Reinforced concrete column earthquake damage mode prediction method, equipment, medium and product

Abstract

The invention discloses a reinforced concrete column earthquake damage mode prediction method and device, a medium and a product, and relates to the technical field of earthquake prediction, and the method comprises the steps: constructing a training set; expanding the characteristic parameter value set in the training set bending shear failure mode and the training set shear failure mode to obtain an expanded data set; removing outliers in the expanded data set to obtain a high-quality expanded data set; constructing a plurality of models with different kernels; the model comprises all feature parameters in a feature parameter set; determining a final model in all models according to all subsets of the feature parameter set, the training set, the expansion data set and the high-quality expansion data set; the final model is the model corresponding to the subset with the highest prediction precision of the bending-shear failure mode. The method and the device can solve the problem of low prediction accuracy of the bending-shear failure mode.

Description

Technical Field

[0001] The present application relates to the technical field of earthquake prediction, and in particular to a method, device, medium and product for predicting earthquake damage modes of reinforced concrete columns. Background Art

[0002] Reinforced Concrete (RC) columns are important vertical load-bearing and horizontal lateral-resisting members for buildings, bridges and underground structures. They undergo three different failure modes under earthquakes: bending failure, bending-shear failure and shear failure. The seismic performance of columns varies greatly under different failure modes. Bending failure has the strongest seismic performance, followed by bending-shear failure, and shear failure is the weakest. How to accurately predict the earthquake failure mode of reinforced concrete columns has become an important part of their seismic performance evaluation.

[0003] Bending failure mainly occurs when the shear span ratio is large, the axial compression ratio is small and the reinforcement is reasonable. The bending bearing capacity plays a controlling role. The failure process is that the tensile longitudinal reinforcement yields first. After experiencing a large plastic deformation, the concrete in the compression zone reaches the ultimate compressive strain. The horizontal force corresponding to the bending bearing capacity is always less than the shear bearing capacity. The failure characteristics are dense horizontal bending cracks in the plastic hinge area at the column end, and the longitudinal reinforcement yields to form plastic hinges. At the final failure, the concrete is obviously crushed. The whole process absorbs a large amount of seismic energy, which is a ductile failure.

[0004] Shear failure mainly occurs when the shear span ratio is small, the axial compression ratio is large, and the stirrups are insufficiently restrained. The shear bearing capacity plays a controlling role. The failure process is that the stirrups yield first, but the longitudinal reinforcement never yields. After the shear bearing capacity of the reinforced concrete column is reached, the deformation increases rapidly and the horizontal bearing capacity decreases rapidly. The failure characteristics are that the longitudinal reinforcement never yields, while the stirrups yield first or are even broken, and obvious shear slip oblique cracks appear, resulting in brittle failure. No plastic hinge is formed, which is a brittle failure.

[0005] Bending-shear failure is between the two. In the early stage, the bending bearing capacity plays a controlling role. In the later stage, brittle failure occurs due to insufficient shear bearing capacity. It has a certain ductility. The failure process is that as the deformation increases, the longitudinal reinforcement yields first. Due to the development of shear oblique cracks, the effective shear area of ​​the concrete is reduced, and the stirrups yield and shear brittle failure occurs. The failure characteristics are that the longitudinal reinforcement yields and forms plastic hinges. After a certain plastic deformation, the stirrups yield or even break. The concrete strain has not reached the limit value at the time of failure. Due to the limitations of objective conditions and the complexity of its failure mechanism, the occurrence of bending-shear failure cannot be completely avoided in actual design. Therefore, accurate prediction of bending-shear failure has become the most important part of the earthquake failure mode of reinforced concrete columns. Summary of the invention

[0006] The purpose of this application is to provide a method, device, medium and product for predicting earthquake damage modes of reinforced concrete columns, which can solve the problem of low accuracy in predicting bending-shear damage modes.

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

[0008] In a first aspect, the present application provides a method for predicting earthquake damage modes of reinforced concrete columns, comprising:

[0009] Constructing a training set; the training set includes a plurality of characteristic parameter value sets corresponding to sample reinforced concrete columns and seismic failure modes of reinforced concrete columns corresponding to each characteristic parameter value set; the seismic failure modes of reinforced concrete columns include bending failure mode, bending-shear failure mode and shear failure mode; the characteristic parameter value set includes the value of each characteristic parameter in the characteristic parameter set;

[0010] Expanding the characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set to obtain an expanded data set;

[0011] Remove outliers from the expanded data set to obtain a high-quality expanded data set;

[0012] Constructing a plurality of models with different kernels; the models include all feature parameters in the feature parameter set;

[0013] A final model is determined among all models according to all subsets, training sets, expanded data sets and high-quality expanded data sets of the characteristic parameter set; the final model is the model corresponding to the subset with the highest prediction accuracy of the bending-shear failure mode, and the final model is used to predict the seismic failure mode of reinforced concrete columns.

