Method for predicting mechanical properties of cement-based negative Poisson's ratio configuration

Abstract

The invention discloses a cement-based negative Poisson's ratio configuration mechanical property prediction method. The method comprises the following steps: defining parameters of different cement-based negative Poisson's ratio configurations by using a unified standard; performing parametric modeling, and establishing a plurality of groups of analogue simulation models with different configurations; performing finite element analysis, acquiring performance data and parameter data of the configuration, and constructing an original data set; preprocessing the original data set; constructing a machine learning model, and setting an output matrix and a feature matrix; performing feature analysis on the feature matrix, and sorting key factors influencing the mechanical performance; different weights are set according to the sorting sequence, the machine learning regression model is trained and tested, a predictable model is obtained, and the mechanical property is predicted based on the predictable model. According to the method, the parameter characteristics are efficiently analyzed, the internal relation between the configuration parameters and the mechanical properties is fully excavated through machine learning, the research and development cost is reduced, and the material performance evaluation and optimization design process is accelerated.

Description

Technical Field

[0001] This invention relates to the field of mechanical property prediction technology, and in particular to a method for predicting the mechanical properties of cement-based negative Poisson's ratio configurations. Background Technology

[0002] Concrete, as the most widely used and consumed core structural material in the global engineering field, has long supported the construction of various infrastructure projects such as bridges, dams, and tunnels due to its significant advantages, including high strength, low cost, wide availability of raw materials, and ease of casting. However, due to the inherent limitations of its composition, concrete has inherent shortcomings such as tensile strength being only about one-tenth of its compressive strength, extremely small ultimate strain, and significant brittleness. Under complex service environments, it is highly susceptible to microcracks that can rapidly propagate, ultimately leading to structural fracture and damage. This not only severely restricts the long-term service performance and service life of engineering structures but also makes it difficult to meet the stringent requirements for material safety, stability, and durability in major strategic projects.

[0003] Against this backdrop, metamaterials, through structurally controlled artificial design methods, offer a breakthrough direction for the innovation of engineering material performance. Negative Poisson's ratio metamaterials stand out due to their unique "compression-expansion" mechanical response. They exhibit synchronous vertical contraction under uniaxial compression and synchronous vertical expansion under uniaxial tension. This characteristic allows them to actively compact towards the stress point under impact loads, effectively dispersing stress, while also possessing excellent impact resistance, indentation resistance, and high energy absorption efficiency. This material precisely matches the core requirements of major engineering projects for high energy absorption, large deformation, and high strength and toughness, and is expected to fundamentally break through the performance bottlenecks of traditional concrete, becoming a key path to drive the transformation of cement-based material performance and ensure the safety of strategic projects.

[0004] With the rapid development of big data and machine learning technologies, the materials field has begun to explore data-driven methods to predict and optimize material properties. However, there is still a lack of mature and effective technical solutions for predicting the performance of cement-based negative Poisson's ratio configurations. In addition, the configuration parameters are complex and diverse, and there are strong coupling effects between the parameters. Without a systematic classification of configuration features and efficient feature extraction rules, machine learning models often struggle to achieve high-precision predictions. Furthermore, traditional experimental methods are time-consuming and costly, which hinders the research and development process. Summary of the Invention

[0005] The purpose of this invention is to provide a method for predicting the mechanical properties of cement-based negative Poisson's ratio configurations, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, this invention provides a method for predicting the mechanical properties of cement-based negative Poisson's ratio configurations, comprising the following steps: S1. Obtain various cement-based negative Poisson's ratio configurations, and use a unified standard to define the parameters of different cement-based negative Poisson's ratio configurations, and convert the information of different cement-based negative Poisson's ratio configurations into quantitative parameters that can be recognized by machine learning. S2. Using parametric modeling, change and record configuration parameters to establish multiple sets of simulation models for different cement-based negative Poisson's ratio configurations; S3. The finite element analysis was used to simulate and analyze the multiple sets of simulation models of different cement-based negative Poisson's ratio configurations, and the performance data and parameter data of the cement-based negative Poisson's ratio configurations were collected to form the original data feature set. S4. Preprocess the original dataset and divide the preprocessed data feature set into a training set and a test set. S5. Construct a machine learning model, setting the performance data as the output matrix and the parameter data as the feature matrix; S6. Perform feature analysis on the feature matrix and rank the key factors that influence mechanical performance; S7. Set different weights according to the sorting order, train the machine learning regression model using the training set, evaluate the model performance through the error index of the test set, obtain a predictable model, and predict the mechanical properties of different cement-based negative Poisson's ratio configurations based on the predictive model.

