Performance prediction and mix proportion multi-objective optimization design method and system for common concrete and ultra-high performance concrete

Abstract

The invention discloses a common concrete and ultra-high performance concrete performance prediction and mix proportion multi-objective optimization design method and system, and the method comprises the steps: obtaining a comprehensive data set which covers the mix proportion, performance data and cost data of common and ultra-high performance concrete; correlation analysis is carried out, and influence factors strongly correlated with the target variable are determined; constructing a concrete performance prediction model by using a machine learning algorithm, training and testing, then carrying out interpretability analysis, and selecting an optimal concrete performance prediction model; constructing a mix proportion optimization design objective function and an ultra-high performance concrete cost function; establishing a multi-objective optimization design mathematical model in combination with constraint conditions, a cost function and an optimization objective function; and solving by adopting a multi-objective evolutionary algorithm, generating an approximate Pareto frontier solution set, and determining an optimal mix proportion optimization design scheme. According to the method, the concrete performance can be predicted, the mix proportion with the optimal comprehensive performance and cost effectiveness is obtained, and multi-objective optimization design of the concrete mix proportion is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of architectural design, and particularly relates to a method and system for performance prediction and multi-objective optimization design of mix proportion of ordinary concrete and ultra-high performance concrete. Background Art

[0002] Concrete is composed of cement, aggregates, water, admixtures and mineral admixtures, and is an important structural material in construction engineering. Its mix proportion design is complex. The key lies in selecting appropriate proportions to obtain the required performance, and it is necessary to consider workability, mechanical properties, durability and cost. It is difficult to optimize all performance and cost targets simultaneously. Therefore, in the process of mix proportion design, multi-objective optimization design is required.

[0003] There are many deficiencies in traditional concrete mix proportion design methods, such as low efficiency, strong subjectivity and high dependence on human experience. Common methods such as the unit weight method and the volume method have a large workload and lack the trade-off between multiple objectives. Although the full factor design method can transform the problem into multi-objective optimization, constructing the relationship between influencing factors and concrete performance is complex, and the coupling relationship between factors poses a challenge to the optimization algorithm. In recent years, machine learning technology has provided a new approach for concrete mix proportion design. By establishing a concrete performance prediction model, an optimization algorithm is used for design as the objective function. However, traditional machine learning modeling requires profound domain experience. Optimizing the performance of the model requires a large amount of hyperparameter tuning and model selection, and manual operation has the risk of overfitting. At the same time, multi-objective optimization of concrete workability, mechanical properties, durability and cost involves objective conflicts. Common methods transform multiple objectives into a single objective, and the weight assignment depends on experience, subjective judgment or historical data, and only local optimal solutions can be found, ignoring other possible excellent solutions and not achieving the global optimum.

[0004] Therefore, completing the machine learning process to optimize the concrete mix proportion design requires a large amount of time and effort. It is necessary to overcome the complexity in the modeling process and solve the objective conflicts and weight assignment problems in multi-objective optimization. Only by comprehensively searching in the solution space and finding a more comprehensive and excellent concrete mix proportion design scheme can the problems existing in the existing concrete mix proportion design technology, such as low efficiency, strong subjectivity, difficulty in achieving global optimum, and complexity of modeling and optimization, be overcome. Summary of the Invention

[0005] The object of the present invention is to provide a method and system for performance prediction and multi-objective optimization design of mix proportion of ordinary concrete and ultra-high performance concrete to solve the problems existing in the above-mentioned prior art.

[0006] On the one hand, to achieve the above object, the present invention provides a method for performance prediction and multi-objective optimization design of mix proportion of ordinary concrete and ultra-high performance concrete, including: obtaining a comprehensive data set, where the comprehensive data set includes mix proportion data, performance data and cost data of concrete, and the concrete includes ordinary concrete and ultra-high performance concrete; performing correlation analysis on the data in the comprehensive data set based on preset target variables to obtain influencing factors with strong correlation with the preset target variables; where the preset target variables include the compressive strength performance of ordinary concrete, the slump performance of ordinary concrete, the porosity performance of ordinary concrete, the compressive strength performance of ultra-high performance concrete and the flexural strength performance of ultra-high performance concrete; constructing concrete performance prediction models corresponding to each of the preset target variables based on a machine learning algorithm, training and testing the concrete performance prediction models corresponding to each of the preset target variables based on the influencing factors to obtain the trained and tested concrete performance prediction models;

[0007] Performing interpretability analysis on each of the trained and tested concrete performance prediction models, determining the optimal concrete performance prediction model corresponding to each preset target variable based on the results of the interpretability analysis; setting constraint conditions, setting corresponding mix proportion optimization design objective functions based on the optimal concrete performance prediction models of each preset target variable, and constructing a raw material cost objective function for ultra-high performance concrete; constructing a multi-objective mix proportion optimization design mathematical model based on the constraint conditions, the raw material cost objective function and each mix proportion optimization design objective function; solving the multi-objective mix proportion optimization design mathematical model based on a multi-objective evolutionary algorithm to generate an approximate Pareto front solution set; determining the mix proportion optimization design scheme with the best comprehensive performance based on the approximate Pareto front solution set.

[0008] Optionally, before performing correlation analysis on the data in the comprehensive data set based on the preset target variables, it further includes cleaning the comprehensive data set to obtain a preprocessed comprehensive data set, and performing correlation analysis based on the preprocessed comprehensive data set.

[0009] Optionally, the performing correlation analysis on the data in the comprehensive data set based on the preset target variables specifically includes: analyzing the correlation between each data in the comprehensive data set and the preset target variables based on the Spearman correlation coefficient, and determining the influencing factors with strong correlation with the preset target variables based on the results of the correlation analysis.

[0010] Optionally, the machine learning algorithm includes an automated machine learning algorithm and a traditional machine learning algorithm.

[0011] Optionally, the training and testing of the concrete performance prediction model corresponding to each of the preset target variables specifically include: The comprehensive data set is divided into a training set and a testing set, where the training set accounts for 80% of the comprehensive data set, and the testing set accounts for 20% of the comprehensive data set. The training set is used to train the concrete performance prediction model, and the testing set is used to verify the concrete performance prediction model. The prediction results of the concrete performance prediction model are evaluated according to the target loss function to obtain the trained and tested concrete performance prediction model.

[0012] Optionally, the interpretability analysis of each trained and tested concrete performance prediction model specifically includes: calculating the SHAP values of each influencing factor in each algorithm model based on the SHAP method to obtain the interpretability analysis result, and determining the optimal concrete performance prediction model corresponding to each preset target variable from the interpretability analysis result.

