Rubber sand concrete uniaxial compressive strength prediction method based on hybrid intelligent model

Through the backpropagation neural network based on lion group optimization combined with a hybrid intelligent model, the problem of experimental time-consuming and low accuracy in the prediction of uniaxial compressive strength of rubber-sand concrete is solved, and the uniaxial compressive strength of rubber-sand concrete is realized. Efficient and accurate prediction, suitable for the research and development of new shock absorbing materials.

CN118918994APending Publication Date: 2024-11-08INST OF ROCK & SOIL MECHANICS CHINESE ACAD OF SCI

Patent Information

Application Number
CN202410965459.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The prior art has problems such as time-consuming and labor-consuming experiments and low prediction accuracy in the prediction of single-axis compressive strength of rubber-sand concrete, and a single intelligent algorithm is difficult to meet the needs.

Method used

The backpropagation neural network based on lion group optimization is adopted to combine a hybrid intelligent model. By screening important features, building training sets and test sets, the neural network weights and thresholds are optimized to achieve efficient and accurate prediction of the uniaxial compressive strength of rubber-sand concrete.

Benefits of technology

It significantly accelerates the prediction speed and improves accuracy, providing a fast and efficient method suitable for the research and development and performance research of new shock absorbing materials.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118918994A_ABST
    Figure CN118918994A_ABST
Patent Text Reader

Abstract

The invention discloses a rubber-sand concrete uniaxial compressive strength prediction method based on a hybrid intelligent model, and the method comprises the steps: collecting actual rubber-sand concrete uniaxial compressive strength corresponding to different experimental parameters as actual output characteristic values, and recording a group of experimental parameters corresponding to each rubber-sand concrete uniaxial compressive strength test as an input characteristic group; screening the types of the input features of the input feature group, and constructing a training set and a test set; using the training set to train a lion group optimization-based back propagation neural network; and predicting the to-be-processed input feature group to obtain a predicted output feature. According to the method, the back-propagation neural network is optimized by using the lion group model to establish the hybrid intelligent model, the characteristics of different intelligent algorithms are fully utilized, the training speed is increased, the prediction precision is improved, and a rapid, efficient and accurate method is provided for predicting the uniaxial compressive strength of the rubber-sand concrete. And the method has wide applicability and practicability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent prediction of uniaxial compressive strength of new shock-absorbing materials (rubber-sand concrete) for underground engineering, and specifically relates to a method for predicting uniaxial compressive strength of rubber-sand concrete based on a hybrid intelligent model. Background Art

[0002] Materials suitable for the shock-absorbing design of underground engineering structures are usually required to have the characteristics of light weight, low elastic modulus and high damping. In recent years, flexible shock-absorbing technology has become a research hotspot for earthquake damage control of tunnel structures, and the research results of corresponding shock-absorbing materials are also very rich. The rubber in discarded tires has high elastic properties and good damping characteristics. It is of great significance to recycle discarded tires and develop rubber-sand concrete shock-absorbing materials. Shock-absorbing materials for underground engineering need to have good energy-absorbing and shock-absorbing properties while also having a certain strength to support the structure and stabilize it. Therefore, the uniaxial compressive strength of rubber-sand concrete materials is one of the important indicators for judging whether the material can be used as a shock-absorbing material for underground engineering. The prediction of the uniaxial compressive strength of rubber-sand concrete has important practical value and scientific significance.

[0003] At present, domestic and foreign scholars mostly use experimental research methods to study the mechanical properties of rubber-sand concrete materials. However, the large number of specimens required for the experiment and the lengthy test process are extremely time-consuming, labor-intensive and financially intensive. Some scholars have also used the method of constructing empirical formulas to predict the uniaxial compressive strength of rubber-sand concrete materials, but this method has the disadvantages of large discreteness and difficulty in ensuring the accuracy of the prediction results. With the leap of computer technology and the development of artificial intelligence algorithms, many studies have applied intelligent algorithms to the field of material mechanical property prediction. Common algorithms include artificial neural networks, random forests, extreme gradient boosting algorithms, etc., but these single intelligent algorithm models have problems such as slow learning speed and difficulty in meeting the requirements for prediction accuracy.

