulse emission rate;

[0024] Re-evaluate the fitness of the optimal meta-parameters in the geopolymer concrete model to obtain the optimal solution of the meta-parameters;

[0025] Determine whether the stopping criterion is met and output the optimal solution when the number of iterations exceeds the preset maximum number of iterations; otherwise, return to step 3 and recalculate.

[0026] Furthermore, the process of collecting geopolymer concrete training sample data in step 1 to train a convolutional neural network to construct a geopolymer concrete model includes the following steps:

[0027] The experimental data of strength influencing factors in the training sample data set are input into the input layer of the geopolymer concrete strength prediction model;

[0028] Positively outputting the concrete strength prediction value through step 4 under the connection weight and bias conditions of the initial geopolymer concrete strength prediction model;

[0029] The error between the predicted strength value and the actual measured value is back-propagated from the output layer to the input layer; during the back-propagation process, the geopolymer concrete strength prediction model is continuously updated so that the predicted strength value is close to the corresponding true value;

[0030] Repeat the above steps until the error reaches a certain threshold, and then end the training of the geopolymer concrete strength prediction model. Furthermore, the geopolymer concrete strength model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer; wherein:

[0031] The input layer contains 11 neurons, corresponding to the 11 influencing factors of the compressive strength of geopolymer concrete;

[0032] The input layer is followed by two convolutional layers, which are used for extracting intensity-sensitive features and deep representation respectively;

[0033] The pooling layer downsamples the learned deep features to form a feature map;

[0034] The fully connected layer inputs the feature map into the second fully connected layer to perform redundant feature elimination and pattern recognition;

[0035] The output layer outputs the predicted value of the compressive strength of the geopolymer concrete.

[0036] Furthermore, the geopolymer concrete training sample data includes the mass of fly ash and slag in a unit volume of mixture, the ratio of fine aggregate and coarse aggregate to binder, the liquid-to-solid ratio between water and solid, the molar ratio of silica to sodium oxide (Ms), the ratio of sodium oxide to binder (nm), the plasticizer ratio (SP), the curing temperature (CT), the oven curing time (CH), and the concrete age (CA).

[0037] Beneficial effects

[0038] First, the present invention introduces an adaptive inertia weight coefficient into the bat velocity update formula, thereby enhancing the local optimization capability when solving multi-dimensional complex problems. A random term is added to increase the diversity of the bat population, thereby improving the optimization accuracy and efficiency of the algorithm and avoiding the limitations of traditional methods.

[0039] Second: The present invention uses the bat algorithm to optimize the meta-parameters of the deep learning model, proposes an improved bat algorithm to optimize the geopolymer concrete compressive strength prediction model, establishes a complex mapping relationship between the content and proportion of each raw material of the geopolymer mixture and the compressive strength, and realizes the evaluation of the predicted value of the geopolymer concrete compressive strength.

[0040] Third: The present invention is conducive to accurately predicting and evaluating the compressive strength value of geopolymer concrete in industrial production, thereby promoting production. Description of the drawings:

[0041] Figure 1 The CNN model structure for strength prediction of fly ash-slag based aggregate concrete;

[0042] Figure 2 This is the flow chart for predicting the compressive strength of geopolymer concrete based on EBA-CNN;

[0043] Figure 3 Comparison between the strength prediction values ​​based on the proposed method and the actual measured values;

[0044] Figure 4 Based on the regression analysis of the proposed EBA-CNN intensity prediction results;

[0045] Figure 5 Regression analysis based on SVM intensity prediction results;

[0046] Figure 6 Regression analysis based on general CNN intensity prediction results. Specific implementation method:

[0047] The following is combined with Figure 1-6 The present invention is described as follows:

[0048] like Figure 1-2 As shown, the present invention provides a method for predicting the strength of geopolymer concrete, comprising the following steps:

[0049] Training Sample Data Collection and Analysis: We collected extensive measurement data on various mix proportions, curing conditions, ages, and pressure tests of fly ash-slag-based geopolymer concrete from a publicly available comprehensive experimental database. We statistically analyzed the impact of various experimental data on the standard compressive strength of geopolymer concrete, including the mass of fly ash and slag per unit volume of the mixture, the ratio of coarse / fine aggregate and aggregate to binder, the liquid-to-solid ratio of water to solids, the molar ratio of silica to sodium oxide (Ms), the ratio of sodium oxide to binder (nm), the plasticizer content (SP), the curing temperature (CT), the oven curing time (CH), and the concrete age (CA). This dataset was used for model training and validation of the model's predictions.