[0013] Optionally, the constraint conditions include the density of ordinary concrete raw materials, the dosage limit of ordinary concrete raw materials, the dosage ratio limit of ordinary concrete raw materials, the density and price of ultra-high performance concrete raw materials, the dosage limit of ultra-high performance concrete raw materials, the dosage ratio limit of ultra-high performance concrete raw materials, the slump flow of ultra-high performance concrete, the curing method of ultra-high performance concrete, the curing temperature of ultra-high performance concrete, and the curing time of ultra-high performance concrete.

[0014] Optionally, determining the mix proportion optimization design scheme with the best comprehensive performance and cost-benefit based on the Pareto front solution set specifically includes: solving the multi-objective mix proportion optimization design mathematical model based on the multi-objective evolutionary algorithm to obtain several approximate Pareto front solution sets; based on the entropy weight-TOPSIS method, determining the weights by the entropy value method and normalizing each index, and then using the TOPSIS method to comprehensively evaluate each approximate Pareto front solution set, and determining the mix proportion optimization design scheme with the best comprehensive performance and cost-benefit based on the evaluation results.

[0015] On the other hand, to achieve the above object, the present invention provides a performance prediction and mix proportion multi-objective optimization design system for ordinary concrete and ultra-high performance concrete, including:

[0016] A data acquisition module, configured to obtain a comprehensive data set, where the comprehensive data set includes the mix proportion data, performance data, and cost data of concrete, and the concrete includes ordinary concrete and ultra-high performance concrete;

[0017] A correlation analysis module, which is used to perform a correlation analysis on the data in the comprehensive dataset according to a preset target variable to obtain influencing factors with strong correlation with the preset target variable; wherein, the preset target variables include the compressive strength performance of ordinary concrete, the slump performance of ordinary concrete, the porosity performance of ordinary concrete, the compressive strength performance of ultra-high performance concrete, and the flexural strength performance of ultra-high performance concrete;

[0018] A concrete performance prediction model construction module, which is used to construct a concrete performance prediction model corresponding to each of the preset target variables according to a machine learning algorithm, train and test the concrete performance prediction model corresponding to each of the preset target variables based on the influencing factors, and obtain a trained and tested concrete performance prediction model; perform an interpretability analysis on each of the trained and tested concrete performance prediction models, and determine an optimal concrete performance prediction model corresponding to each preset target variable based on the results of the interpretability analysis;

[0019] A multi-objective mix proportion optimization design model construction module, which is used to set constraint conditions, set corresponding mix proportion optimization design objective functions based on the optimal concrete performance prediction models of each preset target variable, and construct a raw material cost objective function for ultra-high performance concrete; construct a multi-objective mix proportion optimization design mathematical model based on the constraint conditions, the raw material cost objective function, and each mix proportion optimization design objective function;

[0020] A multi-objective evolutionary algorithm solving module, which is used to solve the multi-objective mix proportion optimization design mathematical model according to a multi-objective evolutionary algorithm to generate an approximate Pareto front solution set; determine an optimal mix proportion optimization design scheme with the best comprehensive performance based on the approximate Pareto front solution set.

[0021] The technical effect of the present invention is:

[0022] The present invention combines machine learning, multi-objective evolutionary algorithms, and performance evaluation methods, can find the global optimal solution to realize the optimization design of the concrete mix proportion. Through this method, the concrete performance can be effectively predicted, and the mix proportion with the best comprehensive performance and cost-benefit can be found. By comprehensively considering multiple performance indicators and cost factors, the present invention provides a scientific, efficient, and economical concrete mix proportion optimization design method, which helps to improve the technical level and engineering quality of the concrete industry. Description of the Drawings

[0023] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0024] Figure 1Flow chart of the concrete performance prediction and multi-objective mix proportion rapid design method and system based on automated machine learning in the embodiments of the present invention;

[0025] Figure 2 Taylor diagram of the prediction effect of normal concrete performance based on automated machine learning algorithm and traditional machine learning algorithm in the embodiments of the present invention; wherein, Figure 2 (a) in it is the Taylor diagram of the prediction effect of normal concrete compressive strength based on automated machine learning algorithm and traditional machine learning algorithm in the embodiments of the present invention; Figure 2 (b) in it is the Taylor diagram of the prediction effect of normal concrete slump based on automated machine learning algorithm and traditional machine learning algorithm in the embodiments of the present invention; Figure 2 (c) in it is the Taylor diagram of the prediction effect of normal concrete porosity based on automated machine learning algorithm and traditional machine learning algorithm in the embodiments of the present invention;

[0026] Figure 3 Taylor diagram of the prediction effect of ultra-high performance concrete performance based on automated machine learning algorithm and traditional machine learning algorithm in the embodiments of the present invention; Figure 3 (a) in it is the Taylor diagram of the prediction effect of ultra-high performance concrete compressive strength based on automated machine learning algorithm and traditional machine learning algorithm in the embodiments of the present invention; Figure 3 (b) in it is the Taylor diagram of the prediction effect of ultra-high performance concrete flexural strength based on automated machine learning algorithm and traditional machine learning algorithm in the embodiments of the present invention;

[0027] Figure 4 Schematic diagram of the evaluation index of the convergence performance of the multi-objective evolutionary algorithm approximate Pareto front in the embodiments of the present invention; wherein, Figure 4 (a) in it is the HV 2D schematic diagram in the embodiments of the present invention; Figure 4 (b) in it is the schematic diagram of the disadvantages of the HV evaluation index in the embodiments of the present invention;

[0028] Figure 5 Login module of the performance prediction and mix proportion multi-objective optimization design system of normal concrete and ultra-high performance concrete in the embodiments of the present invention; wherein, Figure 5 (a) in it is the user login module in the embodiments of the present invention; Figure 5 (b) in it is the user registration module in the embodiments of the present invention.

[0029] Figure 6 Concrete performance prediction module of the performance prediction and mix proportion multi-objective optimization design system of normal concrete and ultra-high performance concrete in the embodiments of the present invention; wherein, Figure 6 (a) in it is the normal concrete compressive strength performance prediction module in the embodiments of the present invention; Figure 6In (b) is the slump performance prediction module for normal concrete in the embodiment of the present invention; Figure 6 In (c) is the porosity performance prediction module for normal concrete in the embodiment of the present invention.

[0030] Figure 7 Is the concrete performance prediction module of a performance prediction and mix proportion multi-objective optimization design system for normal concrete and ultra-high performance concrete in the embodiment of the present invention; wherein, Figure 7 In (a) is the compressive strength performance prediction module for ultra-high performance concrete in the embodiment of the present invention; Figure 7 In (b) is the flexural strength performance prediction module for ultra-high performance concrete in the embodiment of the present invention.

[0031] Figure 8 Is the mix proportion multi-objective optimization design module of a performance prediction and mix proportion multi-objective optimization design system for normal concrete and ultra-high performance concrete in the embodiment of the present invention; wherein, Figure 8 In (a) is the mix proportion multi-objective optimization design module for normal concrete in the embodiment of the present invention; Figure 8 In (b) is the mix proportion multi-objective optimization design module for ultra-high performance concrete in the embodiment of the present invention.