[0004] Therefore, it is urgent to propose a new intelligent prediction method in the field of uniaxial compressive strength prediction of rubber-sand concrete, which can realize efficient and accurate intelligent prediction of the uniaxial compressive strength of rubber-sand concrete and provide assistance for the development of new shock-absorbing materials. Summary of the invention

[0005] The purpose of the present invention is to address the above-mentioned problems existing in the prior art and provide a method for predicting the uniaxial compressive strength of rubber-sand concrete based on a hybrid intelligent model, so as to achieve efficient and accurate prediction of the uniaxial compressive strength of rubber-sand concrete and provide support for the research and development and performance research of new shock-absorbing materials.

[0006] The above-mentioned purpose of the present invention is achieved by the following technical means:

[0007] The method for predicting the uniaxial compressive strength of rubber sand concrete based on the hybrid intelligent model includes the following steps:

[0008] Step 1: Conduct a uniaxial compressive test on rubber-sand concrete, collect the actual uniaxial compressive strength of rubber-sand concrete corresponding to different experimental parameters as the actual output feature value, and record a set of experimental parameters corresponding to each uniaxial compressive test of rubber-sand concrete as the input feature group;

[0009] Step 2: Filter the types of input features of the input feature group, and use the filtered input feature group to build a training set and a test set;

[0010] Step 3: construct a back propagation neural network based on lion group optimization, and use the training set to train the back propagation neural network based on lion group optimization to obtain a trained back propagation neural network based on lion group optimization;

[0011] Step 4: Use the trained lion group optimized back propagation neural network to predict the input feature group to be processed and obtain the predicted output feature value.

[0012] As mentioned above, step 2 includes the following steps:

[0013] Step 2.1: construct an initial data set based on each input feature group obtained in step 1, and construct a shadow data set based on the initial data set;

[0014] Step 2.2: Use the prediction model to determine the importance scores of each category of input features in the shadow data set, and obtain the maximum importance score based on the importance scores of each category of input features, and set the maximum importance score as the importance score threshold;

[0015] Step 2.3: Use the prediction model to determine the importance scores of various input features in the initial data set, and filter out the types of input features in the initial data set whose importance scores are greater than the importance score threshold;

[0016] Step 2.4: According to the various input features screened out in the initial data set in step 2.3, draw the correlation coefficient diagram of various input features based on the Pearson intraclass correlation test, calculate the intraclass correlation coefficient of various input features, and traverse various input features in each input feature group of the initial data set. If the intraclass correlation coefficient between the input feature of the current traversal category and other categories is higher than the set intraclass correlation coefficient threshold, and the number of other categories higher than the set intraclass correlation coefficient threshold is greater than or equal to the set value, eliminate the input features of the traversal category, obtain the input feature group screened in the initial data set, and use the input feature group screened in the initial data set to construct the training set and test set.

[0017] Constructing shadow data from the initial dataset as described above includes:

[0018] The shadow data set is a two-dimensional matrix. Each column represents the corresponding input feature type. The input features in each row constitute an input feature group. The types of input features in the input feature group in the shadow data set are the same as those in the input feature group in the initial data set. The input feature values ​​of each type in the initial data set are randomly shuffled. For each type of randomly shuffled input feature, it is input into the input features of the corresponding column in the shadow data set in the shuffled order.

