Method for evaluating variability of mechanical properties of steel bars and predicting strength based on large model

Abstract

The application relates to a method for evaluating the variability of mechanical properties and predicting the strength of steel bars based on a large model. The method collects basic information, geometric parameters, mechanical property test data and test environment data of a steel bar sample, constructs a structured detection data set, extracts and normalizes the detection data, constructs a steel bar feature matrix, and constructs a mechanical property label based on the test results. A deep neural network model is used to establish the mapping relationship between the steel bar features and the mechanical properties, and the yield strength, tensile strength and elongation after fracture of the steel bar are predicted. On this basis, statistical analysis is performed on the prediction results, the average value, standard deviation and coefficient of variation are calculated, the variability of the steel bar performance is evaluated, a comprehensive quality evaluation index is constructed, the strength and quality of the steel bar are evaluated, and the model is updated and optimized through new data.

Description

Technical Field

[0001] This application relates to the field of intelligent detection and artificial intelligence analysis technology for engineering materials, and in particular to a method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model. Background Technology

[0002] With the continuous expansion of my country's transportation infrastructure construction and the increasing demands for structural safety, steel reinforcement, as the main load-bearing material in concrete structures, directly affects the safety and durability of bridges, tunnels, highways, and large-scale transportation engineering structures. The yield strength, tensile strength, and elongation after fracture of steel reinforcement are typically tested through tensile tests, and the quality of the steel reinforcement material is evaluated according to relevant standards. In engineering testing practice, to ensure the reliability of test results and the stability of engineering material quality, testing institutions usually need to systematically test and record data from a large number of steel reinforcement samples, gradually accumulating a large amount of data on steel reinforcement mechanical properties. Therefore, how to utilize existing test data to systematically analyze and predict the performance of steel reinforcement materials, improve material quality assessment capabilities, and enhance the efficiency of test data utilization has become an important requirement in the field of engineering material testing.

[0003] In related technologies, the evaluation of the mechanical properties of reinforcing steel mainly relies on the results of traditional tensile tests. The performance of the reinforcing steel material is evaluated through statistical analysis of the test data or empirical judgment. For example, some testing institutions perform statistical analysis on the performance of the same batch of reinforcing steel by calculating the mean, standard deviation, or coefficient of variation of the steel strength indicators. Meanwhile, some studies have attempted to use traditional machine learning methods to predict the mechanical properties of reinforcing steel, such as support vector machines, random forests, or shallow neural network models. These methods can achieve the analysis and prediction of reinforcing steel mechanical property data to a certain extent, but they usually require manually constructing feature parameters and rely on limited sample data for modeling and analysis.

[0004] However, the above methods still have certain limitations in practical applications. First, traditional statistical analysis methods usually only analyze single batches of test data, making it difficult to reflect the overall variation of steel reinforcement mechanical properties under different production batches, different testing cycles, and different engineering environments. Second, existing machine learning methods have limited capabilities in processing multi-source test data and complex nonlinear relationships, making it difficult to fully utilize the large-scale test data accumulated over a long period in the testing system. In addition, most existing technologies focus on the analysis of single test results, lacking the ability to systematically evaluate the overall variability characteristics of steel reinforcement mechanical properties, and also making it difficult to achieve stable prediction of steel reinforcement strength indicators.

[0005] To address the aforementioned issues, a method for assessing the variability of the mechanical properties of reinforcing steel and predicting its strength based on a large model is now designed. Summary of the Invention

[0006] This application provides a method for evaluating the variability of steel bar mechanical properties and predicting strength based on a large model, in order to solve the technical problems in related technologies such as low utilization rate of steel bar mechanical property test data, difficulty in systematically evaluating the variability of mechanical properties of different batches of steel bars, and difficulty in effectively predicting steel bar strength indicators based on historical test data.

[0007] Firstly, a method for assessing the variability of the mechanical properties of reinforcing steel and predicting its strength based on a large model is provided, including the following steps:

[0008] S1. Rebar Testing Data Acquisition: Collect basic information, geometric parameter data, mechanical property test data, and test environment data of rebar samples, and construct a structured rebar testing dataset;

[0009] S2. Construction of rebar feature data: The rebar detection data is processed by feature extraction and organization to construct a rebar feature vector, and then normalized and statistically extracted to form a rebar feature matrix;

[0010] S3. Construction of Mechanical Property Labels for Steel Reinforcing Bars: Based on the tensile test results of steel reinforcing bars, the yield strength, tensile strength and elongation after fracture are extracted to construct a label vector and label matrix for the mechanical properties of steel reinforcing bars.