[0032] Figure 9 Is the data upload module of a performance prediction and mix proportion multi-objective optimization design system for normal concrete and ultra-high performance concrete in the embodiment of the present invention. Specific embodiments

[0033] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0034] Such as Figure 1 - Figure 4As shown in the figure, in this embodiment, a performance prediction and mix ratio multi-objective optimization design method for normal concrete and ultra-high performance concrete is provided, including: obtaining a comprehensive data set, where the comprehensive data set includes mix ratio data, performance data, and cost data of concrete, and the concrete includes normal concrete and ultra-high performance concrete; performing a correlation analysis on the data in the comprehensive data set based on a preset target variable to obtain influencing factors with strong correlation with the preset target variable; where the preset target variables include normal concrete compressive strength performance, normal concrete slump performance, normal concrete porosity performance, ultra-high performance concrete compressive strength performance, and ultra-high performance concrete flexural strength performance; constructing a prediction model corresponding to each of the preset target variables based on a machine learning algorithm, training and testing based on the influencing factors and the concrete performance prediction models corresponding to each of the preset target variables to obtain the trained and tested concrete performance prediction models; performing an interpretability analysis on each of the trained and tested concrete performance prediction models, and determining the optimal concrete performance prediction model corresponding to each preset target variable based on the interpretability analysis results; setting constraint conditions, setting a corresponding mix ratio optimization design objective function based on the optimal concrete performance prediction model of each preset target variable, and constructing a raw material cost objective function for ultra-high performance concrete; constructing a multi-objective mix ratio optimization design mathematical model based on the constraint conditions, the raw material cost objective function, and each mix ratio optimization design objective function; solving the multi-objective mix ratio optimization design mathematical model based on a multi-objective evolutionary algorithm to generate an approximate Pareto front solution set; and determining an optimal mix ratio optimization design plan with the best comprehensive performance based on the approximate Pareto front solution set.

[0035] On the other hand, as Figures 5 - 9 shown in the figure, this embodiment provides a performance prediction and mix ratio multi-objective optimization design system for normal concrete and ultra-high performance concrete, including: a login module, a concrete performance prediction system based on an automated machine learning algorithm and a traditional machine learning algorithm, a concrete mix ratio optimization design system based on an automated machine learning algorithm and a multi-objective evolutionary algorithm, and a data upload system. The login module specifically includes a user login and a user registration module.

[0036] The concrete performance prediction system based on an automated machine learning algorithm and a traditional machine learning algorithm specifically includes:

[0037] A data acquisition module, configured to obtain a comprehensive data set, where the comprehensive data set includes mix ratio data, performance data, and cost data of concrete, and the concrete includes normal concrete and ultra-high performance concrete;

[0038] A correlation analysis module, configured to perform a correlation analysis on the data in the comprehensive dataset according to a preset target variable, so as to obtain influencing factors with strong correlation with the preset target variable; wherein, the preset target variables include the compressive strength performance of normal concrete, the slump performance of normal concrete, the porosity performance of normal concrete, the compressive strength performance of ultra-high performance concrete, and the flexural strength performance of ultra-high performance concrete;

[0039] A concrete performance prediction model construction module, configured to construct a concrete performance prediction model corresponding to each of the preset target variables according to a machine learning algorithm, train and test the concrete performance prediction model corresponding to each of the preset target variables based on the influencing factors, so as to obtain a trained and tested concrete performance prediction model; perform an interpretability analysis on each of the trained and tested concrete performance prediction models, and determine an optimal concrete performance prediction model corresponding to each preset target variable based on the results of the interpretability analysis;

[0040] A concrete mix proportion optimization design system and a data upload system based on an automated machine learning algorithm and a multi-objective evolutionary algorithm, comprising:

[0041] A multi-objective mix proportion optimization design model construction module, configured to set constraint conditions, set a corresponding mix proportion optimization design objective function based on the optimal concrete performance prediction model of each preset target variable, and construct a raw material cost objective function for ultra-high performance concrete; construct a multi-objective mix proportion optimization design mathematical model based on the constraint conditions, the raw material cost objective function, and each mix proportion optimization design objective function;

[0042] A multi-objective evolutionary algorithm solving module, configured to solve the multi-objective mix proportion optimization design mathematical model according to a multi-objective evolutionary algorithm, so as to generate an approximate Pareto front solution set; determine an optimal mix proportion optimization design scheme with comprehensive performance based on the approximate Pareto front solution set.

[0043] The data upload system specifically includes a data upload module.

[0044] The concrete performance prediction system based on the automated machine learning algorithm and the traditional machine learning algorithm is built based on the optimal concrete performance prediction model. The concrete performance prediction module includes a normal concrete performance prediction module and an ultra-high performance concrete performance prediction module. The input module is used to input concrete mix proportion parameters, and the submission module is used to submit the mix proportion parameters and give corresponding concrete performance prediction results.

[0045] The concrete mix proportion optimization design system based on the automated machine learning algorithm and the multi-objective evolutionary algorithm is built based on the multi-objective mix proportion optimization design mathematical model. The concrete mix proportion optimization design module includes a normal concrete mix proportion optimization design module and a ultra-high performance concrete mix proportion optimization design module. The input module is used to input the limit values of concrete mix proportion parameters, and the submission module is used to submit the limit values of mix proportion parameters and give the corresponding concrete mix proportion results.

[0046] The data upload system specifically includes a data upload module for uploading concrete mix proportion parameters, performance data, and cost data.

[0047] In the first aspect of this embodiment, that is, the concrete performance prediction method based on the automated machine learning algorithm and the traditional machine learning algorithm, specifically includes:

[0048] 1. Concrete performance prediction method based on the automated machine learning algorithm and the traditional machine learning algorithm

[0049] The concrete performance prediction method based on the automated machine learning algorithm and the traditional machine learning algorithm mainly includes the prediction methods for the compressive strength, slump, and porosity of normal concrete, as well as the prediction methods for the compressive strength and flexural strength of ultra-high performance concrete, which can be interpreted as:

[0050] (1) Collect the mix proportions of normal concrete and ultra-high performance concrete and their corresponding concrete performance data and cost data, establish a comprehensive data set, and perform automatic data cleaning and feature engineering and other work using the automated machine learning algorithm;

[0051] (2) Based on the comprehensive data set, conduct a preliminary analysis of the data using the Spearman correlation coefficient;

[0052] (3) Establish performance prediction models for normal concrete and ultra-high performance concrete based on the automated machine learning algorithm and the traditional machine learning algorithm;

[0053] In step (1), based on the literature, obtain a large amount of mix proportions of normal concrete and ultra-high performance concrete and their corresponding concrete performance data and cost data, and perform automatic data cleaning and feature engineering and other work through the automated machine learning algorithm to establish a comprehensive data set, specifically including:

