Lightgbm-based concrete chloride ion diffusion coefficient evaluation method under LDHs-containing corrosion inhibitor

Through the evaluation method of chlorine ion diffusion coefficient of concrete under LDHs-containing corrosion inhibitor based on lightgbm, the problem of low prediction accuracy of chlorine ion diffusion coefficient in marine engineering is solved, and efficient and reliable concrete structure analysis and design is achieved.

CN119007895AInactive Publication Date: 2024-11-22CCCC FOURTH HARBOR ENG INST CO LTD +1

Patent Information

Application Number
CN202411472241.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-22
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art cannot effectively predict the chloride ion diffusion coefficient of concrete in the marine environment, resulting in durability problems of marine engineering structures. The traditional methods are low in accuracy, high in cost or complex in process.

Method used

The method of evaluating the chloride ion diffusion coefficient of concrete under LDHs-containing corrosion inhibitor based on lightgbm was adopted. By training the data set and adjusting the hyperparameters, an efficient chloride ion diffusion coefficient detection model was constructed, and the lightgbm algorithm was used for prediction.

Benefits of technology

It improves the prediction accuracy and reliability of the chloride ion diffusion coefficient, helps the analysis and design of offshore concrete structures, and avoids the problem of low accuracy of traditional methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119007895A_ABST
    Figure CN119007895A_ABST
Patent Text Reader

Abstract

The invention relates to a light gbm-based concrete chloride ion diffusion coefficient evaluation method under a corrosion inhibitor containing LDHs (layered double hydroxides). The method comprises the following steps: acquiring test piece parameters of to-be-tested concrete; the test piece parameters are input into a preset chloride ion diffusion coefficient detection model, the chloride ion diffusion coefficient of the to-be-detected concrete is obtained, the chloride ion diffusion coefficient detection model is obtained based on data set training, the data set comprises erosion test data of a plurality of concrete in the chlorine salt environment, and the test piece parameters are determined according to the erosion test data. The chloride ion diffusion coefficient detection model is constructed by adopting a lightgbm algorithm model. The method disclosed by the invention has very high accuracy and reliability, and is beneficial to realizing analysis and design of a marine concrete structure or a reinforced concrete structure.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of safety assessment of marine concrete structures, and in particular to a method for evaluating the chloride ion diffusion coefficient of concrete containing an LDHs corrosion inhibitor based on LightGBM. Background Art

[0002] Concrete is recognized as a building material with a wide range of applications, easy pouring and molding, low price and good durability. Compared with other building materials, its lower cost and multifunctional use have become the main advantages of concrete materials. Marine engineering structures are usually built in coastal areas. Common chloride ions and sulfate ions in seawater will corrode concrete structures. Among them, chloride ions have the greatest impact on concrete structures. In this environment where ion corrosion is very likely to occur, the durability of concrete is very prominent. The destruction of concrete structures due to durability in the later stage usually requires a lot of manpower and material resources, causing damage to social and economic benefits. Therefore, it is very meaningful to improve the durability of concrete by modifying concrete materials.

[0003] Modern concrete service life design methods mainly provide sufficient concrete thickness to cover the steel bars to reduce the transmission rate of chloride ions in the concrete and the chloride ion concentration in the pore solution. Traditional standards oversimplify the complex mechanism of chloride ion corrosion of concrete and fail to provide a large number of examples of marine engineering concrete's actual resistance to chloride ion corrosion. According to a large number of field observations, there will be a large amount of concrete damage caused by seawater erosion in the next few decades, as well as more marine engineering maintenance work. There are currently a limited number of methods to improve the ability of concrete to resist ion corrosion, some of which are too expensive or complex, and some are not efficient enough. The new generation of steel reinforced concrete uses stainless steel bars, but the price is 5 or even 10 times that of ordinary carbon steel bars. Cathodic protection is more effective, but its application range is narrow and it is not widely used. Concrete protective coatings cannot last for a long time (10-20 years), so they need to be replaced regularly, resulting in their own maintenance cycle and increased manpower and material resources. Anti-corrosion and rust inhibitors have also been proposed, but their long-term effects are not reliable.