[0011] S4. Construction of a large-scale model for predicting the mechanical properties of reinforcing bars: The deep learning model is trained using the aforementioned feature matrix and mechanical property label matrix of reinforcing bars to establish the mapping relationship between reinforcing bar features and mechanical properties, and the prediction results of the mechanical properties of reinforcing bars are output.

[0012] S5. Evaluation of the variability of steel bar mechanical properties: Based on the predicted results of steel bar mechanical properties, calculate the statistical characteristic parameters of steel bar mechanical properties, including the mean, standard deviation and coefficient of variation, to evaluate the fluctuation of steel bar performance;

[0013] S6. Reinforcing bar strength prediction and quality assessment: The predicted strength of the reinforcing bar is compared with the standard strength, and a comprehensive quality evaluation index is constructed by combining the coefficient of variation to assess the quality of the reinforcing bar material;

[0014] S7. Model Update and Performance Optimization: Introduce new rebar detection data into the model training process and update model parameters through incremental learning to improve model prediction accuracy and adaptability.

[0015] In some embodiments, during the rebar testing data acquisition step, basic rebar information is collected and managed through a laboratory information management system, and a unique testing identification code is generated for each rebar sample to achieve full-process association and traceability of rebar testing data. The basic rebar information includes at least the rebar specification and model, rebar manufacturer, rebar production batch, rebar arrival time, project name, testing commissioning unit, and testing number. Through the aforementioned unique identification code, the basic rebar information is uniformly associated with geometric parameter data, mechanical performance test data, and test environment data to form a complete rebar testing record. This ensures the consistency and traceability of testing data during acquisition, storage, and subsequent analysis, and provides a reliable data foundation for subsequent rebar feature data construction and model training.

[0016] In some embodiments, the geometric parameter data acquisition includes obtaining the diameter, length, and surface morphology features of the reinforcing bar. The diameter is determined by taking multiple measurements at different locations along the reinforcing bar's axial direction and averaging them to reduce the impact of localized wear or non-uniformity on the measurement results. The length is obtained using a laser rangefinder or measuring tool. The surface morphology information is obtained by acquiring image data of the reinforcing bar surface using an image acquisition device to characterize surface corrosion, cracks, or defects. The geometric parameter data is input into the detection system database through automatic acquisition or manual entry and is associated with the reinforcing bar detection number for storage, thereby achieving a structured expression of the reinforcing bar's geometric features and providing basic data support for subsequent feature extraction and model analysis.

[0017] In some embodiments, the test environment data includes test temperature, test humidity, test time, and test equipment number. This test environment data is collected synchronously and recorded in the testing database during the steel reinforcement mechanical property test. Since environmental conditions affect the test results of the steel reinforcement mechanical properties, introducing test environment data as a feature variable can reflect the impact of environmental factors on the test results during subsequent feature construction. The test environment data is stored in association with the steel reinforcement test number and, together with the steel reinforcement geometric parameter data and mechanical property test data, constitutes a complete test data record, thereby improving the accuracy and reliability of steel reinforcement test data analysis.

[0018] In some embodiments, the rebar feature data construction step involves uniformly organizing the geometric features, production information features, inspection environment features, and time features in the rebar inspection data, and constructing rebar feature vectors. During feature processing, the min-max normalization method is used to standardize each feature variable, mapping features of different dimensions to a unified numerical range to eliminate the impact of differences in feature scales on model training. Simultaneously, statistical feature information is extracted by calculating the statistical mean and standard deviation of the feature data to reflect the overall distribution of the rebar inspection data. Based on this, all rebar sample feature vectors are combined to form a rebar feature matrix, which serves as the input data for the subsequent rebar mechanical property prediction model.

[0019] In some embodiments, the large-scale model for predicting the mechanical properties of reinforcing bars is a deep neural network model. The model includes an input layer, at least one hidden layer, and an output layer. The input layer is used to receive the feature matrix data of the reinforcing bars. The hidden layer is used to perform multi-layer mapping on the input features through a weight matrix and a nonlinear activation function to extract deep feature information from the reinforcing bar detection data. The output layer is used to output the prediction results of the mechanical properties of the reinforcing bars. The model realizes the complex relationship modeling between the feature data of the reinforcing bars and the mechanical performance indicators through multi-layer nonlinear mapping, thereby improving the model's ability to express and predict multi-source data.