[0054] Extract the cement dosage, fine aggregate dosage, coarse aggregate dosage, water dosage, high-range water reducer dosage, fly ash dosage, accelerating agent dosage, silica fume dosage, curing time, and compressive strength of the ordinary concrete mix proportion as influencing factors, input them into the dataset, and use them for predicting the compressive strength performance of ordinary concrete; extract the cement dosage, fine aggregate dosage, coarse aggregate dosage, water dosage, high-range water reducer dosage, fly ash dosage, accelerating agent dosage, silica fume dosage, and slump of the ordinary concrete mix proportion as influencing factors, input them into the dataset, and use them for predicting the slump performance of ordinary concrete; extract the binder dosage, fly ash dosage, blast furnace slag dosage, high-range water reducer dosage, water-binder ratio, fine-to-coarse aggregate ratio, curing method, curing time, and porosity of the ordinary concrete mix proportion, input them into the dataset, and use them for predicting the porosity performance of ordinary concrete;

[0055] Extract the cement dosage, silica fume dosage, fly ash dosage, slag dosage, nano-SiO 2 dosage, limestone powder dosage, quartz powder dosage, water dosage, aggregate dosage, water reducer dosage, steel fiber dosage, curing temperature, curing time, and compressive strength of the ultra-high performance concrete mix proportion as influencing factors, input them into the dataset, and use them for predicting the compressive strength performance of ultra-high performance concrete; extract the cement dosage, silica fume dosage, fly ash dosage, slag dosage, nano-SiO 2 dosage, limestone powder dosage, quartz powder dosage, water dosage, aggregate dosage, water reducer dosage, steel fiber dosage, curing temperature, curing time, specimen size, and flexural strength of the ultra-high performance concrete mix proportion as influencing factors, input them into the dataset, and use them for predicting the flexural strength performance of ultra-high performance concrete;

[0056] Clean the dataset by means of screening, deduplication, etc.; 80% of the dataset is divided into the training set, and 20% is divided into the test set, where the test set is used to train the concrete performance prediction model, and the test set is used to verify the concrete performance prediction model.

[0057] In step (2), the Spearman correlation coefficient is used for preliminary data analysis. The Spearman correlation coefficient is used to measure the correlation between influencing factors. Its value has nothing to do with the specific values of the influencing factors and only relates to the magnitude relationship between the influencing factors. Its value ranges from -1 to 1. When it is close to 1 or -1, it indicates a strong positive or negative correlation between the influencing factors. When it is close to 0, it indicates almost no relationship. The specific calculation formula is as follows:

[0058]

[0059] where, d i represents the difference in the rank values of the i-th data pair; N represents the total number of samples.

[0060] In step (3), the automated machine learning algorithms include: H2O, LightAutoML, Mljar, AutoGluon, EvalML, FlaML, and TPOT; the traditional machine learning algorithms include: SVR, RF, and MLP.

[0061] Use the coefficient of determination (R 2 ), mean squared error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and explained variance (EV) as the evaluation metrics for the prediction results of the concrete performance prediction model based on the automated machine learning algorithms and traditional machine learning algorithms. Among them, R 2 is used to reflect the matching degree between the predicted value and the actual value, and its value is less than 1. The closer R 2 is to 1, the better the model fits the training data; MSE is used to reflect the error of each data point in the regression model; MAE is used to reflect the average value of the distance between the predicted value and the actual value; MAPE is used to reflect the relative error between the predicted value and the actual value; EV is used to represent the proportion of the data variance explained by the model, and its value ranges from 0 to 1. The closer it is to 1, the better the model's ability to explain the actual data variance. The specific calculation formulas are as follows:

[0062]

[0063] Among them, y i is the actual value, is the predicted value, is the average value of the predicted values, and Var{y} is the variance of y.

[0064] In step (3), use the standard deviation (S), Spearman correlation coefficient, and root mean square error (RMSE) to evaluate the prediction results of the automated machine learning algorithms and traditional machine learning algorithms; S reflects the breadth of the data distribution and the degree of data deviation from the average; RMSE reflects the difference between the predicted value and the actual value and is the average value of the mean squared error. The closer S is to 1, and the smaller the Spearman correlation coefficient and RMSE, the smaller the difference between the predicted value and the actual value of the model. The specific calculation formulas are as follows:

[0065]

[0066] Among them, y i is the actual value, is the predicted value.

[0067] Adopt the SHAP theory to conduct an interpretability analysis of the concrete performance prediction model based on the automated machine learning algorithm. Use the training set as the target of theoretical analysis, calculate the contribution of each influencing factor to the model's predicted value (i.e., the SHAP value), and interpret the prediction of the model.

[0068] Based on the dataset, by evaluating the accuracy and error of automated machine learning algorithms and traditional machine learning algorithms, and conducting an interpretability analysis of automated machine learning algorithms and traditional machine learning algorithms based on the SHAP theory, it is determined that the AutoGluon algorithm has the best effect in predicting the porosity and compressive strength of normal concrete, as well as the compressive strength and flexural strength of ultra-high performance concrete; it is determined that the Mljar algorithm has the best effect in predicting the slump performance of normal concrete.

[0069] 2. Concrete Performance Prediction System Based on Automated Machine Learning Algorithm and Traditional Machine Learning Algorithm

[0070] A concrete performance prediction system based on automated machine learning algorithms and traditional machine learning algorithms specifically includes:

[0071] The concrete performance prediction system based on automated machine learning algorithms and traditional machine learning algorithms includes a prediction system for the compressive strength, slump, and porosity of normal concrete, as well as a prediction system for the compressive strength and flexural strength of ultra-high performance concrete, which can be interpreted as:

[0072] (1) Concrete Performance Prediction Module: The concrete performance prediction module based on automated machine learning algorithms and traditional machine learning algorithms includes a normal concrete performance prediction module and an ultra-high performance concrete performance prediction module. Among them, the normal concrete performance prediction module specifically includes a normal concrete slump prediction module, a compressive strength prediction module, and a porosity prediction module. Users can select the corresponding normal concrete performance prediction module based on their needs; the ultra-high performance concrete performance prediction module specifically includes an ultra-high performance concrete compressive strength prediction module and a flexural strength prediction module. Users can select the corresponding ultra-high performance concrete performance prediction module based on their needs.

[0073] (2) Input Module: Users input the mix ratio parameters of normal or ultra-high performance concrete in the corresponding concrete performance prediction module.

[0074] (3) Submission Module: After users complete the operation of the input module, they submit the mix ratio parameters. For the mix ratio parameters submitted by users, concrete performance prediction results corresponding to the users' mix ratio parameters are given.

[0075] Specifically, in the input module, the input module includes a list of concrete mix ratio parameters. After users select the corresponding concrete performance prediction module, they complete the input within the list of concrete mix ratio parameters.