[0007] Preferably, the cement-based negative Poisson's ratio configuration in step S1 includes two-dimensional double arrow, concave hexagon, quadrangular star, unidirectional three-rib hand shape, unidirectional four-rib hand shape, unidirectional six-rib hand shape, three-dimensional double arrow, inverted unidirectional three-rib hand shape, and inverted unidirectional four-rib hand shape.

[0008] Preferably, the unified standards in step S1 include: defining the width of the minimum symmetry unit of each configuration as the cell half-length, defining the length of the minimum symmetry unit of each configuration as the cell downsloping rib, defining the thickness of the skeleton of each configuration as the rod thickness; defining the stretching height of the configurations other than the three-dimensional double-arrow configuration as the cell depth; defining the minimum angle used to describe the skeleton connection in each configuration as the interior angle; defining the angle between the cell half-length and the cell downsloping rib used to limit the configuration in non-rotationally symmetric configurations as the exterior angle; and defining the symmetry rotation angle of the minimum symmetry unit as the exterior angle in rotationally symmetric configurations.

[0009] Preferably, step S3 involves using finite element analysis to perform simulation analysis on multiple sets of simulation models for different cement-based negative Poisson's ratio configurations, including: Using Python, the parameters of multiple sets of simulation models of different cement-based negative Poisson's ratio configurations set in parametric modeling are imported into the finite element analysis software to obtain the corresponding cement-based negative Poisson's ratio configuration models; The cement-based negative Poisson's ratio configuration model is divided into components: the main body and the upper and lower support plates. Define material properties, and after assembly, create a mechanical state analysis step before and after a dynamic explicit impact. Boundary conditions are set, the lower support plate is fixed to the ground, and the upper support plate is displaced vertically downwards with a certain amplitude. The main mesh of the cement-based negative Poisson's ratio configuration is generated, and the established mechanical model is subjected to finite element analysis to obtain the calculation results of dynamic explicit impact.

[0010] Preferably, in step S3, based on the calculation results of dynamic explicit impact, the performance data and parameter data of the cement-based negative Poisson's ratio configuration are collected to form an original data feature set. The performance data includes bearing capacity, displacement and energy dissipation, and the parameter data includes cell half-length distance, cell lower oblique rib, inner angle, outer angle, cell depth, number of horizontal and vertical cells.

[0011] Preferably, the preprocessing in step S4 includes removing all-zero features, outlier removal, and standardization.

[0012] Preferably, the feature matrix is ​​subjected to feature dimensionality reduction analysis based on PCA, feature correlation analysis is performed based on correlation heatmap, importance ranking is performed based on LightGBM algorithm, and feasibility analysis is performed based on SHAP.

[0013] Preferably, the importance ranking based on the LightGBM algorithm includes: constructing a prediction model based on the LightGBM algorithm, measuring and extracting key influencing factors by using the Gini index when each feature splits in the decision tree, and screening core features that have a significant regulatory effect on bearing capacity, displacement, and energy consumption.

[0014] Preferably, for bearing capacity, the priority of feature importance is as follows: interior angle, cell bottom rib, cell half-length distance, exterior angle, number of horizontal cells, number of vertical cells, and cell depth; for displacement, the priority is as follows: cell bottom rib, interior angle, exterior angle, cell half-length distance, number of vertical cells, number of horizontal cells, and cell depth; for energy consumption, the priority is as follows: cell bottom rib, interior angle, exterior angle, cell half-length distance, number of horizontal cells, number of vertical cells, and cell depth.