[0014] Optionally, the step of expanding the characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set to obtain an expanded data set specifically includes:

[0015] Processing the characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set to obtain an expanded characteristic parameter value set under the bending-shear failure mode and an expanded characteristic parameter value set under the shear failure mode;

[0016] The extended characteristic parameter value set under the bending-shear failure mode, the extended characteristic parameter value set under the shear failure mode, and the characteristic parameter value set under the bending failure mode in the training set are determined as an extended data set.

[0017] Optionally, removing outliers in the expanded data set to obtain a high-quality expanded data set specifically includes:

[0018] The characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set are deleted from the characteristic parameter value set under the expanded bending-shear failure mode and the characteristic parameter value set under the expanded shear failure mode to obtain a target parameter set;

[0019] Eliminate outliers from the target parameter set according to the training set to obtain a characteristic parameter value set under the bending-shear failure mode after elimination and a characteristic parameter value set under the shear failure mode after elimination;

[0020] It is determined that the characteristic parameter value set under the bending-shear failure mode after elimination, the characteristic parameter value set under the shear failure mode after elimination and the training set are high-quality expanded data sets.

[0021] Optionally, a final model is determined among all models based on all subsets of the feature parameter set, the training set, the augmented data set, and the high-quality augmented data set, specifically including:

[0022] Determine the best model among all models based on the training set, augmented dataset, and high-quality augmented dataset;

[0023] Construct a subset set, and construct a data set corresponding to each subset in the subset set based on the high-quality expanded data set; the subset set includes all subsets of the feature parameter set;

[0024] The data sets corresponding to each subset are input into the optimal model to obtain the prediction accuracy of the bending-shear failure mode corresponding to each subset, and the characteristic parameters in the optimal model are simplified according to the subset with the highest prediction accuracy of the bending-shear failure mode to obtain a simplified model;

[0025] The simplified model is hyperparameter optimized to obtain the final model.

[0026] Optionally, the best model among all models is determined based on the training set, the augmented data set, and the high-quality augmented data set, specifically including:

[0027] For any model, a ten-fold equal-proportional cross-validation method is used to determine the prediction accuracy of the model on the training set, the expanded data set, and the high-quality expanded data set;

[0028] Determine the optimal data set according to the prediction accuracy of each model on the training set, the expanded data set and the high-quality expanded data set; the optimal data set is one of the training set, the expanded data set and the high-quality expanded data set;

[0029] The model with the highest prediction accuracy on the optimal data set is determined to be the optimal model.

[0030] Optionally, a data set corresponding to each subset in the subset set is constructed according to the high-quality expanded data set, specifically including:

[0031] For any subset set, the values ​​of all feature parameters included in the subset in the high-quality expanded data set are determined as the data set corresponding to the subset.

[0032] Optionally, the simplified model is subjected to hyperparameter optimization to obtain a final model, specifically including:

[0033] The random forest optimization algorithm is used to optimize the hyperparameters of the simplified model to obtain the final model.

[0034] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for predicting earthquake damage modes of reinforced concrete columns.

[0035] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for predicting earthquake damage modes of reinforced concrete columns.

[0036] In a fourth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for predicting earthquake damage modes of reinforced concrete columns.

[0037] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0038] The present application provides a method, device, medium and product for predicting earthquake damage modes of reinforced concrete columns. The method generates high-quality expanded data by expanding minority data (a set of characteristic parameter values ​​under bending-shear damage mode and shear damage mode) and eliminating outliers from the expanded data. The amount of minority data is increased while ensuring data quality, thereby solving the problem of low accuracy caused by the small number of bending-shear damage sample data. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 This is a flow chart of a method for predicting earthquake damage modes of reinforced concrete columns in one embodiment of the present application;

[0041] Figure 2This is a schematic diagram of a method for predicting earthquake damage modes of reinforced concrete columns in one embodiment of the present application;

[0042] Figure 3 It is a scatter plot of some feature parameters in the training set;

[0043] Figure 4 Scatter plot of some feature parameters in the expanded data set;

[0044] Figure 5 Scatter plots of some feature parameters in the high-quality augmented dataset;

[0045] Figure 6 It is a scatter plot of some characteristic parameters of the bending-shear failure mode in the training set;

[0046] Figure 7 Scatter plots of some characteristic parameters in the bending-shear failure mode in the expanded data set;