[0019] As described above, using the training set to train the back propagation neural network based on lion group optimization includes the following steps:

[0020] A lion group model of individual lions and hunting boundaries is constructed. The individual lions are respectively assigned a set of weights and thresholds. Based on the corresponding weights and thresholds of the individual lions, weights and thresholds are assigned to the back propagation neural network. The input feature group in the training set is input into the back propagation neural network. The back propagation neural network outputs the corresponding predicted output eigenvalues. The error value between the predicted output eigenvalues ​​and the corresponding actual output eigenvalues ​​is converted into fitness, and the fitness of each individual lion is calculated. The lion group model updates the weights and thresholds of the back propagation neural network by changing the number of individuals in the lion group. Finally, after all iterations are completed, the population size corresponding to the minimum fitness value that does not change before the iteration is completed or the change rate is less than the set threshold is recorded as the best lion group, and the weights and thresholds corresponding to the lion king in the best lion group are used as the optimal weights and optimal thresholds in the back propagation neural network model.

[0021] As described above, the back propagation neural network in step 3 includes an input layer, a hidden layer and an output layer.

[0022] The prediction method of uniaxial compressive strength of rubber sand concrete based on hybrid intelligent model also includes the use of root mean square error RMSE, correlation coefficient R, determination coefficient R 2 , mean absolute error MAE, mean square error MSE and square error SSE evaluate the difference between the predicted output eigenvalues ​​and the corresponding actual output eigenvalues ​​of the back-propagation neural network based on lion group optimization.

[0023] As mentioned above, the root mean square error RMSE is based on the following formula:

[0024]

[0025] The correlation coefficient R is based on the following formula:

[0026]

[0027] The coefficient of determination R 2 Based on the following formula:

[0028]

[0029] The mean absolute error MAE is based on the following formula:

[0030]

[0031] The mean square error MSE is based on the following formula:

[0032]

[0033] The squared error SSE is based on the following formula:

[0034]

[0035] Where n represents the number of input feature groups in the test set; and i Respectively represent the predicted output feature value corresponding to the input feature group and the actual output feature value corresponding to the input feature group; and They represent the average value of the predicted output feature values ​​corresponding to the input feature group and the average value of the actual output feature values ​​corresponding to the input feature group, respectively.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The present invention utilizes the lion group model to optimize the back propagation neural network to establish a hybrid intelligent prediction model, fully utilizes the characteristics of different intelligent algorithms, accelerates the training speed and improves the prediction accuracy, and provides a fast, efficient and accurate method for predicting the uniaxial compressive strength of rubber-sand concrete. The present invention has mature theory and concise algorithm, and is widely applicable and practical in this technical field. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of the present invention;

[0039] Figure 2 It is a schematic diagram of input feature selection of the present invention;

[0040] Among them, (a) is a graph of importance scores of various input features calculated by implementing steps 2.1-2.3 in Example 2; (b) is a graph of correlation coefficients between input features and between input features and prediction targets;

[0041] X1 is the mass of rubber; X2 is the mass of sand; X3 is the mass of cement; X4 is the rubber particle size; X5 is the mass of the specimen; X6 is the density of the specimen; X7 is the diameter of the specimen; X8 is the length of the specimen; Y1 is the uniaxial compressive strength of rubber-sand concrete;

[0042] Figure 3 This is a diagram of optimization results of a back propagation neural network based on lion group optimization according to Example 2 of the present invention;

[0043] Among them: (a) is the adaptability value of each lion individual during the optimization process; (b) is a curve chart of the optimal fitness value and iteration time obtained during the optimization process;

[0044] Figure 4 This is a comparison chart of the experimental results of Example 2 of the present invention;

[0045] Among them: (a) is a schematic diagram of the comparison of the actual uniaxial compressive strength of rubber-sand concrete and the predicted uniaxial compressive strength of rubber-sand concrete of different specimens; (b) is a schematic diagram of the prediction index of the back propagation neural network based on lion group optimization. DETAILED DESCRIPTION

[0046] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below in conjunction with embodiments. It should be understood that the implementation examples described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0047] Embodiment 1:

[0048] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting the uniaxial compressive strength of rubber sand concrete based on a hybrid intelligent model, comprising the following steps:

[0049] Step 1: Based on the uniaxial compressive test of rubber-sand concrete, the actual uniaxial compressive strength of rubber-sand concrete corresponding to different experimental parameters is collected to build a uniaxial compressive strength database of rubber-sand concrete. The experimental parameters affecting the uniaxial compressive strength of rubber-sand concrete are used as input features, and the actual uniaxial compressive strength of rubber-sand concrete is used as the actual output feature value; for each uniaxial compressive test of rubber-sand concrete, a set of experimental parameters involved is recorded as the input feature group.