[0050] Initial configuration of the convolutional neural network; the entire model network consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer, with a total of 7 layers. The input layer contains 11 neurons, corresponding to the 11 influencing factors of the compressive strength of geopolymer concrete; the input layer is followed by two convolutional layers, which are used for extracting strength-sensitive features and deep representation respectively; then, a global maximum pooling layer is used to downsample the learned deep features to form a feature map; then, the feature map is input into two fully connected layers for redundant feature elimination and pattern recognition; finally, the predicted value of the compressive strength of geopolymer concrete is output in the output layer. The process of collecting geopolymer concrete training sample data to train the convolutional neural network and construct a geopolymer concrete model includes the following steps:

[0051] Training process of geopolymer concrete strength prediction model with initial configuration;

[0052] The experimental data of strength influencing factors in the training sample data set are input into the input layer of the geopolymer concrete strength prediction model;

[0053] Positively outputting the concrete strength prediction value through step 4 under the connection weight and bias conditions of the initial geopolymer concrete strength prediction model;

[0054] The error between the predicted strength value and the actual measured value is back-propagated from the output layer to the input layer; during the back-propagation process, the geopolymer concrete strength prediction model is continuously updated so that the predicted strength value is close to the corresponding true value;

[0055] Repeat the above steps until the error reaches a certain threshold, and then end the training of the geopolymer concrete strength prediction model.

[0056] Optimization of meta-parameters in geopolymer concrete model through improved bat algorithm: The bat algorithm in this invention has two important improvements compared with traditional methods.

[0057] The adaptive inertia weight coefficient τ of the hyperbolic tangent function is introduced into the bat speed update formula, and the expression is shown in formula (1):

[0058] (1)

[0059] Among them, it represents the current iteration number; τ min and τ max Represent the minimum and maximum inertia weights respectively; ω is a constant that adjusts the diffusion area of ​​the curve and is set to 10; N ts is the total number of iterations.

[0060] Add a random term to the velocity update formula The purpose is to improve the diversity of bat groups. Therefore, after combining the two improvements, the updated expression of bat speed in EBA is shown in formula (2):

[0061]

[0062] in, represents the speed of the i-th bat at time t; represents the position of the i-th bat at time t minus the optimal solution of all individuals in the current iteration; x * Indicates the optimal solution for all individuals in the current iteration; f i represents the pulse frequency of the i-th bat; represents a random item; x r represents the position of a randomly selected bat; lc is the learning factor. The learning factor is shown in formula (3):

[0063] lc=lc min +(lc max -lc min )θ (3)

[0064] Among them, lc min and lc max denote the minimum and maximum learning factors respectively; θ is a random number between 0 and 1.

[0065] Through the above analysis, the calculation process of the meta-parameter optimization of the geopolymer concrete model based on the improved bat algorithm (EBA) includes:

[0066] 1) Determine the initial parameters based on the geopolymer concrete strength prediction model, including bat population size, solution dimension, initial pulse emission rate, maximum and minimum pulse frequency, maximum and minimum inertia weight, maximum and minimum learning factor, and maximum number of iterations;

[0067] 2) According to the initial values ​​of the geopolymer concrete strength prediction model parameters, the positions of the parameters in the population are randomly initialized;

[0068] 3) Based on the fitness of each bat, find the single bat with the best meta-parameter fitness, which is considered the optimal solution;

[0069] 4) Update the speed and position of the optimal element parameters in the geopolymer concrete model according to the following formula;

[0070]

[0071] in: represents the speed of the i-th bat at time t; x r represents the position of a randomly selected bat; f irepresents the pulse frequency of the i-th bat; lc is the learning factor;

[0072] 5) Local search process, generating a random number between 0 and 1, and using the random walk model to perform local search on the geopolymer concrete model to update the position of each current optimal meta-parameter;

[0073] 6) In the initial stage of iterative search, a random number between 0 and 1 is generated, and the geopolymer concrete model is iteratively searched to update each optimal meta-parameter loudness and pulse emission rate;

[0074] 7) Re-evaluate the fitness of the optimal meta-parameters in the geopolymer concrete model to obtain the optimal solution of the meta-parameters;

[0075] 8) Determine whether the stopping criterion is met, and output the optimal solution when the number of iterations exceeds the preset maximum number of iterations; otherwise, return to step 3) and recalculate.

[0076] The model was retrained using the dataset to obtain an optimized prediction model. Eleven factors, including the raw materials, mix ratio, and curing conditions of the geopolymer concrete to be predicted, were input into the optimized prediction model, which then output the predicted value of the geopolymer concrete's compressive strength.