[0004] LDHs is the abbreviation of layered double hydroxides. It is an inorganic non-metallic material that has been considered to have great potential for adsorbing waste ions in recent years. LDHs can be expressed by the general chemical formula [M 2+ 1-x M 3+ x(OH)2]A n- x / n ·yH2O, where M 2+ and M 3+ represents any divalent or trivalent metal cation, An- is the anion between the layers, x is M 3+ / ( M 2+ +M 3+ ) molar ratio. The layered structure of LDHs is composed of positively charged (M 2+ , M 3+ )(OH)6 is similar to the octahedron of magnesia, which is arranged periodically to construct the main layer, and is composed of negatively charged interlayer anions and water molecules. Under certain conditions, certain functional substances (atoms, molecules and ions) can overcome the interaction between the layers of the layered compound and reversibly insert into the interlayer gap (sometimes ion exchange), expand the layer space, and form a strong interaction force with the layer to become a layer column material, giving it rich optical, electrical, magnetic, catalytic and other properties, making it a class of new functional materials with very attractive prospects. The interlayer anions form an interaction force with the main layer through hydrogen bonds, so the interlayer anions can be exchanged out by other anions that are easier to enter the interlayer, achieving the purpose of ion exchange.

[0005] Different methods have certain defects in the current technical measures to improve the chloride ion corrosion resistance of concrete. Therefore, it is necessary to find a more effective and direct method to directly remove harmful chloride ions. One of the methods is to find a high-efficiency chloride ion adsorbent. By grasping the research trends and hot spots in recent years, introducing layered double hydroxides (LDHs) into the field of building materials has become a new research direction.

[0006] Therefore, studying the chloride ion diffusion behavior of concrete containing LDHs corrosion inhibitors and accurately predicting the chloride ion diffusion coefficient are of great significance to the construction of marine engineering structures. However, there are many factors that affect the diffusion of chloride ions in concrete, including the material composition of concrete, pore structure, environmental conditions, etc., which makes the calculation of the chloride ion diffusion coefficient more complicated. The traditional method of calculating the chloride ion diffusion coefficient often directly applies Fick's second law to calculate, without comprehensively considering the many factors that affect the concrete's resistance to chloride ion corrosion. At the same time, Fick's second law has some limitations, such as idealized assumptions and time-varying diffusion coefficients.

[0007] Therefore, in order to overcome the low accuracy of traditional empirical formula prediction methods, a LightGBM-based method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitors is urgently needed. Summary of the invention

[0008] The purpose of the present invention is to propose a method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on LightGBM to solve the problems existing in the above-mentioned prior art. The method has very high accuracy and reliability, and is helpful to realize the analysis and design of marine concrete structures or reinforced marine concrete structures.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] A method for evaluating chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on LightGBM, comprising:

[0011] Obtaining the specimen parameters of the concrete to be tested;

[0012] The specimen parameters are input into a preset chloride ion diffusion coefficient detection model to obtain the chloride ion diffusion coefficient of the concrete to be tested, wherein the chloride ion diffusion coefficient detection model is obtained based on data set training, the data set includes several concrete corrosion test data in a chloride salt environment, and the chloride ion diffusion coefficient detection model is constructed using a lightgbm algorithm model.

[0013] Optionally, the corrosion test data include: material parameter information of the specimen, corrosion environment parameters, and chloride ion diffusion coefficient, wherein the material parameter information includes the type of cement, water-cement ratio, sand-binder ratio, sand ratio, LDHs dosage, and LDHs calcination conditions in the mix ratio, and the corrosion environment parameters include sodium chloride concentration and corrosion time.

[0014] Optionally, training the chloride ion diffusion coefficient detection model based on the data set includes:

[0015] The material parameter information and corrosion environment parameters of the specimen are used as input variables, and the chloride ion diffusion coefficient is used as an output variable to train the LightGBM algorithm model.

[0016] Optionally, training the lightgbm algorithm model includes:

[0017] Set the hyperparameters to be adjusted;

[0018] Dividing the data set into a training set and a test set;

[0019] The lightgbm algorithm model is trained based on the training set. During the training process, the hyperparameters to be adjusted are adjusted by grid search and five-fold cross validation method until the preset accuracy is reached, and the trained model is tested by the test set.