[0015] Preferably, in step S7, the weights are set in a stepwise manner according to the sorting order, with higher priority giving greater weight.

[0016] Therefore, this invention employs the aforementioned method for predicting the mechanical properties of cement-based negative Poisson's ratio configurations. It systematically classifies and encodes the parameters of typical cement-based negative Poisson's ratio configurations and performs parametric modeling. Through automated simulation, it collects multi-dimensional index data (such as bearing capacity, displacement, energy consumption, etc.), classifies and analyzes the parameter characteristics, and uses machine learning methods to achieve accurate output of the mechanical properties of the configuration to be predicted. This overcomes the problems of long testing cycles and high costs in traditional methods, and improves prediction accuracy and R&D efficiency.

[0017] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the cement-based negative Poisson's ratio configuration parameters of the present invention; Figure 3 This is a feature-reduced 3D PCA analysis diagram of the present invention; Figure 4 This is a heatmap of the feature correlation matrix of the present invention; Figure 5 This is a ranking diagram of the feature importance of the present invention; Figure 6 This is a summary diagram of the SHAP of the present invention; Figure 7 This is a regression fitting graph showing the true values ​​and predicted values ​​of the present invention; Figure 8 This is a comparison chart of the prediction results for the test set of this invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. In the description of the present invention, it should be noted that the terms "upper," "lower," "inner," "outer," etc., indicating orientation or positional relationships are based on the orientation or positional relationships shown in the accompanying drawings, or the orientation or positional relationships commonly used when the product of the invention is in use. They are only for the convenience of describing the present invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention.

[0020] Example like Figure 1As shown, this invention provides a method for predicting the mechanical properties of cement-based negative Poisson's ratio configurations, comprising the following steps: S1, such as Figure 2 As shown, various cement-based negative Poisson's ratio configurations are obtained, and parameters of different cement-based negative Poisson's ratio configurations are defined using a unified standard. The information of different cement-based negative Poisson's ratio configurations is converted into quantitative parameters that can be recognized by machine learning.

[0021] The obtained cement-based negative Poisson's ratio configurations are: two-dimensional double arrowhead, concave hexagon, quadrangular star, three-ribbed hand (forward), four-ribbed hand (forward), six-ribbed hand (forward), three-dimensional double arrowhead, three-ribbed hand (reverse), and four-ribbed hand (reverse). Each configuration has a different shape and needs to be defined using a unified standard. Observation shows that each configuration is either axially symmetric or rotationally symmetric; therefore, the minimum symmetric unit width is defined as the cell half-length, the minimum symmetric unit length as the cell downsloping rib, and the skeleton thickness as the rod thickness. Except for the three-dimensional double arrowhead configuration, the remaining configurations are all stretched from two-dimensional configurations; the stretching height is defined as the cell depth. Simultaneously, interior angles are defined to describe the minimum angle connecting the skeleton in the configuration; exterior angles are defined to limit the angle between the cell half-length and the cell downsloping rib in non-rotationally symmetric configurations. In rotationally symmetric configurations, since the angle between the half-length of the cell and the lower oblique rib of the cell is fixed at 90 degrees, the exterior angle is used to define the symmetry rotation angle of the smallest symmetric unit.

[0022] S2. Using parametric modeling, the configuration parameters were changed and recorded to establish 106 sets of simulation models for different cement-based negative Poisson's ratio configurations.