[0050] In this embodiment, the types of input features of the input feature group include rubber quality, sand quality, cement quality and rubber particle size.

[0051] Step 2: Filter the types of input features of the input feature group, use the filtered input feature group to build training sets and test sets, and use the actual output feature values ​​corresponding to the input feature group as prediction targets.

[0052] More specifically, step 2 includes the following steps:

[0053] Step 2.1: The initial data set is constructed based on the input feature groups obtained in step 1, and the shadow data is constructed. The shadow data set is a two-dimensional matrix. Each column represents the corresponding input feature type, and each input feature in each row constitutes an input feature group. The types of input features in the input feature group in the shadow data set are the same as the types of input features in the input feature group in the initial data set. The input feature values ​​of each type in the initial data set are randomly shuffled. For each type of randomly shuffled input feature, they are input into the input features of the corresponding column in the shadow data set in the shuffled order. The types of input features in the input feature group in the shadow data set are kept consistent with the number of input feature types in the input feature group in the initial data set, and only the number of input feature groups in the original data set is expanded.

[0054] Step 2.2: Establish a prediction model (such as random forest) that can output importance scores, use the prediction model to determine the importance scores of each category of input features in the shadow dataset, and determine the maximum importance score based on the importance scores of each category of input features, and set the above maximum importance score as the importance score threshold.

[0055] Step 2.3: On the basis of the prediction model remaining unchanged, the prediction model is used to determine the importance scores of various input features in the initial data set, and the types of input features in the initial data set whose importance scores are greater than the importance score threshold are screened out.

[0056] Step 2.4: According to the various input features screened out in the initial data set in step 2.3, draw the correlation coefficient diagram of various input features based on the Pearson intraclass correlation test, calculate the intraclass correlation coefficients between various input features, and traverse various input features in each input feature group of the initial data set. If the intraclass correlation coefficient between the input feature of the current traversal category and other categories is higher than the set intraclass correlation coefficient threshold (0.7), and the number of other categories higher than the set intraclass correlation coefficient threshold (0.7) is greater than or equal to the set value (for example, 2, 3, etc.), remove the input features of the traversal category, obtain the input feature group screened in the initial data set, and use the input feature group screened in the initial data set to construct the training set and test set.

[0057] Step 3: Construct a back propagation neural network based on lion group optimization (hybrid intelligent model), and use the training set to train the back propagation neural network based on lion group optimization.

[0058] More specifically, in step 3, a back propagation neural network is constructed, which is connected in sequence by an input layer, a hidden layer, and an output layer; a lion group model of a certain number of lion individuals and a hunting boundary is constructed, and the lion individuals are respectively assigned a set of weights and thresholds. Based on the corresponding weights and thresholds of the lion individuals, weights and thresholds are assigned to the back propagation neural network, and the input feature group in the training set is input into the back propagation neural network. The back propagation neural network outputs the corresponding predicted output eigenvalue, that is, outputs the predicted uniaxial compressive strength of rubber-sand concrete, and realizes the prediction of the uniaxial compressive strength of rubber-sand concrete, and the error value between the predicted output eigenvalue and the corresponding actual output eigenvalue (that is, the actual uniaxial compressive strength of rubber-sand concrete) is converted into fitness (FF), the fitness of each lion individual is calculated, and the fitness of each lion individual is regarded as the fitness value of the back propagation neural network based on lion group optimization. Among them, the lion individual with the smallest fitness value is selected as the lion king, and the weights and thresholds corresponding to the lion king are used as the optimal weights and optimal thresholds of the back-propagation neural network based on lion group optimization; by changing the number of lion individuals, the lion group model updates the weights and thresholds of the back-propagation neural network according to its own logical rules within a limited iteration coefficient, that is, iterative calculation of the fitness (FF) value of the lion individual is carried out; finally, after all iterations are completed, the population size corresponding to the smallest fitness (FF) value that no longer changes before the iteration is completed (or the rate of change is less than the set threshold) is recorded as the best lion group, and the weights and thresholds corresponding to the lion king in the best lion group are used as the optimal weights and optimal thresholds in the back-propagation neural network model, thereby building a back-propagation neural network based on lion group optimization.