[0076] In the input module for predicting the slump performance of normal concrete, the list of mix proportion parameters includes: cement dosage, fine aggregate dosage, coarse aggregate dosage, water dosage, high-range water reducer dosage, fly ash dosage, accelerating admixture dosage, and silica fume dosage; in the input module for predicting the compressive strength performance of normal concrete, the list of mix proportion parameters includes: cement dosage, fine aggregate dosage, coarse aggregate dosage, water dosage, high-range water reducer dosage, fly ash dosage, accelerating admixture dosage, silica fume dosage, and curing time; in the input module for predicting the porosity performance of normal concrete, the list of mix proportion parameters includes: binder dosage, fly ash dosage, blast furnace slag dosage, high-range water reducer dosage, water-binder ratio, fine-to-coarse aggregate ratio, curing method, and curing time.

[0077] In the input module for predicting the compressive strength performance of ultra-high performance concrete, the list of mix proportion parameters includes: cement dosage, silica fume dosage, fly ash dosage, slag dosage, nano-SiO 2 dosage, limestone powder dosage, quartz powder dosage, water dosage, aggregate dosage, water reducer dosage, steel fiber dosage, curing temperature, and curing time; in the input module for predicting the flexural strength performance of ultra-high performance concrete, the list of mix proportion parameters includes: cement dosage, silica fume dosage, fly ash dosage, slag dosage, nano-SiO 2 dosage, limestone powder dosage, quartz powder dosage, water dosage, aggregate dosage, water reducer dosage, steel fiber dosage, curing temperature, curing time, and specimen size.

[0078] In the submission module, after completing the submission module operation, the mix proportion parameters provided by the user are calculated using a concrete performance prediction method based on automated machine learning algorithms and traditional machine learning algorithms. After the calculation is completed, the corresponding prediction results of each algorithm are given, including the prediction results of automated machine learning algorithms: H2O, LightAutoML, Mljar, AutoGluon, EvalML, FlaML, and TPOT; and the prediction results of traditional machine learning algorithms: SVR, RF, and MLP.

[0079] In the second aspect of this embodiment, the multi-objective optimization design method and system for concrete mix proportion based on automated machine learning algorithms and multi-objective evolutionary algorithms include:

[0080] 1. The multi-objective optimization design method for concrete mix proportion based on automated machine learning algorithms and multi-objective evolutionary algorithms

[0081] The multi-objective optimization design method for concrete mix proportion based on automated machine learning algorithms and multi-objective evolutionary algorithms mainly includes the multi-objective optimization design method for normal concrete mix proportion and the multi-objective optimization design method for ultra-high performance concrete mix proportion, which can be interpreted as:

[0082] (1) Set the objective functions: For the concrete performance prediction method based on automated machine learning algorithms and traditional machine learning algorithms, set the AutoGluon algorithm as the objective functions for the porosity, compressive strength of normal concrete, and the compressive strength, flexural strength, and cost mix ratio design of ultra-high performance concrete; set the Mljar algorithm as the objective function for the slump mix ratio design of normal concrete; construct the objective function for the raw material cost of ultra-high performance concrete, and establish the mapping relationship between the mix ratio and concrete performance.

[0083] (2) Set the constraint conditions: Unify the raw material densities of normal concrete, establish the limits on the amounts of normal concrete raw materials used, and set the limits on the proportion of normal concrete raw materials used; unify the raw material densities and prices of ultra-high performance concrete, establish the limits on the amounts of ultra-high performance concrete raw materials used, set the limits on the proportion of ultra-high performance concrete raw materials used, and constrain the slump of ultra-high performance concrete and the curing method, curing temperature, and curing time of ultra-high performance concrete.

[0084] (3) Establish the two-objective mix ratio optimization design mathematical models for the compressive strength and slump of normal concrete; the three-objective mix ratio optimization design mathematical models for the compressive strength, slump, and porosity of normal concrete; the two-objective mix ratio optimization design mathematical models for the compressive strength and flexural strength of ultra-high performance concrete; the three-objective mix ratio optimization design mathematical models for the compressive strength, flexural strength, and cost of ultra-high performance concrete.

[0085] (4) Use four multi-objective evolutionary algorithms, namely NSGA-II, NSGA-III, AGEMOEA-II, and SMSEMOA, to solve the multi-objective mix ratio optimization design mathematical models. Compare the convergence performance, solution set distribution, solution set quality, and their uniformity and diversity of the approximate Pareto front under the four multi-objective evolutionary algorithms, and evaluate the applicability of the four multi-objective evolutionary algorithms to the concrete mix ratio optimization design problem.

[0086] (5) Use the entropy weight-TOPSIS method to determine the weights by the entropy value method and normalize each index, and then use the TOPSIS method to comprehensively evaluate the corresponding approximate Pareto front solution sets of the four multi-objective evolutionary algorithms. Based on the evaluation results, determine the mix ratio optimization design scheme with the best comprehensive performance and cost-benefit.

[0087] Specifically: The construction of a mathematical model for multi-objective optimization problems requires objective functions and constraint conditions. To establish the mapping relationship between mix proportions and concrete properties, in this embodiment, based on a dataset, by evaluating the accuracy and error of automated machine learning algorithms and traditional machine learning algorithms, and using SHAP theory for interpretability analysis of automated machine learning algorithms and traditional machine learning algorithms, it is determined that the AutoGluon algorithm has the best effect in predicting the porosity and compressive strength of normal concrete, as well as the compressive strength and flexural strength of ultra-high performance concrete; it is determined that the Mljar algorithm has the best effect in predicting the slump of normal concrete. Therefore, in this embodiment, the AutoGluon algorithm is used as the objective function for the design of the porosity, compressive strength of normal concrete, and the compressive strength, flexural strength, and cost mix proportions of ultra-high performance concrete; the Mljar algorithm is set as the objective function for the design of the slump mix proportion of normal concrete.

[0088] To achieve cost control of ultra-high performance concrete, in this embodiment, an objective function for the raw material cost of ultra-high performance concrete is set, and the specific formula is as follows:

[0089] Cost = ∑Cost i *m i

[0090] Where, Cost i (C, F,..., Fiber) respectively correspond to the raw material costs, and m i (C, F,..., Fiber) respectively correspond to the raw material masses.

[0091] In step (2), the constraint conditions for the mathematical model of multi-objective optimization design of concrete in this embodiment are: unify the densities of the raw materials of normal concrete, establish the limits on the amounts of the raw materials of normal concrete, and set the limits on the proportion of the amounts of the raw materials of normal concrete, as shown in Tables 1 and 2; unify the densities and prices of the raw materials of ultra-high performance concrete, establish the limits on the amounts of the raw materials of ultra-high performance concrete, and set the limits on the proportion of the amounts of the raw materials of ultra-high performance concrete, as shown in Tables 3 and 4; based on the "Technical Requirements for Ultra-High Performance Concrete (UHPC)" (T / CECS10107 - 2020), the spread of ultra-high performance concrete is constrained by whether the prediction result is greater than 650 mm. Unify the curing method of ultra-high performance concrete to hot water curing, set the curing temperature to 90 °C, and set the curing time to 28 days.