[0023] S3. Finite element analysis was used to simulate and analyze multiple sets of different cement-based negative Poisson's ratio configurations, and performance and parameter data of the cement-based negative Poisson's ratio configurations were collected to form an original data feature set. Specifically, this includes: Using Python, the parameters of multiple sets of simulation models for different cement-based negative Poisson's ratio (NPPR) configurations, set in parametric modeling, were imported into the ABAQUS finite element analysis software to obtain the corresponding cement-based NPPR configuration models. The cement-based NPPR configuration models were then divided into a main body and upper and lower support plates; material properties were defined, and analysis steps were created after assembly. Specifically, two analysis steps were created for the main body of the configuration based on dynamic explicit impact simulation: the first analysis step was for the mechanical state before the dynamic explicit impact, and the second analysis step was for the mechanical state after the dynamic explicit impact. After creating the analysis steps, boundary conditions were set in the interaction module of the ABAQUS software. The lower support plate was fixed to the ground, and the upper support plate was set to displace vertically downwards with a certain amplitude. Then, the main body of the cement-based NPPR configuration was meshed, with the mesh control attribute being tetrahedral structured mesh. The established mechanical model was then calculated in the job module of the ABAQUS software to obtain the calculation results of the dynamic explicit impact.

[0024] By using Python for code writing and execution, a complete ABAQUS simulation program can be written according to the above process. This program can read the code of different configuration models that have already been established and perform batch simulations, saving time.

[0025] Simultaneously, based on the calculation results of dynamic explicit impact, performance data and parameter data of cement-based negative Poisson's ratio configurations were collected to form a raw data feature set. The performance data includes bearing capacity, displacement, and energy dissipation, while the parameter data includes cell half-length distance, cell lower diagonal ribs, interior angles, exterior angles, cell depth, number of horizontal ribs, and number of vertical ribs. A summary table of the raw data information for 106 sets of mechanical performance samples of cement-based negative Poisson's ratio configurations is shown in Table 1.

[0026] Table 1. Summary of raw data information for 106 groups of cement-based negative Poisson's ratio configuration mechanical properties.

[0027] Table 1 describes the original dataset of steel rebar rust inhibitor sample data using the mean, standard deviation, and five-number summary. The minimum, 0.25 (first quartile), 0.5 (second quartile), 0.75 (third quartile), and maximum values ​​in the header row of Table 1 are the five key values ​​of the five-number summary. They are an important part of descriptive statistics and help to quickly understand the characteristics of the dataset.

[0028] S4. Preprocess the original dataset and divide the preprocessed data feature set into training set and test set.

[0029] The preprocessing process includes removing all-zero features, outlier removal, and standardization. The cement-based Poisson's ratio configuration information and parameters are converted into machine learning-recognizable quantitative features. The preprocessed data feature set is then divided into training and testing sets in an 8:2 ratio. For the three features of bearing capacity, displacement, and energy dissipation, columns with values ​​of 0 in all samples are removed, and extreme outliers caused by statistical or simulation errors are eliminated using the 3σ criterion. All features are standardized to have a mean of 0 and a standard deviation of 1 to eliminate the impact of dimensional differences on model training.

[0030] S5. Construct a machine learning model. Set the performance data, namely bearing capacity (N), displacement (mm), and energy consumption (J), as the output matrix Y of the machine learning model. Set the parameter data, namely cell half-length, cell underslope rib, inner angle, outer angle, cell depth, number of horizontal cells, and number of vertical cells, as the feature matrix X to prepare for subsequent model training and prediction.

[0031] S6. Perform feature analysis on the feature matrix and rank the key factors that influence mechanical performance.

[0032] Specifically, based on PCA, feature dimensionality reduction analysis and feature correlation analysis are performed on the feature matrix to identify key parameters affecting mechanical performance.

[0033] The PCA dimensionality reduction algorithm (n_components=3) is used to process the feature matrix and construct a three-dimensional principal component space coordinate system. Energy consumption is used as a color mapping parameter to intuitively explore the spatial distribution patterns of samples in the principal component coordinate system from a macroscopic perspective. Figure 3 The results show that the point cloud data exhibits high dispersion and significant color distribution span. This phenomenon reveals that there are significant differences between samples in the main feature dimensions, and the differences in energy consumption may be due to the combined effect of multiple factors.