[0059] In this embodiment, the number of neurons in the input layer, hidden layer and output layer are 4, 10 and 1 respectively.

[0060] The calculation formula of the fitness (FF) is:

[0061]

[0062] Where w and b represent weight and threshold respectively; n is the total number of input feature groups in the training set; T i and P i are the actual output feature value and predicted output feature value of the i-th input feature group respectively.

[0063] The input feature group in the test set is input into the trained back-propagation neural network based on lion group optimization, and the corresponding predicted uniaxial compressive strength of rubber-sand concrete is obtained in real time.

[0064] The root mean square error (RMSE), correlation coefficient (R), and coefficient of determination (R 2), mean absolute error (MAE), mean square error (MSE) and square error (SSE) were used to evaluate the difference between the predicted output eigenvalues ​​and the corresponding actual output eigenvalues ​​of the back propagation neural network based on lion group optimization, and to verify the prediction performance of the back propagation neural network based on lion group optimization for the uniaxial compressive strength of rubber-sand concrete; the root mean square error (RMSE), correlation coefficient (R), determination coefficient (R 2 ), the calculation models of mean absolute error (MAE), mean square error (MSE) and square error (SSE) are:

[0065]

[0066]

[0067] Where n represents the number of input feature groups in the test set; and i They represent the predicted output eigenvalues ​​(i.e., predicted uniaxial compressive strength of rubber-sand concrete) corresponding to the input feature group and the actual output eigenvalues ​​(i.e., actual uniaxial compressive strength of rubber-sand concrete) corresponding to the input feature group, respectively; and They represent the average value of the predicted output feature values ​​corresponding to the input feature group and the average value of the actual output feature values ​​corresponding to the input feature group, respectively.

[0068] Step 5: Input the input feature group to be predicted of the same input feature type as that in the training set into the trained back propagation neural network based on lion group optimization to obtain the corresponding predicted uniaxial compressive strength of rubber-sand concrete in real time.

[0069] Example 2

[0070] A hybrid intelligent prediction method for uniaxial compressive strength of rubber-sand concrete based on back-propagation neural network and lion group optimization.

[0071] Among them, the types of input features in the input feature group include rubber mass, sand mass, cement mass, rubber particle size, specimen mass, specimen density, specimen diameter, and specimen length, a total of 8 types of input features.

[0072] Implement Example 1.

[0073] Figure 2 The importance score statistics of various input features in the initial data set after implementing steps 2.1 to 2.3 are shown. It can be seen that the importance scores of various input features in the initial data set are greater than the importance score threshold. The feature selection results show that the above input features all contribute to the prediction target and need to be retained.