[0020] Optionally, the hyperparameters to be adjusted include: the number of weak learners, the learning rate, the maximum tree depth, and the maximum number of leaf nodes.

[0021] Optionally, setting the hyperparameter to be adjusted includes:

[0022] Preset the values ​​of the number of weak learners, learning rate, maximum tree depth, and maximum number of leaf nodes;

[0023] Based on the preset values, the values ​​are arranged and combined to obtain several hyperparameter combination schemes.

[0024] Optionally, adjusting the hyperparameter to be adjusted by a grid search method includes:

[0025] During the training process of the lightgbm algorithm model, the hyperparameters of the model are set to the hyperparameter combination scheme, and the prediction accuracy under the hyperparameter combination scheme is obtained;

[0026] All hyperparameter combination schemes are traversed, and the hyperparameter combination scheme with the highest prediction accuracy is selected as the final hyperparameter of the model.

[0027] Optionally, the data set is divided into: 80% as a training set and 20% as a test set.

[0028] The beneficial effects of the present invention are:

[0029] The present invention adopts the LightGBM algorithm to predict the chloride ion diffusion coefficient of concrete. The model is a tree-based integrated machine learning algorithm with the characteristics of high efficiency, flexibility and portability. The LightGBM algorithm iteratively trains the decision tree model through the gradient boosting framework to minimize the gradient of the loss function to improve the overall model performance, avoiding the defect of low prediction accuracy of traditional machine learning, and compared with the traditional machine learning database. More experimental parameters and larger experimental data sets are selected, with very high accuracy and reliability, which is helpful for the analysis and design of structures with concrete or reinforced concrete structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention 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 invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0031] Figure 1 The present invention is a schematic flow chart of a method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on LightGBM. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] The present embodiment proposes a method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on LightGBM, including:

[0035] Obtaining the specimen parameters of the concrete to be tested;

[0036] The specimen parameters are input into the preset chloride ion diffusion coefficient detection model to obtain the chloride ion diffusion coefficient of the concrete to be tested, wherein the chloride ion diffusion coefficient detection model is obtained based on data set training, and the data set includes several concrete corrosion test data in chloride salt environment, and the chloride ion diffusion coefficient detection model is constructed using the LightGBM algorithm model.

[0037] Furthermore, the corrosion test data include: material parameter information of the specimen, corrosion environment parameters, and chloride ion diffusion coefficient, wherein the material parameter information includes the type of cement, water-cement ratio, sand-binder ratio, sand ratio, LDHs dosage, and LDHs calcination conditions in the mix ratio, and the corrosion environment parameters include sodium chloride concentration and corrosion time.

[0038] Further, training a chloride ion diffusion coefficient detection model based on the data set includes:

[0039] The material parameter information and corrosion environment parameters of the specimen are taken as input variables, and the chloride ion diffusion coefficient is taken as the output variable to train the LightGBM algorithm model.

[0040] Furthermore, the training of the lightgbm algorithm model includes:

[0041] Set the hyperparameters to be adjusted;

[0042] Divide the dataset into training and testing sets;

[0043] The LightGBM algorithm model is trained based on the training set. During the training process, the hyperparameters to be adjusted are adjusted through grid search and five-fold cross validation method until the preset accuracy is reached, and the trained model is tested through the test set.

[0044] Furthermore, the hyperparameters to be adjusted include: the number of weak learners, the learning rate, the maximum tree depth, and the maximum number of leaf nodes.

[0045] Furthermore, setting the hyperparameters to be adjusted includes:

[0046] Preset the number of weak learners, learning rate, maximum tree depth, and maximum number of leaf nodes;

[0047] Based on the preset values, we can perform permutations and combinations to obtain several hyperparameter combination schemes.

[0048] Furthermore, the hyperparameters to be adjusted by the grid search method include:

[0049] During the training process of the LightGBM algorithm model, the hyperparameters of the model are set to the hyperparameter combination scheme, and the prediction accuracy under the hyperparameter combination scheme is obtained;

[0050] Traverse all hyperparameter combinations and select the hyperparameter combination with the highest prediction accuracy as the final hyperparameter of the model.

[0051] Furthermore, the data set is divided into 80% as a training set and 20% as a test set.