[0092] The densities of the raw materials of normal concrete and the limits on the amounts of the raw materials are shown in Table 1:

[0093] Table 1 Densities of the raw materials of normal concrete and limits on the amounts of the raw materials

[0094]

[0095] Note: Among them, C represents cement, FA represents fine aggregate, CA represents coarse aggregate, W represents water, SP represents water reducer, F represents fly ash, and SF represents silica fume.

[0096] The proportion limit of raw material dosage for normal concrete is shown in Table 2:

[0097] Table 2 Proportion Limit of Raw Material Dosage for Normal Concrete

[0098]

[0099] Note: Among them, B represents cementitious material, FA represents fine aggregate, CA represents coarse aggregate, W represents water, SP represents water reducer, F represents fly ash, and SF represents silica fume.

[0100] The limit of raw material dosage for normal concrete is restricted by the following formula:

[0101]

[0102] Among them, ρ i represents the density of raw material i, ε is the error, and here it is taken as 0.02.

[0103] The density and price of raw materials for ultra-high performance concrete are shown in Table 3. Here, the price and density of raw materials for ultra-high performance concrete can be modified according to the actual situation:

[0104] Table 3 Price and Density of Raw Materials for Ultra-High Performance Concrete

[0105]

[0106]

[0107] The proportion limit of raw material dosage for ultra-high performance concrete is shown in Table 4:

[0108] Table 4 Proportion Limit of Raw Material Dosage for Ultra-High Performance Concrete

[0109]

[0110] Note: Among them, B represents cementitious material, FA represents fine aggregate, W represents water, SP represents water reducer, F represents fly ash, SF represents silica fume, S represents blast furnace slag, NS represents nano-SiO 2 , L represents limestone, QP represents quartz powder, and Fiber represents steel fiber.

[0111] The limit of raw material dosage for ultra-high performance concrete is restricted by the following formula:

[0112] (∑m i / ρ i -1) 2 ≤ε 2

[0113] Among them, m i is the mass of the corresponding raw material, ρ i is the density of the corresponding raw material, and ε is the error, which is taken as 0.02 here.

[0114] In step (2), based on the "Technical Requirements for Ultra-High Performance Concrete (UHPC)" (T / CECS10107-2020), the slump flow of ultra-high performance concrete is constrained by whether the predicted result is greater than 650 mm. The curing method of ultra-high performance concrete is unified as hot water curing, the curing temperature is set at 90 °C, and the curing time is set at 28 days.

[0115] In step (4), this embodiment uses four multi-objective evolutionary algorithms, namely NSGA-II, NSGA-III, AGEMOEA-II, and SMSEMOA, to solve the multi-objective mix proportion optimization design mathematical model. Compare the convergence performance, solution set distribution, solution set quality, and their uniformity and diversity of the approximate Pareto front under the four multi-objective evolutionary algorithms, and evaluate the applicability of the four multi-objective evolutionary algorithms to the concrete mix proportion optimization design problem. The parameter settings of the multi-objective evolutionary algorithms are shown in Table 5. For the NSGA-Ⅲ algorithm, the Rieszs-energy method is used to generate 500 reference points in the objective space to guide the search process towards diversity and balance. The convergence performance, solution set distribution, solution set quality, and their uniformity and diversity indicators of the approximate Pareto front are evaluated by the convergence evaluation index HV, the uniformity evaluation index SP, the diversity evaluation index Δ

[0116] The parameter settings of the multi-objective evolutionary algorithms are shown in Table 5:

[0117] Table 5 Parameter Settings of Multi-Objective Evolutionary Algorithms

[0118]

[0119] For the NSGA-Ⅲ algorithm, the Rieszs-energy method is used to generate 500 reference points in the objective space to guide the search process towards diversity and balance.

[0120] In step (4), the convergence performance, solution set distribution, solution set quality, and their uniformity and diversity indicators of the approximate Pareto front are evaluated by the convergence evaluation index HV, the uniformity evaluation index SP, the diversity evaluation index Δ Line and the solution set quality evaluation index C.

[0121] The hypervolume (HV) index quantifies the frontier solution set S and the given reference point y refThe hypervolume between them is used to evaluate the performance of the Pareto front set. The hypervolumes of all solutions on the Pareto front are summed up to obtain the HV index value of the Pareto front set. The larger the HV index, the closer the approximate front obtained by the algorithm is to the true Pareto front. In the objective space, for the set S with respect to the reference point y ref The hypervolume is defined as:

[0122]

[0123] where is the Lebesgue measure. The hypervolume of a single individual is calculated by the following formula:

[0124]

[0125] where represents the set excluding x

[0126] The uniformity index SP measures the uniformity of the Pareto front set by comparing the distances between each front point. The distance d i represents the Manhattan distance between x i and x j in the objective space, where x j is the point with the smallest Manhattan distance to x i and x j is not equal to x i . When solving the Manhattan distance, different dimensions of different objectives have a great influence on d i , so it is necessary to normalize the front points in the objective space in advance. The calculation formula of d i is as follows:

[0127]

[0128] where F k (x i ) represents the k-th objective of the individual xi. The uniformity index SP is the standard deviation of di, and the calculation formula is as follows:

[0129]

[0130] where is the average value of d i , and |S| represents the number of individuals in the Pareto front. The smaller the SP value, the better the uniformity, and the larger the SP value, the worse the uniformity.

[0131] Diversity evaluation refers to evaluating the diversity and expansibility of the solution set through Δ Line without the need for the true Pareto front. Let β be N equally divided intervals in the range of 0 to 1 The midpoint, where N is the number of solutions in the Pareto set S, then for the i-th objective Δ Line is defined as:

[0132]

[0133] where F i (s) is the normalized approximate solution of solution s for the i-th objective. The closer the i-th objective Δ Line is to 0, the more evenly distributed the obtained approximate front is along the i-th objective. The overall Δ Line is defined as:

[0134]

[0135] where M represents M objectives. The closer Δ Line is to 0, the more diverse its distribution is. The larger its value, the more concentrated the value distribution is, that is, the worse the diversity is.