[0074] After implementing step 2.4, it is shown that the internal correlation between the specimen quality and the rubber quality, sand quality and cement quality is much higher than the internal correlation coefficient between other types of input features, which means that the actual contribution of the specimen quality in the prediction process can be replaced by the above three parameters (rubber quality, sand quality and cement quality). Therefore, the 7 necessary input features (labeled as acceptable and low internal correlation) after screening, a total of 70 input feature groups containing the above input features, were used to establish the training set and test set. The training set includes 56 input feature groups, and the test set includes 14 input feature groups. The statistical information of different types of input features in the training set and test set is shown in Table 1:

[0075] Table 1 Statistics of different types of input features in the training set and test set

[0076]

[0077] In the process of optimizing the back propagation neural network through the lion group optimization algorithm, the total number of lion individuals and the maximum number of iterations are important parameters that determine the reliability of the back propagation neural network based on lion group optimization. Figure 3 It can be seen that in the convergence curve of the minimum fitness value during the 400 iterations, after 200 iterations, the minimum fitness value remains unchanged, which shows that the selected parameters can output stable prediction results. At the same time, considering the computational efficiency of the back propagation neural network based on lion group optimization, the maximum number of iterations of the back propagation neural network based on lion group optimization can be set to 200. In addition, the iteration time required to reach the optimal fitness value of the back propagation neural network based on lion group optimization increases with the increase in the size of lion individuals. Therefore, the size of the number of lion individuals can be determined according to the optimal fitness value, so that the back propagation neural network based on lion group optimization has the best accuracy. Figure 3 It can be determined that the optimal population size of the back propagation neural network based on lion group optimization is 75.

[0078] In step 4, the input feature group in the test set is input into the hybrid intelligent high-precision model of the training number (i.e., the back propagation neural network based on lion group optimization) to obtain the uniaxial compressive strength prediction result of rubber-sand concrete in real time.

[0079] Table 2 Different error indicators of back propagation neural network based on lion group optimization

[0080]

[0081] From the error index results, we can see that the R value of the hybrid intelligent prediction model is 0.97, R 2 The value is 0.91, which is very close to 1. Figure 4The correlation comparison analysis results of the hybrid intelligent model for the prediction of the uniaxial compressive strength of rubber-sand concrete and the experimental results show that the data points for predicting the uniaxial compressive strength of rubber-sand concrete are basically distributed near the best fitting line "predicted output eigenvalue = actual output eigenvalue", indicating that the prediction results of the back propagation neural network based on lion group optimization are accurate and reliable.

[0082] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A method for predicting the uniaxial compressive strength of rubber sand concrete based on a hybrid intelligent model, characterized in that: The following steps are involved: Step 1: Conduct a uniaxial compressive test on rubber-sand concrete, collect the actual uniaxial compressive strength of rubber-sand concrete corresponding to different experimental parameters as the actual output feature value, and record a set of experimental parameters corresponding to each uniaxial compressive test of rubber-sand concrete as the input feature group; Step 2: Filter the types of input features of the input feature group, and use the filtered input feature group to build a training set and a test set; Step 3: construct a back propagation neural network based on lion group optimization, and use the training set to train the back propagation neural network based on lion group optimization to obtain a trained back propagation neural network based on lion group optimization; Step 4: Use the trained lion group optimized back propagation neural network to predict the input feature group to be processed and obtain the predicted output feature value.

2. The method for predicting the uniaxial compressive strength of rubber sand concrete based on a hybrid intelligent model according to claim 1 is characterized in that: The step 2 comprises the following steps: Step 2.1: construct an initial data set based on each input feature group obtained in step 1, and construct a shadow data set based on the initial data set; Step 2.2: Use the prediction model to determine the importance scores of each category of input features in the shadow data set, and obtain the maximum importance score based on the importance scores of each category of input features, and set the maximum importance score as the importance score threshold; Step 2.3: Use the prediction model to determine the importance scores of various input features in the initial data set, and filter out the types of input features in the initial data set whose importance scores are greater than the importance score threshold; Step 2.4: According to the various input features screened out in the initial data set in step 2.3, draw the correlation coefficient diagram of various input features based on the Pearson intraclass correlation test, calculate the intraclass correlation coefficient of various input features, and traverse various input features in each input feature group of the initial data set. If the intraclass correlation coefficient between the input feature of the current traversal category and other categories is higher than the set intraclass correlation coefficient threshold, and the number of other categories higher than the set intraclass correlation coefficient threshold is greater than or equal to the set value, eliminate the input features of the traversal category, obtain the input feature group screened in the initial data set, and use the input feature group screened in the initial data set to construct the training set and test set.