[0077] 1. Acquisition and analysis of experimental data

[0078] A total of 850 sets of experimental compressive strength data for fly ash-slag-based geopolymer concrete with different mix proportions, curing conditions, and ages were collected from a comprehensive, publicly available experimental database. The effects of various experimental data on the standard compressive strength of geopolymer concrete were statistically analyzed, including the mass of fly ash and slag per unit volume of the mixture, the ratio of fine and coarse aggregate to binder (FBR & CBR), the liquid-to-solid ratio (W / S), the molar ratio of silica to sodium oxide (Ms), the ratio of sodium oxide to binder (nm), the plasticizer ratio (SP), the curing temperature (CT), the oven curing time (CH), and the concrete age (CA). Table 1 shows the statistical analysis results of the geopolymer concrete strength measurement experimental data set, including the maximum, minimum, median, mode, mean, skewness, kurtosis, and standard deviation. Among these statistical indices, the median, mode, and mean represent the central characteristics of the dataset, while the maximum, minimum, skewness, kurtosis, and standard deviation represent the irregular characteristics of the data. Statistical results show that the strength of fly ash-slag geopolymer concrete is widely affected by the above parameters. In order to ensure the accuracy of strength prediction of geopolymer concrete based on the above parameters, the strength prediction model is required to have sufficient robustness.

[0079] Table 1 Statistical analysis results of geopolymer concrete strength measurement experimental data:

[0080]

[0081]

[0082] 2. Initial configuration and model training of geopolymer concrete model

[0083] Establish as Figure 1 The following table shows the structure of a CNN model with seven layers. The model has nine meta-parameters, including the initial learning rate, gradient decay factor, L2 regularization factor, learning rate decay period, learning rate decay factor, number of cores in the first convolutional layer, number of cores in the second convolutional layer, number of neurons in the first FC layer, and number of neurons in the second FC layer. The following table lists the ranges of these meta-parameters.

[0084] Table 2 CNN meta-parameter value range

[0085]

[0086] 595 (70%) of the 850 data sets were randomly selected for model training. The raw materials, proportions and curing data of each experimental group were used as input parameters of the prediction model.

[0087] 3. Parameter optimization of geopolymer concrete prediction model

[0088] During the model training process, EBA was used to iteratively optimize the model parameters. The optimization target (fitness) was defined as the root mean square error between the model intensity prediction value and the actual result during the training process. The EBA setting parameters were: bat population size N = 30, solution dimension D = 9, initial pulse emission rate R0 = 0, maximum pulse frequency f max =2.8, minimum pulse frequency f min =1.3, maximum inertia weight τ max =0.85, minimum inertia weight τ min =0.15, maximum learning factor lc max =0.6, minimum learning factor lc min = 0.3, loudness and pulse emission rate enhancement coefficient α = γ = 0.9, total number of iterations N ts = 200. All meta-parameters reached their optimal values ​​within 100 iterations, fully demonstrating the rapid convergence of EBA in model parameter optimization. The optimal values ​​of the CNN model meta-parameters obtained based on EBA are shown in the following table.

[0089] Table 3 Optimal values ​​of CNN model meta-parameters obtained based on EBA

[0090]

[0091] The EBA-CNN prediction model is constructed using the optimal values ​​of the meta-parameters in the table, and the training data is analyzed again.

[0092] 4. Verification of the effectiveness of geopolymer concrete strength model

[0093] Run the EBA-CNN prediction model, input the remaining 255 groups (30%) of corresponding experimental data into the prediction model, obtain the intensity prediction value, and compare the prediction results with the experimental results, as shown in Figure 2. Figure 3 The results show that the proposed 1D-CNN can accurately predict the compressive strength of fly ash-slag based aggregate concrete under various mixing situations and curing conditions.

[0094] To demonstrate the superiority of our method, we compared the proposed EBA-CNN prediction method with existing SVM and 1D-CNN prediction models without meta-parameter optimization. We analyzed 255 sets of experimental data using each of the three prediction methods, and performed regression analysis between the predicted intensity values ​​and the actual measured values.

[0095] It is worth noting that for all three models, most of the data points are evenly distributed around the regression line, with only a few data points outside the ±20% boundary line. 2 The value range is between 0 and 1. The larger the value, the better the performance of the model. Among these geopolymer concrete strength prediction methods, the strength prediction ability based on the proposed EBA-CNN is the best, and the determination coefficient R of the test sample is 2 It is 0.9577, which is larger than 0.9042 of the 1D-CNN method without meta-parameter optimization and 0.9019 of the SVM method.