[0020] In some embodiments, during the model training process, the mechanical property labels of steel bars are used as the target output, and the model parameters are optimized by minimizing the error between the model prediction results and the true labels. The error is expressed as mean square error as the loss function, and the model parameters are continuously updated through iterative training so that the model prediction results gradually approach the true values. During the training process, the model is trained in batches on historical steel bar detection data so that it can learn the mapping relationship between steel bar feature data and mechanical properties, thereby improving the model's generalization ability and prediction accuracy.

[0021] In some embodiments, the step of assessing the variability of the mechanical properties of reinforcing steel bars involves statistical analysis of the predicted mechanical property data of reinforcing steel bar samples from the same production batch, calculating the batch average, standard deviation, and coefficient of variation. The coefficient of variation is the ratio of the standard deviation to the average, used to characterize the degree of performance fluctuation of the reinforcing steel bar material. By comparing the coefficient of variation with a preset threshold, the stability of the reinforcing steel bar material performance is determined. When the coefficient of variation exceeds the threshold, it is determined that the batch of reinforcing steel bars has a risk of performance fluctuation, thereby achieving a quantitative assessment of the variability of the reinforcing steel bar material performance.

[0022] In some embodiments, in the steel reinforcement strength prediction and quality assessment step, the ratio of the predicted yield strength of the steel reinforcement to the standard specified strength is calculated, and a comprehensive quality evaluation index is constructed by combining the steel reinforcement performance variation coefficient. The evaluation index is obtained by weighting the strength index and the stability index. The quality of steel reinforcement samples and batches of steel reinforcement materials is evaluated by this comprehensive index. When the evaluation index is lower than a preset threshold, it is determined that the steel reinforcement material has a quality risk, thereby realizing a comprehensive assessment of the quality of steel reinforcement materials.

[0023] In some embodiments, the model update and performance optimization steps involve fusing newly added rebar detection data with existing training data to construct a new training dataset, and using incremental training to update the rebar mechanical property prediction model. During the model update process, the model parameters are iteratively optimized using a gradient descent algorithm to reduce prediction errors. By periodically introducing new data for model training, the model can continuously learn the performance variation characteristics of different batches of rebar materials, thereby improving the model's prediction accuracy and adaptability to changes in data distribution.

[0024] The beneficial effects of the technical solution provided in this application include:

[0025] 1. This application constructs a multi-dimensional feature data system by uniformly collecting and structurally modeling basic information, geometric parameters, mechanical performance test data, and test environment data of reinforcing bars, thereby realizing the integrated utilization of multi-source test data. Compared with existing analysis methods that rely solely on single test data, this application can fully explore the potential information in reinforcing bar test data and improve data utilization efficiency;

[0026] 2. This application employs a large model to perform deep feature extraction and nonlinear modeling of rebar characteristic data, establishing a mapping relationship between rebar characteristics and yield strength, tensile strength, and elongation after fracture. Compared with traditional statistical methods or shallow machine learning models, this approach can more accurately characterize complex data relationships, thereby significantly improving the accuracy of rebar mechanical property prediction.

[0027] 3. This application uses statistical analysis of the predicted mechanical properties of reinforcing steel bars to calculate the average value, standard deviation, and coefficient of variation, thereby achieving a quantitative description of the fluctuations in the material properties of reinforcing steel bars. Compared to existing methods that rely solely on experience, this application can systematically evaluate the performance stability of different batches of reinforcing steel bars, improving the scientific rigor of material quality analysis.

[0028] 4. Based on the prediction of steel reinforcement strength, this application introduces a performance variability index to construct a comprehensive quality evaluation index, thereby achieving a joint evaluation of the strength level and performance stability of steel reinforcement materials. Compared with evaluation methods based solely on a single strength index, this application can more comprehensively reflect the quality status of steel reinforcement materials, thus improving the accuracy and reliability of quality judgment.