[0047] Figure 8 Scatter plots of some characteristic parameters in the bending-shear failure mode in the high-quality expansion dataset;

[0048] Fig. 9 It is a scatter plot of some characteristic parameters of shear failure mode in the training set;

[0049] Fig.10 Scatter plots of some characteristic parameters in the shear failure mode in the expanded data set;

[0050] Fig.11 Scatter plots of some characteristic parameters in shear failure mode in the high-quality expansion dataset;

[0051] Fig.12 When the model is a GPC model, the prediction result diagram of the optimal model in the training set;

[0052] Fig.13 This is the prediction result diagram of the optimal model in the expanded data set when the model is a GPC model;

[0053] Fig.14 When the model is a GPC model, the prediction result diagram of the optimal model in the high-quality expanded data set;

[0054] Fig.15 When the model is a GPC model, the predicted result diagram of the simplified model;

[0055] Fig.16 This is a feature density scatter plot of the interpretability analysis results of the SHAP algorithm;

[0056] Fig.17 It is the absolute average influence map of SHAP value;

[0057] Fig.18 It is the single sample feature influence diagram of SHAP algorithm;

[0058] Fig.19 Feature dependency graph for the interpretability analysis results of the SHAP algorithm;

[0059] Fig. 20 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0061] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0062] In an exemplary embodiment, a method for predicting earthquake damage modes of reinforced concrete columns is provided. Figure 1 As shown, the method comprises the following steps:

[0063] Step 201: Construct a training set. The scatter plot of some feature parameters in the training set is as follows: Figure 3 As shown; the training set includes multiple feature parameter value sets corresponding to the sample reinforced concrete columns and the seismic failure modes of the reinforced concrete columns corresponding to each feature parameter value set; the seismic failure modes of the reinforced concrete columns include bending failure mode, bending-shear failure mode and shear failure mode; the feature parameter value set includes the value of each feature parameter in the feature parameter set. Preferably, each feature parameter in the feature parameter set is a design parameter of reinforced concrete, including the basic information of the reinforced concrete column, which may specifically be: concrete compressive strength (f'c), longitudinal reinforcement yield strength (fyl), stirrup yield strength (fyt), shear span ratio (λ), stirrup spacing / section effective height (s / d) in the densified area, axial compression ratio (n), longitudinal reinforcement ratio (ρ l ), stirrup reinforcement ratio (ρ sv ), longitudinal reinforcement characteristic parameters (α l ) and stirrup characteristic parameters (α sv ).

[0064] Step 202: Expand the characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set to obtain an expanded data set. The scatter plot of some characteristic parameters in the expanded data set is shown in Figure 4The scatter plot of some characteristic parameters in the bending-shear failure mode in the training set is shown in Figure 6 As shown in the figure, the scatter plots of some characteristic parameters in the shear failure mode in the training set are as follows Fig. 9 shown.

[0065] Step 203: Remove outliers in the expanded data set to obtain a high-quality expanded data set. The scatter plot of some feature parameters in the high-quality expanded data set is as follows: Figure 5 As shown in the figure, the scatter plots of some characteristic parameters under the bending-shear failure mode in the high-quality expanded data set are shown in Figure 8 As shown in the figure, the scatter plots of some characteristic parameters under the shear failure mode in the high-quality expanded data set are shown in Fig.11 shown.

[0066] Step 204: construct multiple models with different kernels; the models include all feature parameters in the feature parameter set.

[0067] Step 205: Determine a final model among all models based on all subsets of the characteristic parameter set, the training set, the expanded data set, and the high-quality expanded data set; the final model is the model corresponding to the subset with the highest prediction accuracy of the bending-shear failure mode, and the final model is used to predict the earthquake failure mode of reinforced concrete columns.

[0068] Execute the above steps 201 to 205, generate high-quality expanded data by expanding the characteristic parameter value sets under the bending-shear failure mode and the shear failure mode and removing outliers from the expanded data, increase the number of minority class data while ensuring data quality, and solve the problem of low accuracy caused by small number of bending-shear failure sample data.

[0069] In practical applications, constructing a training set includes:

[0070] Construct an RC column test database. Collect test data of reinforced concrete rectangular columns under low-cycle repeated loads from the Pacific Earthquake Engineering Research Center Structural Information Database (PEER Database) and doctoral and master's degree theses.