[0052] Combine the following Figure 1 A method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on LightGBM proposed in this embodiment is described in detail as follows:

[0053] (1) Collect 150 sets of concrete corrosion data in chloride salt environment, including material parameter information of specimens, curing environment parameters and chloride ion diffusion coefficient of specimens;

[0054] (2) The specimen parameters of 150 groups of concrete chloride environment corrosion tests were used as input variables X, and the chloride ion diffusion coefficient was used as the output variable D to train the LightGBM algorithm;

[0055] (3) Adjust the hyperparameters of the LightGBM algorithm through the grid search method and train and test the model in the training data. The hyperparameters with the best prediction accuracy are used as the final parameters of the model to obtain the trained LightGBM model F(X);

[0056] (4) By inputting the specimen parameters into the lightgbm model F(X), the chloride ion diffusion coefficient of the concrete specimen can be quickly obtained.

[0057] Furthermore, the collected experimental data contain basic parameter information of the specimens, including 9 test data such as cement type, water-cement ratio, sand-binder ratio, sand ratio, LDHs dosage, LDHs calcination condition, sodium chloride concentration, erosion time, chloride ion diffusion coefficient, etc.

[0058] Furthermore, the input variable is: X=[A, W / B, S / B, SA, AM, CS, EM1, EM2], and the output variable D is the value of the chloride ion diffusion coefficient.

[0059] Furthermore, the hyperparameters to be adjusted of the lightgbm algorithm are: the number of weak learners n_esi, the learning rate Learn_Rate, the maximum tree depth max_depth, and the maximum number of leaf nodes max_leaf_nodes.

[0060] Furthermore, the definition method of hyperparameters is: pre-set the possible values ​​of the number of weak learners n_esi, learning rate Learn_Rate, maximum tree depth max_depth, and maximum number of leaf nodes max_leaf_nodes, and then based on the possible value combinations of these parameters, a total of m parameter combination schemes are obtained: Z=[n_esi, Learn_Rate, max_de, max_leaf_nodes].

[0061] Furthermore, the optimal hyperparameter value selection method of the LightGBM algorithm is: traverse m parameter combination schemes, and each time a parameter combination scheme is obtained: Z = [n_esi, Learn_Rate, max_depth, max_leaf_nodes] j∈m;

[0062] Set the parameters in the LightGBM algorithm to the above combination, and use the training set for training to obtain the prediction accuracy under this parameter combination;

[0063] Repeat the above process until m rounds of training;

[0064] The parameter combination with the highest prediction accuracy in round m is taken as the final LightGBM model parameters, and the model training is completed.

[0065] Furthermore, the data set is divided into 80% as a training set and 20% as a test set.

[0066] This embodiment adopts the LightGBM algorithm to predict the chloride ion diffusion coefficient of marine concrete, avoiding the defect of low prediction accuracy of traditional machine learning. Compared with traditional machine learning databases, it selects more experimental parameters and larger experimental data sets, with very high accuracy and reliability, which is helpful for the analysis and design of structures with marine concrete or reinforced concrete structures.

[0067] More specifically, the Lightgbm algorithm model construction process in this embodiment is described in detail below:

[0068] Import the library:

[0069] pandas and numpy for data manipulation

[0070] import pandas as pd;

[0071] import numpy as np;

[0072] #from sklearn.preprocessing import;

[0073] from lightgbm import LGBMRegressor;

[0074] import matplotlib.pyplot as plt

[0075] %matplotlib inline;

[0076] import seaborn as sns;

[0077] from sklearn.model_selection import train_test_split;

[0078] from IPython.core.pylabtools import figsize;

[0079] pd.options.mode.chained_assignment = None;

[0080] pd.set_option('display.max_columns', 60);

[0081] import warnings

[0082] warnings.filterwarnings("ignore");

[0083] data = pd.read_excel('E:\ lightgbm\ new folder.xlsx');

[0084] features = pd.DataFrame(data[['A', 'W / B', 'S / B', 'SA', 'AM', 'CS', 'EM1', 'EM2']]); #Here is the basic information of the input parameters, 'A' is the cement type, 'W / B' is the water-cement ratio, 'S / B' is the sand-binder ratio, 'SA' is the sand ratio, 'AM' is the LDHs dosage, 'CS' is the LDHs calcination condition, 'EM1' is the sodium chloride concentration, and 'EM2' is the erosion time;

[0085] targets = pd.DataFrame(data['D']);#Here is the output parameter 'D' of the model, 'D' represents the value of the chloride ion diffusion coefficient.