[0029] 5. By comparing the predicted strength of the reinforcing bars with the standard strength and combining the results of variability assessment, this application can identify batches of reinforcing bars with large performance fluctuations or insufficient strength, thereby achieving early warning of reinforcing bar material quality risks. Compared with traditional post-inspection judgment methods, this application has forward-looking analytical capabilities, which helps to improve the safety of engineering structures;

[0030] 6. This application introduces a model update and incremental learning mechanism, which can continuously optimize model parameters using newly added detection data, enabling the model to gradually adapt to different production batches and changes in the detection environment. Compared with static models, this application has stronger generalization ability and long-term stability, making it suitable for continuous operation and dynamic application scenarios of rebar detection systems. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 The overall flowchart of the method for evaluating the variability of mechanical properties of steel bars and predicting their strength based on a large model provided in the embodiments of the present invention;

[0033] Figure 2 This is a flowchart of the data acquisition process for rebar inspection provided in an embodiment of the present invention;

[0034] Figure 3 A flowchart for constructing rebar feature data provided in an embodiment of the present invention;

[0035] Figure 4 A flowchart illustrating the construction and training process of the steel bar mechanical property prediction model provided in this embodiment of the invention;

[0036] Figure 5 A flowchart for assessing the variability and quality of reinforcing steel bars provided in this embodiment of the invention;

[0037] Figure 6 This is a schematic diagram of the rebar detection data structure provided in an embodiment of the present invention;

[0038] Figure 7 This is a schematic diagram of the construction of the rebar feature matrix provided in an embodiment of the present invention;

[0039] Figure 8 This is a schematic diagram of the steel bar mechanical property prediction model provided in an embodiment of the present invention;

[0040] Figure 9This is a schematic diagram of the spatial distribution of the variability characteristics of the mechanical properties of reinforcing bars provided in an embodiment of the present invention;

[0041] Figure 10 This is a schematic diagram of the steel reinforcement quality evaluation and model update mechanism provided in an embodiment of the present invention;

[0042] Figure 11 Statistical distribution diagram of mechanical properties of different batches of steel bars provided in embodiments of the present invention;

[0043] Figure 12 A comparative statistical chart of the coefficient of variation of the mechanical properties of steel bars provided in the embodiments of the present invention;

[0044] Figure 13 A comparison chart of predicted strength and standard strength of reinforcing bars and a comprehensive quality evaluation chart provided for embodiments of the present invention;

[0045] Figure 14 The training error and accuracy convergence diagram of the steel bar mechanical property prediction model provided in the embodiments of the present invention;

[0046] Figure 15 This is a comparison chart of prediction performance and computing power consumption before and after incremental model updates provided in an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0048] This application provides a method for evaluating the variability of the mechanical properties of steel bars and predicting their strength based on a large model. This method can solve the technical problems in related technologies, such as low utilization rate of steel bar mechanical property test data, difficulty in systematically evaluating the variability of mechanical properties of different batches of steel bars, and difficulty in effectively predicting steel bar strength indicators based on historical test data.

[0049] Please see Figures 1-15 A method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel bars based on a large model includes the following steps:

[0050] (I) Data Collection for Reinforcing Steel Testing

[0051] The acquisition of steel bar test data is a crucial foundational step in this invention for assessing the variability of steel bar mechanical properties and predicting strength. Its main purpose is to obtain multi-source information data of steel bar materials during the testing process and form a structured test dataset, providing a reliable data foundation for subsequent steel bar mechanical property modeling analysis and variability assessment.

[0052] In the implementation of this invention, the steel bar testing data mainly comes from engineering material testing laboratories or steel bar quality testing systems, and is uniformly collected and managed through systems such as steel bar tensile testing equipment, geometric parameter testing equipment, and laboratory information management system (LIMS).

[0053] 1. Collection of basic information of steel bar samples

[0054] Before steel reinforcement samples enter the testing process, basic information about the samples needs to be registered. This includes:

[0055] Rebar specifications, rebar manufacturer, rebar production batch, rebar arrival time, project name, testing commissioning unit, and rebar testing number.

[0056] The above information is entered into the laboratory information management system, and a unique test identification code is generated for each steel bar sample, thereby realizing unified identification and full-process traceability management of steel bar test data.

[0057] By establishing a unified numbering system, the basic information of steel bar samples can be linked with subsequent test data, providing a basic data structure for steel bar test data management and statistical analysis.