[0136] The solution set quality evaluation index, that is, the metric C(S 1 , S 2 ) is used to evaluate the quality between pairs of solution sets. The metric C(S 1 , S 2 ) can evaluate the quality of two sets of Pareto fronts S 1 and S 2 obtained by different algorithms. It does not require knowledge of the true Pareto front information and does not require manual setting of a reference set. It can directly compare the quality between two Pareto fronts. Its specific calculation formula is as follows:

[0137]

[0138] where S 1 and S 2 represent two sets of Pareto fronts, |S 2 | represents the number of solutions in set S 2 , x 1 ≥ x 2 means that x 1 weakly dominates x 2 . The metric C(S 1 , S 2 ) compares the quality between two sets of solutions by calculating the proportion of solutions in set S 2 that are weakly dominated by at least one solution in set S 1 , and its value ranges from 0 to 1. The higher the value of C(S 1 , S 2 ), the better the quality of set S 1 is than that of set S 2

[0139] In step (5), the entropy weight-TOPSIS method first determines the weights through the entropy weight method, normalizes each index, and then uses the TOPSIS method for comprehensive evaluation. Among them, the entropy weight method calculates the information entropy e of the index, and determines the weight of the index according to the impact of the relative change degree of the index on the overall system. The greater the dispersion degree of the index j, the smaller the entropy e j is, the greater the impact on the comprehensive evaluation, and the corresponding weight ω j is higher; the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) determines the best solution by comparing the similarity between each alternative solution and the ideal solution and the anti-ideal solution.

[0140] The entropy weight method first normalizes different indexes. The normalization formula for positive indexes is:

[0141]

[0142] The normalization formula for negative indexes is:

[0143]

[0144] where is the minimum value of target j, is the maximum value of target j.

[0145] By calculating the proportion of different solutions in the indexes, the proportion matrix is obtained, and its calculation formula is as follows:

[0146]

[0147] The information entropy e of each evaluation index j j is calculated according to the following formula:

[0148]

[0149] where n is the number of solutions.

[0150] The information redundancy d of evaluation index j j refers to the difference between the information entropy e j and its maximum upper bound, and its calculation formula is as follows:

[0151] d j = 1 - e j

[0152] The entropy weight ω refers to the ratio of the information redundancy of evaluation index j to the sum of the information redundancies of m evaluation indexes, that is:

[0153]

[0154] where m is the number of evaluation indexes.

[0155] Multiply the entropy weight by the proportion matrix to obtain the dimensionless index matrix F with self - weights, i.e., F = ω·P. Make a decision on the index matrix F obtained by the entropy weight method based on the TOPSIS method. First, construct the ideal solution F+ and the negative ideal solution F-. The ideal solution refers to the solution where all evaluation indicators reach the optimal values, and the negative ideal solution is the solution where all indicators reach the worst values. The specific formulas are as follows:

[0156]

[0157] By calculating the distances from each scheme i to the ideal solution and the negative ideal solution, we get and The calculation formula is:

[0158]

[0159] Based on and calculate the relative closeness S of each scheme to the ideal solution i :

[0160]

[0161] The smaller the distance to the ideal solution, the larger S i is. The smaller the distance to the negative ideal solution, the larger S i is. According to the magnitudes of S i of each scheme, perform sorting to obtain the scheme with the closest distance to the ideal solution.

[0162] On the other hand, this embodiment provides a concrete mix proportion optimization design system based on the automated machine learning algorithm and the multi - objective evolutionary algorithm: The multi - objective optimization design system for concrete mix proportion based on the automated machine learning algorithm and the multi - objective evolutionary algorithm mainly includes the multi - objective optimization design system for the mix proportion of normal concrete and the multi - objective optimization design system for the mix proportion of ultra - high performance concrete, which can be interpreted as:

[0163] (1) Mix proportion optimization design module: The multi - objective optimization design system for concrete mix proportion based on the automated machine learning algorithm and the multi - objective evolutionary algorithm includes the two - objective (slump and compressive strength) optimization design module for the mix proportion of normal concrete, the three - objective (slump, compressive strength and porosity) optimization design module for the mix proportion of normal concrete, the two - objective (compressive strength and flexural strength) optimization design module for the mix proportion of ultra - high performance concrete, and the three - objective (compressive strength, flexural strength and cost) optimization design module for the mix proportion of ultra - high performance concrete. Users can select the corresponding multi - objective optimization design module for concrete mix proportion based on their needs.

[0164] (2) Input module: Users input the parameter limits of the concrete mix proportion in the corresponding multi - objective optimization design module for concrete mix proportion.

[0165] (3) Submission Module: After the user completes the operations of the input module, the mix ratio parameter limits are submitted. For the mix ratio parameter limits submitted by the user, the concrete mix ratio results corresponding to the user's mix ratio parameter limits are given.

[0166] Specifically: In the input module, the input module includes a list of concrete mix ratio parameter limits. After the user selects the corresponding multi-objective optimization design module for concrete mix ratio, the input is completed within the list of concrete mix ratio parameter limits.

[0167] Based on the density and dosage limits of ordinary concrete raw materials specified in Table 1, the multi-objective optimization design of ordinary concrete mix ratio is carried out. The two-objective (slump and compressive strength) optimization design of ordinary concrete mix ratio is adjusted according to the dosage proportion limits of ordinary concrete raw materials specified in Table 2, and the dosage proportion of raw materials should not exceed the limit range; the three-objective (slump, compressive strength and porosity) optimization design of ordinary concrete mix ratio is adjusted according to the dosage proportion limits of ordinary concrete raw materials specified in Table 2, and the dosage proportion of raw materials should not exceed the limit range. Since the porosity prediction model of ordinary concrete does not consider the silica fume content, the silica fume dosage is set to 0 here.

[0168] Based on the prices and densities of ultra-high performance concrete raw materials specified in Table 3, the multi-objective optimization design of ultra-high performance concrete mix ratio is carried out. The two-objective (compressive strength and flexural strength) and three-objective (compressive strength, flexural strength and cost) optimization designs of ultra-high performance concrete mix ratio are adjusted according to the dosage proportion limits of ultra-high performance concrete specified in Table 4, and the dosage proportion of raw materials should not exceed the limit range.

[0169] In the submission module, after completing the operations of the submission module, the mix ratio parameters set by the user are calculated using the concrete mix ratio optimization design method based on the automatic machine learning algorithm and the multi-objective evolutionary algorithm. After the calculation is completed, the corresponding mix ratio optimization design results of each algorithm are given, including the multi-objective evolutionary algorithms: NSGA-II, NSGA-III, AGEMOEA-II and SMSEMOA mix ratio multi-objective optimization design results.

[0170] Data upload system, specifically including: a data upload module, which can be specifically divided into an ordinary concrete data upload module and an ultra-high performance concrete data module, which can be interpreted as: the user selects the ordinary concrete data upload module or the ultra-high performance concrete data upload module and inputs concrete data, specifically including concrete mix ratio parameters, performance data and cost data.

[0171] Based on the said data upload system, the method and system for performance prediction and multi-objective optimization design of ordinary concrete and ultra-high performance concrete can realize continuous update of the database and continuous iteration and optimization of the model.