[0071] Table 1 Experimental data of rectangular reinforced concrete columns

[0072]

[0073]

[0074] Initially select characteristic parameters and construct data sets. Extract and calculate the values ​​of required characteristic parameters from the RC column test database, obtain the seismic failure mode of rectangular reinforced concrete columns from the RC column test database, and the above two constitute the rectangular reinforced concrete column test data set. The values ​​of characteristic parameters in the rectangular reinforced concrete column test data set are dimensionlessly standardized, and then the seismic failure mode in the rectangular reinforced concrete column test data set is digitized. The output variables are adjusted as follows: bending failure mode is recorded as 1, bending-shear failure mode is recorded as 2, and shear failure mode is recorded as 3. The newly quantized rectangular reinforced concrete column test data set is obtained, and the newly quantized rectangular reinforced concrete column test data set is randomly divided into training set and test set in a ratio of 7:3.

[0075] In practical applications, the prediction methods of seismic damage modes of reinforced concrete columns also include:

[0076] The final model is used to predict the test set, and the SHAP algorithm is used to perform interpretability analysis on the prediction results of the final model. The interpretability analysis results are as follows: Figure 16 to Figure 19 shown. Fig.16 The color in the middle represents the numerical value of the characteristic parameter, and the color gradient (from blue to red) represents the characteristic parameter value from low to high. The larger the SHAP value, the greater the degree to which the characteristic parameter will increase the model prediction value, that is, the larger the SHAP value, the greater the possibility of shear brittle failure of RC columns caused by the characteristic parameter; the smaller the SHAP value, the greater the degree to which the characteristic parameter will reduce the model prediction value, that is, the smaller the SHAP value, the greater the probability of bending failure of RC columns caused by the characteristic parameter. The horizontal position of the scatter point indicates the magnitude and direction of the influence of the characteristic parameter on the predicted value. The farther the scatter point is from the 0 axis, the greater the influence of the characteristic parameter on the scatter point. The distribution width of the scatter point on a certain characteristic parameter row indicates the range of variation of the influence of the characteristic parameter on the predicted value. The wider the distribution, the greater the difference in the influence of the characteristic parameter on different samples. The higher the position of the characteristic parameter in the figure, the more important the characteristic parameter is. The characteristic parameter at the top has the greatest influence on the model prediction value. By observing the distribution of the characteristics under different colors, we can understand the specific influence of the size of the characteristic parameter on the model prediction value. Fig.16 It can be seen that the most important characteristic parameter is the shear span ratio λ, and the SHAP value is distributed between -0.6 and 1. The red sample points are mainly concentrated in the negative area, indicating that when the shear span ratio is high, the predicted value of the model will decrease, that is, bending failure tends to occur. The blue sample points are mainly concentrated in the positive area, indicating that when the shear span ratio is low, the predicted value of the model will increase, that is, bending shear failure or shear failure tends to occur. The scatter distribution is wide, indicating that the shear span ratio has a large difference in the impact on different samples. The longitudinal reinforcement ratio ρ has the least impact on the predicted value among the characteristic parameters l, SHAP values ​​are distributed between -0.2 and 0.2, and the red and blue points are evenly distributed, indicating that the influence of the reinforcement ratio on the model prediction value has no significant directionality. The scattered distribution is relatively concentrated, indicating that the influence of the reinforcement ratio on most samples is relatively stable. Other features such as stirrup reinforcement ratio ρ sv , longitudinal reinforcement yield strength fyl, etc. have a certain influence on the model prediction value and are relatively stable. Fig.17 It can be seen that compared with other characteristic parameters, the shear span ratio has the most significant impact on the model prediction value. Fig.18 Used to explain the predicted value of a specific sample. The horizontal axis represents the predicted value output by the model, and the value increases from left to right. The vertical axis represents the actual value of each feature parameter. The red arrow indicates that the feature parameter will increase the predicted value, and the blue arrow will decrease the predicted value. The SHAP value on the arrow corresponding to the feature parameter indicates the degree of influence of the feature parameter on the predicted value. Fig.18 The E[f(x)] in represents the baseline prediction value of the model, which is the average of all sample prediction values. Fig.18 It can be seen that the shear span ratio λ and stirrup reinforcement ratio ρ sv , longitudinal reinforcement yield strength fyl, longitudinal reinforcement ratio ρ l The characteristic parameters will increase the predicted value of the model (from bending failure to bending-shear failure and shear failure), while the axial compression ratio n, the ratio of the stirrup spacing in the densified area to the effective height of the section s / d, and the stirrup yield strength fyt will reduce the predicted value of the model (from shear failure to bending-shear failure and bending failure). Among them, the shear span ratio λ has the most significant effect on increasing the predicted value of the model, and the axial compression ratio n has the most significant effect on reducing the predicted value of the model. Fig.19 Used to explain the interaction between characteristic parameters. The horizontal axis represents the shear span ratio λ, the left vertical axis represents the SHAP value of the shear span ratio, and the right vertical axis is the name of the comparative characteristic parameter. The color from blue to red represents the different values ​​of the characteristic parameter on the right vertical axis. Blue means a higher stirrup reinforcement ratio, and red means a lower stirrup reinforcement ratio. Fig.19 It can be seen that the shear span ratio has a significant impact on the overall prediction value of the model. Specifically, as the shear span ratio increases, the SHAP value shows a downward trend, that is, the gradual increase in the shear span ratio causes the model's prediction value to gradually decrease (from shear failure to bending-shear failure and bending failure). The stirrup reinforcement ratio has a regulating effect on the shear span ratio: when the stirrup reinforcement ratio is low, the shear span ratio has a more significant effect on increasing the model's prediction value, and when the stirrup reinforcement ratio is high, the shear span ratio has a more significant effect on reducing the model's prediction value.