[0034] Feature correlation analysis is visualized using correlation heatmaps, aiming to quantitatively assess the degree of nonlinear correlation between functional groups and between them and corrosion inhibition efficiency indicators. Figure 4 This intuitively reflects the correlation distribution characteristics of the data features of each sample. For example, the interior angle and the half-length of the cell in the cement-based negative Poisson's ratio configuration are strongly positively correlated, suggesting that such parameters may coexist synergistically. The lower rib of the cell shows a significant negative correlation with parameters such as the half-length of the cell and the interior angle, indicating that there is a high probability of a negative correlation effect between the two.

[0035] A prediction model is constructed based on the LightGBM algorithm. Key influencing factors are extracted and ranked by the Gini index at which each feature splits in the decision tree, and core features with significant regulatory effects on bearing capacity, displacement, and energy consumption are selected. For example... Figure 5As shown, the results indicate that the priority of feature importance is as follows: interior angle > cell downsloping rib > cell half-length distance > exterior angle > number of horizontal features > number of vertical features > cell depth; for displacement, the priority is: cell downsloping rib > interior angle > exterior angle > cell half-length distance > number of vertical features > number of horizontal features > cell depth; for energy dissipation, the priority is: cell downsloping rib > interior angle > exterior angle > cell half-length distance > number of horizontal features > number of vertical features > cell depth. These results clearly reveal that features such as interior angle, cell downsloping rib, cell half-length distance, and exterior angle rank highly in terms of influence on the three key mechanical properties of load-bearing capacity, displacement, and energy dissipation. Among these, the interior angle has the most prominent regulatory effect on load-bearing capacity, while the cell downsloping rib dominates the influence on displacement and energy dissipation. The cell half-length distance and exterior angle also exhibit significant influence in each performance dimension.

[0036] like Figure 6 The diagram shown is a summary of the SHAP (Shapley Additive Explanations) of this invention. By using SHAP to create a swarm diagram, the positive or negative influence of different features such as interior angles and cell undersloping ribs on the model's predicted outputs of load-bearing capacity, displacement, and energy dissipation can be observed. This provides researchers with interpretable mechanistic analysis and structural performance control basis to guide subsequent design. In the "SHAP Value Swarm Diagram - Load-Bearing Capacity," features such as exterior angles, interior angles, and cell undersloping ribs have a significant impact on the model output. Among them, high values ​​(red) and low values ​​(blue) of the interior angle feature are distributed in both the positive and negative regions of the SHAP value, and the scattered points indicate that the relationship between this feature and load-bearing capacity is not a simple linear monotonicity, but may be affected by the interaction of other features. In the "SHAP Value Swarm Diagram - Displacement," features such as interior angles, cell undersloping ribs, and cell half-length distance have a significant impact. For example, the values ​​of cell undersloping ribs are distributed in different ranges of the SHAP value, reflecting a complex mechanism for their control over displacement. In the "SHAP value beehive diagram - energy dissipation," features such as interior angles, number of horizontal elements, and cell depth have a significant impact. For example, high values ​​for the number of horizontal elements are concentrated in the positive SHAP region, while low values ​​are distributed in the negative SHAP region, demonstrating the nonlinear influence of this feature on energy dissipation. In summary, these SHAP beehive diagrams intuitively present the influence patterns of various features on the mechanical properties of negative Poisson's ratio configurations, providing an interpretable analytical tool for a deeper understanding of the structural performance regulation mechanism and optimization of structural design.

[0037] S7. Different weights are assigned according to the sorting order. The machine learning regression model is trained using the training set, and its performance is evaluated using the error index on the test set to obtain a predictable model. Based on this predictive model, the mechanical properties of different cement-based negative Poisson's ratio configurations are predicted. The weights are set in a step-by-step manner according to the sorting order; the higher the priority, the greater the weight.