[0058] 2. Reinforcing bar geometric parameter data acquisition

[0059] After completing the basic information registration, the geometric parameters of the reinforcing bar samples are tested. The main geometric parameters of the reinforcing bars include: diameter, length, and surface morphology. In the actual testing process, a reinforcing bar diameter measuring instrument, laser rangefinder, or machine vision inspection equipment can be used to collect the geometric parameters of the reinforcing bars.

[0060] Among them, the diameter of the reinforcing bar is usually measured by multiple points, that is, the diameter is measured multiple times at different locations of the reinforcing bar, and the average value is calculated as the diameter parameter of the reinforcing bar, so as to reduce the measurement error caused by local wear or surface unevenness.

[0061] The length of the reinforcing bars can be measured using a laser rangefinder or a ruler and recorded in the inspection database.

[0062] For information on the surface morphology of reinforcing bars, image data of the reinforcing bar surface can be obtained through image acquisition equipment to record the surface condition of the reinforcing bars, such as rust, cracks or surface defects.

[0063] The aforementioned geometric parameter data are automatically collected by the testing equipment or manually entered into the testing system database and associated with the steel bar sample testing number.

[0064] 3. Data collection for mechanical property tests of reinforcing bars

[0065] After completing the geometric parameter measurements, the mechanical properties of the steel reinforcement samples were tested.

[0066] The mechanical properties of reinforcing steel bars are typically tested using a universal testing machine for tensile testing. During the test, an axial tensile force is applied to the reinforcing steel bar specimen through a loading system, while sensors simultaneously collect real-time data on the stress and deformation of the specimen.

[0067] The testing equipment can automatically record the stress-strain curves of steel reinforcement specimens during loading and calculate the main mechanical properties of the steel reinforcement according to the testing standards, including:

[0068] Yield strength, tensile strength, and elongation after fracture.

[0069] The test results are automatically transmitted to the detection system database through the data interface of the test equipment for storage, thereby reducing errors from manual recording and improving data acquisition efficiency.

[0070] 4. Data Acquisition for the Test Environment

[0071] To ensure the accuracy of the test data analysis, it is also necessary to record the test environment information simultaneously during the data acquisition process. The test environment data mainly includes: test temperature, test humidity, test equipment number, and test time.

[0072] Since environmental conditions may have a certain impact on the results of material tests, recording environmental parameters helps to interpret or correct the test data in subsequent data analysis.

[0073] 5. Structured storage of detection data

[0074] After completing the above-mentioned data collection, the basic information of the reinforcing bars, geometric parameter data, mechanical performance test data, and test environment data are stored in the reinforcing bar testing database.

[0075] Each rebar inspection record in the database includes at least the following data fields:

[0076] The information includes the steel bar inspection number, steel bar specifications, steel bar production batch, steel bar diameter, steel bar length, test time, test environment parameters, yield strength, tensile strength, and elongation after fracture.

[0077] The above data fields can be used to form a sample of structured rebar inspection data.

[0078] 6. Data quality control

[0079] To ensure the reliability of rebar testing data, data quality control is required during the data acquisition process.

[0080] First, the integrity of the collected raw data is checked to ensure that each rebar inspection record contains the necessary data fields. Second, outlier identification is performed on the inspection data. For example, when the inspection data deviates significantly from the normal range, the inspection results are confirmed through manual review or repeated testing. Third, duplicate inspection samples are uniformly numbered and managed to avoid duplicate statistics of inspection data.

[0081] The above data quality control measures can improve the accuracy and reliability of rebar testing data.

[0082] 7. Construction of Reinforcing Steel Inspection Data Set

[0083] After completing data collection and quality control, the rebar inspection data was compiled into a rebar inspection data set: ;

[0084] in: This represents the feature information of the i-th steel bar sample, including steel bar geometric parameters, production batch information, and test environment parameters, etc.

[0085] This represents the mechanical performance indicators of the i-th steel bar sample, including yield strength, tensile strength, and elongation after fracture.

[0086] Through the above steps, a rebar detection dataset containing multi-source information can be constructed, providing a data foundation for subsequent training of rebar mechanical performance prediction models and variability assessment analysis.

[0087] (II) Construction of Reinforcing Steel Feature Data

[0088] After completing the collection of rebar inspection data and forming a structured inspection database, it is necessary to organize and construct the features of the collected raw inspection data to form feature data suitable for subsequent rebar mechanical property modeling and analysis. The main purpose of constructing rebar feature data is to extract information that reflects the performance characteristics of rebar materials from the raw inspection data and to uniformly express different types of data, thereby forming a multi-dimensional feature representation.