[0077] In practical applications, the feature parameter value set under the bending-shear failure mode in the training set and the feature parameter value set under the shear failure mode in the training set are expanded to obtain an expanded data set, specifically including:

[0078] The feature parameter value set under the bending-shear failure mode in the training set and the feature parameter value set under the shear failure mode in the training set are processed to obtain the feature parameter value set under the expanded bending-shear failure mode and the feature parameter value set under the expanded shear failure mode. Specifically, the SVMSMOTE algorithm (Synthetic Minority Oversampling Technique under Support Vector Machine) can be used for processing. The SVMSMOTE algorithm is a variant of the SMOTE algorithm (Synthetic Minority Oversampling Technique), which focuses on generating minority class data along the decision boundary (classification boundary), and the generated minority class data is more discriminative. The scatter plot of some feature parameters under the bending-shear failure mode in the expanded data set is shown in the figure below. Figure 7 As shown in the figure, the scatter plots of some characteristic parameters under the shear failure mode in the expanded data set are shown in Fig.10 shown.

[0079] It is determined that the characteristic parameter value set under the expanded bending-shear failure mode and the characteristic parameter value set under the expanded shear failure mode and the training set are the expanded data sets.

[0080] In practical applications, the removal of outliers in the expanded data set to obtain a high-quality expanded data set specifically includes:

[0081] The target parameter set is obtained by deleting the characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set from the characteristic parameter value set under the expanded bending-shear failure mode and the characteristic parameter value set under the expanded shear failure mode.

[0082] According to the training set, outliers are eliminated from the target parameter set to obtain a characteristic parameter value set under the bending-shear failure mode after elimination and a characteristic parameter value set under the shear failure mode after elimination. The spinous processes can be eliminated using a weighted KNN algorithm (K-nearest neighbor algorithm, K-NearestNeighbor).

[0083] It is determined that the characteristic parameter value set under the bending-shear failure mode after elimination and the characteristic parameter value set under the shear failure mode after elimination and the training set are high-quality expanded data sets.

[0084] In practical applications, outliers are eliminated from the target parameter set according to the training set to obtain a characteristic parameter value set in the bending-shear failure mode after elimination and a characteristic parameter value set in the shear failure mode after elimination, specifically:

[0085] The weighted KNN algorithm is adopted to eliminate the characteristic parameter value set under the bending-shear failure mode in the target parameter set according to the data characteristics of the characteristic parameter value set under the bending-shear failure mode in the training set to obtain the eliminated characteristic parameter value set under the bending-shear failure mode.

[0086] The weighted KNN algorithm is adopted to eliminate the characteristic parameter value set under the shear failure mode in the target parameter set according to the data characteristics of the characteristic parameter value set under the shear failure mode in the training set to obtain the characteristic parameter value set under the shear failure mode after elimination.

[0087] In practical applications, the model is a GPC (Gaussian process classification) model, and building multiple models with different kernels is to build multiple GPC models with different kernels, specifically:

[0088] To build the GPC model framework, the GPC model requires a certain kernel function (kernel). Existing single kernels include radial basis function kernel, Matérn kernel, rational quadratic kernel, white noise kernel, etc. Different kernels can capture different characteristics of data. 37 different combined kernels are created by adding or dot multiplying single kernels. The combined kernel can give full play to the advantages of different single kernels in capturing different characteristics of data and capture as many characteristics of data as possible. The combined kernel and the four single kernels form a kernel group to obtain 41 GPC models with different kernels.

[0089] In practical applications, the final model is determined among all models based on all subsets of feature parameter sets, training sets, augmented data sets, and high-quality augmented data sets, including:

[0090] Determine the best model among all models based on the training set, augmented dataset, and high-quality augmented dataset.