[0086] X, X_test, y, y_test = train_test_split(features, targets, test_size= 0.2,random_state =0)

[0087] print(X.shape)

[0088] print(X_test.shape)

[0089] print(y.shape)

[0090] print(y_test.shape);

[0091] X.to_csv('data / training_features.csv', index = False)

[0092] X_test.to_csv('data / testing_features.csv', index = False)

[0093] y.to_csv('data / training_labels.csv', index = False)

[0094] y_test.to_csv('data / testing_labels.csv', index = False);

[0095] X= pd.read_csv('data / training_features.csv')

[0096] y=pd.read_csv('data / training_labels.csv')

[0097] X_test=pd.read_csv('data / testing_features.csv')

[0098] y_test=pd.read_csv('data / testing_labels.csv');

[0099] import math

[0100] from sklearn.metrics import r2_score

[0101] from sklearn.metrics import mean_squared_error

[0102] from sklearn.metrics import mean_absolute_error

[0103] from sklearn.metrics import mean_absolute_percentage_error

[0104] def zb(y_true, y_pred):

[0105] r2=r2_score(y_true,y_pred)

[0106] mse=mean_squared_error(y_true,y_pred)

[0107] rmse=math.sqrt(mse)

[0108] mae=mean_absolute_error(y_true,y_pred)

[0109] mape=mean_absolute_percentage_error(y_true,y_pred)

[0110] return r2,mse,rmse,mae,mape;

[0111] from sklearn.model_selection import RandomizedSearchCV, GridSearchCV

[0112] trees_grid = { #When searching for tree-based models, use trees_grid; when searching for other models, use param_gird #'max_depth':[3,4,5,6,7,8],'subsample':[0.1,0.2,0.3,0.4,0.5,0.6],'learning_rate':[0.01,0.05,0.1,0.15,0.2],

[0113] #'n_estimators':[100,200,300,400,500],'min_child_weight':[1,3,5]#xgboost #'n_estimators':[20,40,60,80,100,120],'max_features':['auto','sqrt'],'max_depth':[None,1,2,3,4,5,6],

[0114] #'min_samples_leaf':[1,2,4,8],'min_samples_split':[2,5,10] #(RF)#'max_depth':[3,4,5,6,7,8],'subsample':[0.1,0.2,0.3,0.4,0.5,0.6,0.7],'learning_rate':[0.05,0.1,0.15,0.2,0.25,0.3],

[0115] #'n_estimators':[100,200,300,400,500],'num_leaves':[1,2,3,4,5,6,7] #lightgbm model

[0116] 'max_depth':[3,4,5,6],'subsample':[0.1,0.2,0.3,0.4],'learning_rate':[0.05,0.1,0.15,0.2],

[0117] 'n_estimators':[100,200,300],'num_leaves':[1,2,3,4,5] #lightgbm model

[0118] #'kernel': ['rbf'],'C': [ 0.01, 0.1, 1, 10, 100,1000],'gamma': [0.001, 0.01, 0.1, 1, 10, 100] #'max_depth':[None,1,2,3,4,5,6,7,8,9,10],'max_features':['None','auto','sqrt','log2'],'min_samples_leaf':[1,2,4,8,10],

[0119] #'min_samples_split':[2,4,6,8,10] #'max_depth':[4],'subsample':[0.5],'learning_rate':[0.1],'n_estimators':

[400] ,'min_child_weight':

[0120] }

[0121] #model = XGBRegressor()

[0122] model = LGBMRegressor(random_state=42) #lightgbm model

[0123] #model = RandomForestRegressor() #Random Forest Model

[0124] #model= SVR() #Support vector machine model

[0125] #model = DecisionTreeRegressor() #Decision tree model;

[0126] grid_search = GridSearchCV(estimator=model,

[0127] param_grid=trees_grid,

[0128] cv=5,

[0129] scoring='neg_root_mean_squared_error',

[0130] verbose=1,

[0131] n_jobs=-1,

[0132] return_train_score=True);