[0089] 1. Processing of raw detection features

[0090] Based on the rebar testing data collected in step (I), each rebar testing record typically contains multiple types of information, including basic rebar information, geometric parameter data, test environment data, and mechanical performance test data. When constructing features, the aforementioned raw data is first categorized and organized, and information reflecting the rebar material properties and testing conditions is extracted as feature data.

[0091] In the implementation of this invention, the characteristic data of reinforcing bars mainly includes the following categories:

[0092] (1) Geometric characteristics of reinforcing bars, including parameters such as diameter and length of reinforcing bars;

[0093] (2) Characteristics of steel bar production information, including steel bar production batch, steel mill number, etc.;

[0094] (3) Detect environmental characteristics, including environmental parameters such as test temperature and test humidity;

[0095] (4) Detection time characteristics, including test date or detection cycle information.

[0096] By organizing the above information, we can obtain the original feature set of the steel bar test samples.

[0097] 2. Construction of Reinforcing Bar Feature Vectors

[0098] After completing the original feature processing, it is necessary to unify the various feature data to form the feature vector of the steel bar sample.

[0099] For the i-th rebar sample, its feature vector can be represented as: ;

[0100] in: This shows the diameter of the nth rebar sample; This represents the length of the rebar in the nth rebar sample; Indicates steel bar production batch information; Indicates the detection time characteristic; This indicates the characteristics of the test environment.

[0101] By using the above method, detection data from different sources can be uniformly represented as numerical feature vectors, which facilitates subsequent data processing and analysis by the model.

[0102] 3. Feature normalization processing

[0103] Since the numerical ranges of different feature variables differ—for example, the diameter of steel bars is usually measured in millimeters, while environmental humidity is expressed as a percentage—it is necessary to normalize the feature data to avoid the impact of different units on the subsequent modeling process.

[0104] In this invention, the minimum-maximum normalization method is used to standardize the feature data, and its calculation method is as follows:

[0105] ;

[0106] in: Represents the normalized eigenvalues; This represents the minimum value of the corresponding feature; This represents the maximum value of the corresponding feature.

[0107] Normalization can map the value ranges of different features to the same interval, thereby improving the stability of subsequent model training.

[0108] 4. Statistical Feature Extraction for Reinforcing Steel Inspection

[0109] In the steel bar testing data, the testing data of different batches of steel bars usually have certain statistical characteristics. Therefore, in the feature construction process, it is also necessary to extract the statistical characteristics of the steel bar testing data to reflect the overall distribution of steel bar material properties.

[0110] In this invention, the central tendency of rebar characteristics is represented by calculating the statistical mean of rebar inspection data. The calculation formula is as follows:

[0111] ;

[0112] in: The value represents the characteristic mean; 𝑛 represents the number of steel bar inspection samples; This represents the feature value of the nth steel bar sample.

[0113] Meanwhile, to describe the dispersion of rebar inspection data, the standard deviation of the characteristic data can be calculated: ;;

[0114] By statistically analyzing the mean and standard deviation, we can reflect the distribution characteristics of steel bar test data among different samples, providing basic information for subsequent analysis of the variability of steel bar mechanical properties.

[0115] 5. Feature Matrix Construction

[0116] After completing feature vector construction, normalization, and statistical feature extraction, the feature data of all rebar samples can be combined to form a rebar feature matrix: ;

[0117] Where: M represents the feature matrix of rebar inspection data; n represents the number of rebar samples; Let represent the normalized feature vector of the i-th rebar sample.

[0118] By constructing a steel reinforcement feature matrix, the multidimensional feature data of different steel reinforcement samples can be uniformly represented, thereby providing input data for the subsequent establishment of a steel reinforcement mechanical property prediction model.

[0119] 6. Feature Data Output

[0120] After the above feature construction steps, a set of rebar feature data containing multi-dimensional information can be obtained. This feature data not only includes the geometric parameters of the rebar, but also integrates the production batch information, testing environment information, and testing time information, thus forming a data representation that can reflect the comprehensive characteristics of the rebar material.

[0121] The constructed steel reinforcement characteristic data will serve as input data for subsequent steel reinforcement mechanical performance analysis models, used for steel reinforcement mechanical performance variability assessment and strength prediction analysis.