[0091] Construct a subset set, and construct a data set corresponding to each subset in the subset set based on the high-quality expanded data set; the subset set includes all subsets of the feature parameter set.

[0092] The data sets corresponding to each subset are input into the optimal model to obtain the prediction accuracy of the bending-shear failure mode corresponding to each subset. The characteristic parameters in the optimal model are simplified according to the subset with the highest prediction accuracy of the bending-shear failure mode to obtain the simplified model. When the model is the GPC model, the prediction results of the optimal model in the training set are shown in the figure below: Fig.12 As shown; when the model is the GPC model, the prediction results of the optimal model in the expanded data set are shown in Fig.13 As shown; when the model is the GPC model, the prediction results of the optimal model in the high-quality expanded data set are shown in Fig.14 As shown, when the model is a GPC model, the prediction results of the simplified model are shown in Fig.15shown.

[0093] The simplified model is hyperparameter optimized to obtain the final model.

[0094] In practical applications, based on the training set, the expanded data set, and the high-quality expanded data set, combined with the ten-fold equal-proportional cross-validation method, the prediction results of the models corresponding to all kernels in the kernel group are calculated and the optimal kernel on each data set is found, and a longitudinal comparison between data sets is performed: the prediction results of the model corresponding to the same kernel on the original data, the expanded data set, and the high-quality expanded data set are used to explore the effect of the SVMSMOTE algorithm and the SVMSMOTE-weighted KNN algorithm on data enhancement. A horizontal comparison between kernels is performed, and the prediction results of the model corresponding to the kernel group on the same data set are used to explore the effect of the combined kernel on algorithm enhancement. The optimal data set - the high-quality expanded data set is found through longitudinal comparison, and the optimal kernel under the optimal data set is found through horizontal comparison to establish a model. Therefore, the optimal model among all models is determined based on the training set, the expanded data set, and the high-quality expanded data set, including:

[0095] For any model, the ten-fold equal-proportional cross-validation method is used to determine the prediction accuracy of the model on the training set, the expanded data set, and the high-quality expanded data set.

[0096] The optimal data set is determined according to the prediction accuracy of each model on the training set, the expanded data set and the high-quality expanded data set; the optimal data set is one of the training set, the expanded data set and the high-quality expanded data set.

[0097] The model with the highest prediction accuracy on the optimal data set is determined to be the optimal model.

[0098] In practical applications, the datasets corresponding to each subset in the subset set are constructed based on the high-quality expanded dataset, including:

[0099] For any subset set, the values ​​of all feature parameters included in the subset in the high-quality expanded data set are determined as the data set corresponding to the subset.

[0100] In practical applications, the data sets corresponding to each subset are input into the optimal model to obtain the prediction accuracy of the bending-shear failure mode corresponding to each subset. The characteristic parameters in the model are simplified according to the subset with the highest prediction accuracy of the bending-shear failure mode to obtain the simplified model, which is as follows:

[0101] According to the optimal model, the exhaustive search strategy is used to eliminate the feature parameters with small influence to simplify the model while ensuring the prediction accuracy of the model. Specific steps: According to the exhaustive search strategy, for a feature parameter set containing k feature parameters, there are 2 k-1 non-empty feature parameter candidate subset. Taking k=10 as an example, first, create 2 10 -1 has a total of 1023 feature subsets; then create a data set of the corresponding subset under the high-quality expanded data set; then select the optimal subset of bending-shear failure modes based on the bending-shear failure prediction accuracy and create a simplified model based on the feature parameters in the optimal subset.

[0102] In practical applications, the simplified model is subjected to hyperparameter optimization to obtain the final model, including:

[0103] The random forest optimization algorithm is used to optimize the hyperparameters of the simplified model to obtain the final model.