[0133] grid_search.fit(X, y);

[0134] results = pd.DataFrame(grid_search.cv_results_)

[0135] results.sort_values('mean_test_score', ascending = False).head(5);

[0136] final_model = grid_search.best_estimator_;

[0137] Evaluation metrics for the training set, test set, and entire dataset

[0138] final_model.fit(X,y)

[0139] model_pred = final_model.predict(X_test)

[0140] model_zb_test = zb(y_test,model_pred)

[0141] model_pred_train=final_model.predict(X)

[0142] model_zb_train = zb(y,model_pred_train)

[0143] model_pred_all=final_model.predict(features)

[0144] model_zb_all=zb(targets,model_pred_all);

[0145] model_zb_train;

[0146] model_zb_test;

[0147] model_zb_all;

[0148] df = pd.DataFrame(model_pred, columns=['pred']);

[0149] df.to_excel('pred_test_dt.xlsx', index=False)

[0150] df1 = pd.DataFrame(model_pred_train, columns=['pred_train'])

[0151] df1.to_excel('pred_train_dt.xlsx', index=False).

[0152] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on lightgbm, characterized in that: include: Obtaining the specimen parameters of the concrete to be tested; The specimen parameters are input into a preset chloride ion diffusion coefficient detection model to obtain the chloride ion diffusion coefficient of the concrete to be tested, wherein the chloride ion diffusion coefficient detection model is obtained based on data set training, the data set includes several concrete corrosion test data in a chloride salt environment, and the chloride ion diffusion coefficient detection model is constructed using a lightgbm algorithm model.

2. The method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on lightgbm according to claim 1 is characterized in that: The corrosion test data include: material parameter information of the specimen, corrosion environment parameters, and chloride ion diffusion coefficient, wherein the material parameter information includes the type of cement, water-cement ratio, sand-binder ratio, sand ratio, LDHs dosage, and LDHs calcination conditions in the mix ratio, and the corrosion environment parameters include sodium chloride concentration and corrosion time.

3. The method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on lightgbm according to claim 2 is characterized in that: Training the chloride ion diffusion coefficient detection model based on the data set includes: The material parameter information and corrosion environment parameters of the specimen are used as input variables, and the chloride ion diffusion coefficient is used as an output variable to train the LightGBM algorithm model.

4. The method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on lightgbm according to claim 3 is characterized in that: Training the lightgbm algorithm model includes: Set the hyperparameters to be adjusted; Dividing the data set into a training set and a test set; The training set is input into the lightgbm algorithm model. During the training process, the hyperparameters to be adjusted are adjusted by grid search and five-fold cross validation method until the preset accuracy is reached, and the trained model is tested by the test set.

5. The method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on lightgbm according to claim 4 is characterized in that: The hyperparameters to be adjusted include: the number of weak learners, the learning rate, the maximum tree depth, and the maximum number of leaf nodes.

6. The method for evaluating the chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on lightgbm according to claim 5 is characterized in that: Setting the hyperparameter to be adjusted includes: Preset the values ​​of the number of weak learners, learning rate, maximum tree depth, and maximum number of leaf nodes; Based on the preset values, the values ​​are arranged and combined to obtain several hyperparameter combination schemes.

7. The method for evaluating chloride ion diffusion coefficient of concrete containing LDHs corrosion inhibitor based on lightgbm according to claim 6 is characterized in that: Adjusting the hyperparameter to be adjusted by the grid search method includes: During the training process of the lightgbm algorithm model, the hyperparameters of the model are set to the hyperparameter combination scheme, and the prediction accuracy under the hyperparameter combination scheme is obtained; All hyperparameter combination schemes are traversed, and the hyperparameter combination scheme with the highest prediction accuracy is selected as the final hyperparameter of the model.

Citation Information

Patent Citations

  • Heating ventilation air conditioning system fault diagnosis method based on LightGBM algorithm and grid search algorithm

    CN113378908A

  • Machine learning-based cement-based material performance optimization method and system

    CN118277887A

  • Recycled concrete carbonization depth prediction method based on XGBoost

    CN118551298A

Cited By

  • Chloride ion diffusion coefficient prediction method based on CatBoost model

    CN120412838A