[0038] Specifically, the training set is used to model a LightGBM machine learning regression model suitable for tabular data and multiple output features, and the model is then evaluated based on the test set error metrics (MSE, MAE, R). 2 The hyperparameters of the network model (MAPE) are adjusted, and the model configuration is further optimized using cross-validation or grid search methods to finally obtain a predictive model that can predict the mechanical properties under different negative Poisson's ratio configurations. For the cement-based negative Poisson's ratio configuration to be predicted, the sample features are preprocessed and then input into the pre-trained regression prediction model to output the corresponding prediction results.

[0039] When performing energy consumption prediction, the model hyperparameters are optimized using Grid Search combined with 5-fold cross-validation, with R... 2 The optimal configuration for the scoring criteria was selected as follows: learning_rate: 0.05, max_depth: -1, min_child_samples: 10, n_estimators: 600, num_leaves: 31, reg_lambda: 5.0, subsample: 0.8 (learning rate: 0.05, maximum tree depth: -1, minimum number of samples per leaf node: 10, number of iterations: 600, number of leaf nodes: 31, L2 regularization coefficient: 5.0, sample rate: 0.8). The error comparison between the training and test sets is shown in Table 2, and the regression fitting plot of the true and predicted values ​​is shown in [see table]. Figure 7 See the comparison of prediction results on the test set. Figure 8 From the error results of the training and test sets in Table 2, we can see that R... 2 All values ​​are greater than 0.91, indicating that the established machine learning model has a good fitting ability for energy consumption performance and is suitable for regression training and prediction tasks.

[0040] Table 2 Comparison of errors between training and test sets

[0041] For bearing capacity and displacement indices, similar parameters to those used in the energy dissipation method were used for prediction, and good prediction results were achieved. Specifically, the bearing capacity training set R... 2 The value is 0.992, and the test set R is... 2 The value is 0.901; the displacement training set R 2 The value is 0.972, and the test set R is... 2 The value is 0.911. The above results show that the machine learning model established in this invention has a good fitting ability to the dataset obtained from automated simulation, and is suitable for regression training and prediction tasks.

[0042] To verify the effectiveness of the prediction model of this invention, it was compared with the XGBoost model and the Catboost model using the same training and test sets. The error results of the training and test sets of different models are shown in Table 3.

[0043] Table 3 Comparison of errors between training and test sets for different models

[0044] As can be seen from the table above, the dataset performs significantly better in the LightGBM and XGBoost models than in the Catboost model. However, in the XGBoost model, although the training set R... 2 Slightly better than the LightGBM model, however, its test set R 2 The decrease indicates that the data in the XGBoost model may have been overfitted, leading to its poor performance in the prediction process.

[0045] Therefore, this invention adopts the above-mentioned method for predicting the mechanical properties of cement-based negative Poisson's ratio configurations. By effectively classifying and encoding the configuration features, it achieves efficient accumulation of simulation data. Through efficient analysis of parameter features, it fully explores the intrinsic relationship between configuration parameters and mechanical properties using machine learning, thereby reducing R&D costs and accelerating the process of material performance evaluation and optimization design.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting the mechanical properties of cement-based negative Poisson's ratio configurations, characterized in that, Including the following steps: S1. Obtain various cement-based negative Poisson's ratio configurations, and use a unified standard to define the parameters of different cement-based negative Poisson's ratio configurations, and convert the information of different cement-based negative Poisson's ratio configurations into quantitative parameters that can be recognized by machine learning. S2. Using parametric modeling, change and record configuration parameters to establish multiple sets of simulation models for different cement-based negative Poisson's ratio configurations; S3. The finite element analysis was used to simulate and analyze the multiple sets of simulation models of different cement-based negative Poisson's ratio configurations, and the performance data and parameter data of the cement-based negative Poisson's ratio configurations were collected to form the original data feature set. S4. Preprocess the original dataset and divide the preprocessed data feature set into a training set and a test set. S5. Construct a machine learning model, setting the performance data as the output matrix and the parameter data as the feature matrix; S6. Perform feature analysis on the feature matrix and rank the key factors that influence mechanical performance; S7. Set different weights according to the sorting order, train the machine learning regression model using the training set, evaluate the model performance through the error index of the test set, obtain a predictable model, and predict the mechanical properties of different cement-based negative Poisson's ratio configurations based on the predictive model.

2. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 1, characterized in that, In step S1, the cement-based negative Poisson's ratio configurations include two-dimensional double arrows, concave hexagons, quadrangular stars, unidirectional three-rib hand shapes, unidirectional four-rib hand shapes, unidirectional six-rib hand shapes, three-dimensional double arrows, inverse three-rib hand shapes, and inverse four-rib hand shapes.

3. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 2, characterized in that, The unified standards in step S1 include: defining the width of the minimum symmetry unit of each configuration as the cell half-length, defining the length of the minimum symmetry unit of each configuration as the cell downsloping rib, defining the thickness of the skeleton of each configuration as the rod thickness; defining the stretching height of the configurations other than the three-dimensional double-arrow configuration as the cell depth; defining the minimum angle used to describe the skeleton connection in each configuration as the interior angle; defining the angle between the cell half-length and the cell downsloping rib used to limit non-rotationally symmetric configurations as the exterior angle; and defining the symmetry rotation angle of the minimum symmetry unit as the exterior angle in rotationally symmetric configurations.

4. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 1, characterized in that, Step S3 involves using finite element analysis to simulate and analyze multiple sets of different cement-based negative Poisson's ratio configurations, including: Using Python, the parameters of multiple sets of simulation models of different cement-based negative Poisson's ratio configurations set in parametric modeling are imported into the finite element analysis software to obtain the corresponding cement-based negative Poisson's ratio configuration models; The cement-based negative Poisson's ratio configuration model is divided into components: the main body and the upper and lower support plates. Define material properties, and after assembly, create a mechanical state analysis step before and after a dynamic explicit impact. Boundary conditions are set, the lower support plate is fixed to the ground, and the upper support plate is displaced vertically downwards with a certain amplitude. The main mesh of the cement-based negative Poisson's ratio configuration is generated, and the established mechanical model is subjected to finite element analysis to obtain the calculation results of dynamic explicit impact.

5. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 4, characterized in that: In step S3, based on the calculation results of dynamic explicit impact, the performance data and parameter data of the cement-based negative Poisson's ratio configuration are collected to form the original data feature set. The performance data includes bearing capacity, displacement and energy dissipation, and the parameter data includes cell half-length distance, cell lower diagonal rib, inner angle, outer angle, cell depth, number of horizontal and vertical cells.

6. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 1, characterized in that: Preprocessing in step S4 includes removing all-zero features, outlier removal, and standardization.

7. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 5, characterized in that, Step S6 involves feature analysis of the feature matrix, including: performing feature dimensionality reduction analysis based on PCA, performing feature correlation analysis based on correlation heatmap, ranking importance based on the LightGBM algorithm, and performing feasibility analysis based on SHAP.

8. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 7, characterized in that: Importance ranking based on the LightGBM algorithm includes: constructing a prediction model based on the LightGBM algorithm, measuring and ranking key influencing factors by the Gini index when each feature splits in the decision tree, and screening core features that have significant regulatory effects on bearing capacity, displacement, and energy consumption.

9. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 8, characterized in that: For load-bearing capacity, the priority of features is as follows: interior angle, cell bottom rib, cell half-length distance, exterior angle, number of horizontal features, number of vertical features, and cell depth. For displacement, the priority is as follows: cell bottom rib, interior angle, exterior angle, cell half-length distance, number of vertical features, number of horizontal features, and cell depth. For energy dissipation, the priority is as follows: cell bottom rib, interior angle, exterior angle, cell half-length distance, number of horizontal features, number of vertical features, and cell depth.

10. The method for predicting the mechanical properties of cement-based negative Poisson's ratio configuration according to claim 9, characterized in that: In step S7, the weights are set in a step-by-step manner according to the sorting order, with higher priority giving greater weight.

Images