[0104] Due to the complex mechanical mechanism and numerous influencing factors of the failure mode of reinforced concrete columns under earthquake action, the research on the bending failure mechanism is relatively in-depth, while the shear force is relatively complex, and the research on the failure mechanism of bending-shear and shear failure is still insufficient. At present, there is no unified failure mode prediction method at home and abroad. The known methods usually include traditional prediction methods and machine learning algorithms. The former is based on mechanical mechanisms and engineering experience to drive empirical or semi-empirical formulas based on characteristic parameters (influencing factors). For example, according to the design parameters in the test database (reinforced concrete column size, stirrup longitudinal reinforcement configuration, loading axial force, etc.), the required characteristic parameters (shear capacity demand ratio, shear span ratio, stirrup reinforcement ratio) are calculated, and the discrimination criteria of the earthquake failure mode of the reinforced concrete column are established according to the characteristic parameters. According to the discrimination criteria, the probability of the occurrence of three failure modes of the reinforced concrete column to be judged under the earthquake action is determined, and the failure mode with the highest probability is taken as the dominant earthquake failure mode of the reinforced concrete column. However, the prediction accuracy of the bending-shear failure mode is low, and the bending-shear failure cannot be effectively identified. The latter mines the potential information of data based on algorithms. In the existing machine learning algorithm technology, web crawler technology is used to build a reinforced concrete experimental database, select characteristic parameters and build a deep neural network model, train the model and optimize the model. Although the model has a high prediction accuracy for bending failure, it cannot effectively identify bending-shear failure, and the prediction accuracy for bending-shear failure is only 66%. Based on the Bayesian parameter estimation method, the simplified model is obtained by gradually eliminating unimportant characteristic parameters combined with the coefficient of variation. The frequency-based probability method is used to divide the boundaries of the continuous values ​​predicted by the model, and the bending failure interval, bending-shear failure interval and shear failure interval are divided to predict the failure mode of reinforced concrete. The model has strong interpretability, but the overall prediction accuracy and the prediction accuracy of the bending-shear failure mode are low. The prediction accuracy for bending-shear failure is only 50% (rectangle) and (58%) circle, respectively, and both cannot effectively identify bending-shear failure. Both traditional prediction methods and machine learning algorithms are based on experimental data, but the experimental data itself has certain class imbalance problems (i.e., the amount of bending failure data is large, and the amount of bending-shear and shear failure data is small). In addition, bending-shear failure is between bending and shear failure, and its data has certain similarities with the other two categories. As a result, the prediction accuracy of bending-shear failure in the prediction results is much lower than that of the other two categories, and accurate prediction cannot be made.

[0105] like Figure 2The method for predicting earthquake damage modes of reinforced concrete columns provided in the present application is shown as follows: a training set and a test set are obtained based on an RC column experimental database; a minority class data in the training set (a set of characteristic parameter values ​​under the bending-shear damage mode and a set of characteristic parameter values ​​under the shear damage mode in the training set) are subjected to SVMSMTOE oversampling to obtain an expanded set of characteristic parameter values ​​under the bending-shear damage mode and an expanded set of characteristic parameter values ​​under the shear damage mode; the expanded set of characteristic parameter values ​​under the bending-shear damage mode, the expanded set of characteristic parameter values ​​under the shear damage mode and the majority class data in the training set (the set of characteristic parameter values ​​under the bending damage mode) are determined as an expanded data set; a KNN elimination process is performed on the expanded set of characteristic parameter values ​​under the bending-shear damage mode and the expanded set of characteristic parameter values ​​under the shear damage mode to obtain a high-quality expanded data set; a GPC model is established based on the high-quality expanded data set, a ten-fold equal-proportional cross-validation method is used to determine the final model, the final model is used to predict the damage mode of RC columns, and a SHAP interpretability analysis is performed on the final model in combination with the prediction set. This application combines the SVMSMOTE algorithm with the weighted KNN algorithm to enhance the data to form a high-quality expanded data set, increase the number of samples of bending-shear failure and shear failure, effectively solve the problem of insufficient bending-shear failure test data, and improve the accuracy of bending-shear failure prediction. Moreover, the GPC algorithm model is a non-parametric model, which is very flexible when processing complex data sets. It can automatically adjust the complexity of the model, adapt to various types of data structures and capture different features in the data by selecting and combining different kernel functions, create a combined kernel to find the optimal kernel for the seismic failure mode of reinforced concrete columns, and enhance the algorithm's sensitivity to data. It effectively solves the problem of low accuracy in the prediction of bending-shear failure.

[0106] Existing machine learning algorithm technology can not only mine potential information in data to predict results, but also analyze the intrinsic relationship between damage patterns and feature parameters, but the prediction results lack certain interpretability. The SHAP algorithm is an algorithm used to explain the output of machine learning models. It is based on the Shapley value in cooperative game theory and assigns a SHAP value to each feature parameter to indicate the contribution of the feature parameter to the model prediction results. The main steps include baseline prediction, feature contribution calculation, Shapely value assignment, etc. The SHAP algorithm is combined to perform interpretability analysis on the prediction results, explore the intrinsic relationship between feature parameters and prediction results, and effectively solve the problem of the lack of interpretability of prediction results.

[0107] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig. 20As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store training sets. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting earthquake damage modes of reinforced concrete columns is implemented.

[0108] Those skilled in the art will understand that Fig. 20 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above-mentioned method embodiments when executing the computer program.

[0109] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the above-mentioned method embodiments are implemented.

[0110] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the above-mentioned method embodiments are implemented.