3. The method for predicting the uniaxial compressive strength of rubber sand concrete based on a hybrid intelligent model according to claim 2 is characterized in that: The constructing of shadow data according to the initial data set comprises: The shadow data set is a two-dimensional matrix. Each column represents the corresponding input feature type. The input features in each row constitute an input feature group. The types of input features in the input feature group in the shadow data set are the same as those in the input feature group in the initial data set. The input feature values ​​of each type in the initial data set are randomly shuffled. For each type of randomly shuffled input feature, it is input into the input features of the corresponding column in the shadow data set in the shuffled order.

4. The method for predicting uniaxial compressive strength of rubber sand concrete based on hybrid intelligent model according to claim 3 is characterized in that: The method of training the back propagation neural network based on lion group optimization using the training set includes the following steps: A lion group model of individual lions and hunting boundaries is constructed. The individual lions are respectively assigned a set of weights and thresholds. Based on the corresponding weights and thresholds of the individual lions, weights and thresholds are assigned to the back propagation neural network. The input feature group in the training set is input into the back propagation neural network. The back propagation neural network outputs the corresponding predicted output eigenvalues. The error value between the predicted output eigenvalues ​​and the corresponding actual output eigenvalues ​​is converted into fitness, and the fitness of each individual lion is calculated. The lion group model updates the weights and thresholds of the back propagation neural network by changing the number of individuals in the lion group. Finally, after all iterations are completed, the population size corresponding to the minimum fitness value that does not change before the iteration is completed or the change rate is less than the set threshold is recorded as the best lion group, and the weights and thresholds corresponding to the lion king in the best lion group are used as the optimal weights and optimal thresholds in the back propagation neural network model.

5. The method for predicting uniaxial compressive strength of rubber sand concrete based on hybrid intelligent model according to claim 1 is characterized in that: The back propagation neural network in step 3 includes an input layer, a hidden layer and an output layer.

6. The method for predicting uniaxial compressive strength of rubber sand concrete based on hybrid intelligent model according to claim 5 is characterized in that: It also includes the use of root mean square error RMSE, correlation coefficient R, and determination coefficient R 2 , mean absolute error MAE, mean square error MSE and square error SSE evaluate the difference between the predicted output eigenvalues ​​and the corresponding actual output eigenvalues ​​of the back-propagation neural network based on lion group optimization.

7. The method for predicting uniaxial compressive strength of rubber sand concrete based on hybrid intelligent model according to claim 6 is characterized in that: The root mean square error RMSE is based on the following formula: The correlation coefficient R is based on the following formula: The coefficient of determination R 2 Based on the following formula: The mean absolute error MAE is based on the following formula: The mean square error MSE is based on the following formula: The squared error SSE is based on the following formula: Where n represents the number of input feature groups in the test set; and i Respectively represent the predicted output feature value corresponding to the input feature group and the actual output feature value corresponding to the input feature group; and They represent the average value of the predicted output feature values ​​corresponding to the input feature group and the average value of the actual output feature values ​​corresponding to the input feature group, respectively.

Citation Information

Patent Citations

  • Method for predicting concrete compressive strength based on random forest and intelligent algorithm

    CN112069567A

  • Power transmission and transformation project cost model construction method, system and device based on big data and medium

    CN116611785A

  • Building material engineering project budget measuring device, system and method

    CN116642516A

  • Reinforced concrete structure failure mode prediction method based on SSA-ELM

    CN117708924A

Cited By

  • Rubber sand mixed soil compression deformation prediction method and system

    CN119779837A