[0172] As described above, it is only the preferred specific embodiment of the present application. However, the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A performance prediction and mix ratio multi-objective optimization design method for ordinary concrete and ultra-high performance concrete, characterized in that: include: Acquire a comprehensive data set, wherein the comprehensive data set includes mix ratio data, performance data, and cost data of concrete, wherein the concrete includes ordinary concrete and ultra-high performance concrete; Based on the preset target variables, correlation analysis is performed on the data in the comprehensive data set to obtain influencing factors with strong correlation with the preset target variables; wherein the preset target variables include ordinary concrete compressive strength performance, ordinary concrete slump performance, ordinary concrete porosity performance, ultra-high performance concrete compressive strength performance and ultra-high performance concrete flexural strength performance; Constructing a concrete performance prediction model corresponding to each of the preset target variables based on a machine learning algorithm, training and testing the concrete performance prediction model corresponding to each of the preset target variables based on influencing factors, and obtaining a trained and tested concrete performance prediction model; Perform interpretability analysis on each trained and tested concrete performance prediction model, and determine the optimal concrete performance prediction model corresponding to each preset target variable based on the interpretability analysis results; Set constraints, set the corresponding mix optimization design objective function based on the optimal concrete performance prediction model of each preset target variable, and construct the raw material cost objective function of ultra-high performance concrete; Constructing a multi-objective mix optimization design mathematical model based on the constraint conditions, the raw material cost objective function and each mix optimization design objective function; Solving the multi-objective mix optimization design mathematical model based on a multi-objective evolutionary algorithm to generate an approximate Pareto frontier solution set; Based on the approximate Pareto front solution set, a mix ratio optimization design scheme with the best comprehensive performance is determined.

2. The performance prediction and mix ratio multi-objective optimization design method of ordinary concrete and ultra-high performance concrete according to claim 1 is characterized in that: Before performing correlation analysis on the data in the comprehensive data set based on the preset target variable, the method also includes performing data cleaning on the comprehensive data set to obtain a preprocessed comprehensive data set, and performing correlation analysis based on the preprocessed comprehensive data set.

3. The performance prediction and mix ratio multi-objective optimization design method of ordinary concrete and ultra-high performance concrete according to claim 1 is characterized in that: The performing correlation analysis on the data in the comprehensive data set based on the preset target variable specifically includes: The correlation between each data in the comprehensive data set and the preset target variable is analyzed based on the Spearman correlation coefficient, and the influencing factors with strong correlation with the preset target variable are determined based on the correlation analysis results.

4. The performance prediction and mix ratio multi-objective optimization design method of ordinary concrete and ultra-high performance concrete according to claim 1 is characterized in that: The machine learning algorithms include automatic machine learning algorithms and traditional machine learning algorithms.

5. The performance prediction and mix ratio multi-objective optimization design method of ordinary concrete and ultra-high performance concrete according to claim 1 is characterized in that: The training and testing of the concrete performance prediction model corresponding to each of the preset target variables specifically includes: The comprehensive data set is divided into a training set and a test set, wherein the training set accounts for 80% of the comprehensive data set and the test set accounts for 20% of the comprehensive data set, wherein the training set is used to train a concrete performance prediction model, and the test set is used to verify the concrete performance prediction model. The prediction result of the concrete performance prediction model is evaluated according to the target loss function to obtain the trained and tested concrete performance prediction model.

6. The performance prediction and mix ratio multi-objective optimization design method of ordinary concrete and ultra-high performance concrete according to claim 1 is characterized in that: The interpretability analysis of each trained and tested concrete performance prediction model specifically includes: Based on the SHAP method, the SHAP value of each influencing factor in each algorithm model is calculated to obtain the interpretability analysis results. The optimal concrete performance prediction model corresponding to each preset target variable is determined based on the interpretability analysis results.

7. The performance prediction and mix ratio multi-objective optimization design method of ordinary concrete and ultra-high performance concrete according to claim 1 is characterized in that: The constraints include the density of ordinary concrete raw materials, the limit on the amount of ordinary concrete raw materials, the limit on the proportion of ordinary concrete raw materials, the density and price of ultra-high performance concrete raw materials, the limit on the amount of ultra-high performance concrete raw materials, the limit on the proportion of ultra-high performance concrete raw materials, the expansion of ultra-high performance concrete, the curing method of ultra-high performance concrete, the curing temperature of ultra-high performance concrete and the curing time of ultra-high performance concrete.

8. The performance prediction and mix ratio multi-objective optimization design method of ordinary concrete and ultra-high performance concrete according to claim 1 is characterized in that: The said determining the optimal design scheme of the mix ratio with the best comprehensive performance and cost-effectiveness based on the said Pareto frontier solution set specifically includes: Based on the multi-objective evolutionary algorithm, the multi-objective mix proportion optimization design mathematical model is solved to obtain several approximate Pareto frontier solution sets; based on the entropy weight-TOPSIS method, the weights are determined by the entropy value method and each indicator is normalized, and then the TOPSIS method is used to conduct a comprehensive performance evaluation on each approximate Pareto frontier solution set, and the mix proportion optimization design scheme with the best comprehensive performance and cost-effectiveness is determined based on the evaluation results.

9. A performance prediction and mix ratio multi-objective optimization design system for ordinary concrete and ultra-high performance concrete, characterized in that: include: A data acquisition module, used to obtain a comprehensive data set, wherein the comprehensive data set includes mix ratio data, performance data and cost data of concrete, wherein the concrete includes ordinary concrete and ultra-high performance concrete; A correlation analysis module is used to perform correlation analysis on the data in the comprehensive data set according to preset target variables to obtain influencing factors with strong correlation with the preset target variables; wherein the preset target variables include ordinary concrete compressive strength performance, ordinary concrete slump performance, ordinary concrete porosity performance, ultra-high performance concrete compressive strength performance and ultra-high performance concrete flexural strength performance; A concrete performance prediction model construction module is used to construct a concrete performance prediction model corresponding to each of the preset target variables according to a machine learning algorithm, train and test the concrete performance prediction model corresponding to each of the preset target variables based on influencing factors, and obtain the trained and tested concrete performance prediction model; perform interpretability analysis on each of the trained and tested concrete performance prediction models, and determine the optimal concrete performance prediction model corresponding to each of the preset target variables based on the interpretability analysis results; A multi-objective mix optimization design model construction module is used to set constraints, set corresponding mix optimization design objective functions based on the optimal concrete performance prediction model of each preset objective variable, and construct a raw material cost objective function of ultra-high performance concrete; a multi-objective mix optimization design mathematical model is constructed based on the constraints, raw material cost objective function and each mix optimization design objective function; The multi-objective evolutionary algorithm solving module is used to solve the multi-objective mix optimization design mathematical model according to the multi-objective evolutionary algorithm to generate an approximate Pareto front solution set; and determine the mix optimization design scheme with the best comprehensive performance based on the approximate Pareto front solution set.

Images