[0111] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0112] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0113] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0114] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0115] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting earthquake damage modes of reinforced concrete columns, characterized in that: The method for predicting earthquake damage modes of reinforced concrete columns comprises: Constructing a training set; the training set includes a plurality of characteristic parameter value sets corresponding to sample reinforced concrete columns and seismic failure modes of reinforced concrete columns corresponding to each characteristic parameter value set; the seismic failure modes of reinforced concrete columns include bending failure mode, bending-shear failure mode and shear failure mode; the characteristic parameter value set includes the value of each characteristic parameter in the characteristic parameter set; Expanding the characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set to obtain an expanded data set; Remove outliers from the expanded data set to obtain a high-quality expanded data set; Constructing a plurality of models with different kernels; the models include all feature parameters in the feature parameter set; A final model is determined among all models according to all subsets, training sets, expanded data sets and high-quality expanded data sets of the characteristic parameter set; the final model is the model corresponding to the subset with the highest prediction accuracy of the bending-shear failure mode, and the final model is used to predict the seismic failure mode of reinforced concrete columns.

2. The method for predicting earthquake damage modes of reinforced concrete columns according to claim 1, characterized in that: The method of expanding the characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set to obtain the expanded data set specifically includes: Processing the characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set to obtain an expanded characteristic parameter value set under the bending-shear failure mode and an expanded characteristic parameter value set under the shear failure mode; The extended characteristic parameter value set under the bending-shear failure mode, the extended characteristic parameter value set under the shear failure mode, and the characteristic parameter value set under the bending failure mode in the training set are determined as an extended data set.

3. The method for predicting earthquake damage modes of reinforced concrete columns according to claim 2, characterized in that: The step of removing outliers from the expanded data set to obtain a high-quality expanded data set specifically includes: The characteristic parameter value set under the bending-shear failure mode in the training set and the characteristic parameter value set under the shear failure mode in the training set are deleted from the characteristic parameter value set under the expanded bending-shear failure mode and the characteristic parameter value set under the expanded shear failure mode to obtain a target parameter set; Eliminate outliers from the target parameter set according to the training set to obtain a characteristic parameter value set under the bending-shear failure mode after elimination and a characteristic parameter value set under the shear failure mode after elimination; It is determined that the characteristic parameter value set under the bending-shear failure mode after elimination, the characteristic parameter value set under the shear failure mode after elimination and the training set are high-quality expanded data sets.

4. The method for predicting earthquake damage modes of reinforced concrete columns according to claim 1, characterized in that: The final model is determined among all models based on all subsets of feature parameter sets, training sets, augmented datasets, and high-quality augmented datasets, including: Determine the best model among all models based on the training set, augmented dataset, and high-quality augmented dataset; Construct a subset set, and construct a data set corresponding to each subset in the subset set based on the high-quality expanded data set; the subset set includes all subsets of the feature parameter set; The data sets corresponding to each subset are input into the optimal model to obtain the prediction accuracy of the bending-shear failure mode corresponding to each subset, and the characteristic parameters in the optimal model are simplified according to the subset with the highest prediction accuracy of the bending-shear failure mode to obtain a simplified model; The simplified model is hyperparameter optimized to obtain the final model.

5. The method for predicting earthquake damage modes of reinforced concrete columns according to claim 4, characterized in that: Determine the best model among all models based on the training set, augmented dataset, and high-quality augmented dataset, including: For any model, a ten-fold equal-proportional cross-validation method is used to determine the prediction accuracy of the model on the training set, the expanded data set, and the high-quality expanded data set; Determine the optimal data set according to the prediction accuracy of each model on the training set, the expanded data set and the high-quality expanded data set; the optimal data set is one of the training set, the expanded data set and the high-quality expanded data set; The model with the highest prediction accuracy on the optimal data set is determined to be the optimal model.

6. The method for predicting earthquake damage modes of reinforced concrete columns according to claim 4, characterized in that: The datasets corresponding to each subset in the subset set are constructed based on the high-quality expanded dataset, including: For any subset set, the values ​​of all feature parameters included in the subset in the high-quality expanded data set are determined as the data set corresponding to the subset.

7. The method for predicting earthquake damage modes of reinforced concrete columns according to claim 4, characterized in that: The simplified model is optimized for hyperparameters to obtain the final model, including: The random forest optimization algorithm is used to optimize the hyperparameters of the simplified model to obtain the final model.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting earthquake damage modes of reinforced concrete columns according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting earthquake damage modes of reinforced concrete columns according to any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for predicting earthquake damage modes of reinforced concrete columns according to any one of claims 1 to 7 is implemented.

Images