Claims

1. A method for predicting the strength of geopolymer concrete, characterized in that: The method comprises the following steps: Step 1: Collect training sample data of geopolymer concrete strength experiment to train the convolutional neural network and build the initial model for geopolymer concrete strength prediction; Step 2: Optimizing the initial meta-parameters in the initial model for predicting the strength of geopolymer concrete by using the bat algorithm to obtain the optimal values ​​of the meta-parameters; Step 3: Update the initial prediction model according to the optimal values ​​of the meta-parameters to construct an optimized geopolymer concrete strength prediction model; Step 4: Output the predicted compressive strength value of the predicted concrete according to the geopolymer concrete strength prediction model; wherein: The process of optimizing the initial meta-parameters in the geopolymer concrete strength prediction model by using the bat algorithm to obtain the optimal values ​​of the meta-parameters includes the following steps: Constructing the initial values ​​of bat parameters required for the optimization of the parameters of geopolymer concrete strength prediction model; The positions of each bat in the population are initialized by random generation to locate the meta-parameters in the geopolymer concrete strength prediction model; According to the meta-parameters in the geopolymer concrete strength prediction model after positioning, the fitness calculation of a single bat is performed to obtain the optimal meta-parameters; The speed and position of the optimal element parameters in the geopolymer concrete model are updated according to the following formula; in: represents the speed of the i-th bat at time t; represents the position of the i-th bat at time t minus the optimal solution of all individuals in the current iteration; x * Indicates the optimal solution for all individuals in the current iteration; f i represents the pulse frequency of the i-th bat; represents a random item; x r represents the position of a randomly selected bat; lc is the learning factor; τ is the adaptive inertia weight coefficient; that is: Among them: it represents the current iteration number; τ min and τ max Represent the minimum and maximum inertia weights respectively; ω is a constant that adjusts the diffusion area of ​​the curve and is set to 10; N ts is the total number of iterations; The random walk method is used to conduct local search for the meta-parameters of the geopolymer concrete strength prediction model and update the position of each optimal meta-parameter. Iteratively search the meta-parameters of geopolymer concrete strength prediction model to update the loudness and pulse emission rate of each optimal meta-parameter; Re-evaluate the fitness of the optimal meta-parameters in the geopolymer concrete strength prediction model to obtain the optimal solution of the meta-parameters; Determine whether the stopping criterion is met and output the optimal solution when the number of iterations exceeds the preset maximum number of iterations; otherwise, return to step 3 and recalculate.

2. A method for predicting the strength of geopolymer concrete according to claim 1, characterized in that: The process of collecting geopolymer concrete strength experimental training sample data in step 1 to train a convolutional neural network to construct a geopolymer concrete strength prediction model includes the following steps: The experimental data of the factors affecting the strength of geopolymer concrete in the training sample data set are input into the input layer of the geopolymer concrete strength prediction model; Positively outputting the concrete strength prediction value through step 4 under the connection weight and bias conditions of the initial geopolymer concrete strength prediction model; The error between the predicted strength value and the actual measured value is back-propagated from the output layer to the input layer; during the back-propagation process, the geopolymer concrete strength prediction model is continuously updated so that the predicted strength value is close to the corresponding true value; Repeat the above steps until the error reaches a certain threshold, and then end the training of the geopolymer concrete strength prediction model.

3. A method for predicting the strength of geopolymer concrete according to claim 1 or 2, characterized in that: The polymer concrete strength prediction model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer; wherein: The input layer contains 11 neurons, corresponding to the 11 influencing factors of the compressive strength of geopolymer concrete; The input layer is followed by two convolutional layers, which are used for extracting intensity-sensitive features and deep representation respectively; The pooling layer downsamples the learned deep features to form a feature map; The fully connected layer inputs the feature map into the second fully connected layer to perform redundant feature elimination and pattern recognition; The output layer outputs the predicted value of the compressive strength of the geopolymer concrete.

4. A method for predicting the strength of geopolymer concrete according to claim 3, characterized in that: The geopolymer concrete strength experimental training sample data includes the mass of fly ash and slag in a unit volume of mixture, the ratio of fine aggregate and coarse aggregate to binder, the liquid-solid ratio between water and solid, the molar ratio of silicon dioxide to sodium oxide, the ratio of sodium oxide to binder, the plasticizer ratio, the curing temperature, the oven curing time, the concrete age, and the measured strength value.

Images