[0122] (III) Construction of Mechanical Performance Labels for Reinforcing Steel

[0123] After completing the collection of rebar test data and the construction of rebar characteristic data, it is necessary to construct labels for the rebar mechanical property test results to form the target variable data required for model training. The main purpose of constructing rebar mechanical property labels is to extract key mechanical indicators from the rebar tensile test results, and to uniformly express and process them to form label data that can reflect the performance characteristics of rebar materials.

[0124] 1. Source of mechanical property labels

[0125] The mechanical property labels for reinforcing bars are primarily derived from test data obtained during tensile tests. In these tests, an axial tensile force is applied to the reinforcing bar specimen using a universal testing machine, and the stress and strain changes during the loading process are recorded in real time.

[0126] Based on tensile test data, the main mechanical properties of reinforcing steel can be calculated, including:

[0127] Yield strength, tensile strength, and elongation after fracture.

[0128] The aforementioned indicators are important parameters for evaluating the mechanical properties of reinforcing steel materials, and are also the most commonly used judgment indicators in the process of reinforcing steel quality inspection. Therefore, in this invention, the aforementioned mechanical performance indicators are used as label data for the reinforcing steel performance analysis model.

[0129] 2. Construction of Mechanical Performance Label Vectors

[0130] To facilitate unified processing of steel bar mechanical property data, it is necessary to represent steel bar mechanical property indicators as label vectors.

[0131] For the i-th steel reinforcement sample, its mechanical property label can be expressed as: ;

[0132] in: This represents the yield strength of the nth steel bar sample; This represents the tensile strength of the nth steel bar sample; This represents the elongation after fracture of the nth steel bar sample; the mechanical property label is the actual value measured by the steel bar test.

[0133] The above method can be used to convert the results of steel bar tensile tests into a unified label vector representation.

[0134] 3. Mechanical property label data processing

[0135] Since the mechanical performance indicators of different steel bar samples may fluctuate, the label data needs to be sorted and processed during the label construction process to reduce the impact of abnormal data on subsequent analysis results.

[0136] In this invention, the overall performance level of the steel reinforcement sample is represented by calculating the average value of the label data. The calculation method is as follows:

[0137] ;

[0138] in: This represents the average value of the steel reinforcement mechanical properties label; 𝑛 represents the number of steel reinforcement test samples. This indicates the mechanical property label for the nth steel bar sample.

[0139] By calculating the average value of the label data, the overall performance level of the steel reinforcement material in the sample set can be reflected.

[0140] 4. Discrete Feature Description of Labels

[0141] To further describe the distribution of steel bar mechanical property data among different samples, it is also necessary to analyze the dispersion of the label data.

[0142] In this invention, the dispersion of the steel reinforcement mechanical property data is represented by calculating the variance of the label data. The calculation formula is as follows:

[0143] ;

[0144] in: This indicates the variance of the labeled mechanical properties of the reinforcing steel.

[0145] By calculating the label variance, the fluctuation of steel reinforcement material properties across different samples can be reflected.

[0146] 5. Construction of Mechanical Property Tag Matrix

[0147] After completing the label vector construction and data processing, the mechanical property labels of all steel reinforcement samples can be combined to form a label matrix:

[0148] ;

[0149] Where: Y represents the steel bar mechanical property label matrix; n represents the number of steel bar samples; This represents the mechanical property label vector for the i-th steel bar sample.

[0150] By constructing a label matrix, the mechanical performance indicators of all steel bar samples can be uniformly represented, thereby providing target variable data for the training of subsequent steel bar mechanical performance prediction models.

[0151] 6. Tag Data Output

[0152] By following the steps above, the mechanical performance label for reinforcing steel can be constructed. The resulting label data contains the main mechanical performance indicators of the reinforcing steel material and can reflect the performance differences of the reinforcing steel material between different test samples.

[0153] The constructed steel bar mechanical property label matrix, together with the steel bar feature data matrix constructed in the previous steps, will serve as the input and target data for the subsequent steel bar performance analysis model, and will be used for steel bar mechanical property variability assessment and strength prediction analysis.

[0154] (iv) Construction of a large-scale model for predicting the mechanical properties of reinforcing bars

[0155] After completing the collection of rebar inspection data, the construction of rebar characteristic data, and the construction of rebar mechanical performance labels, it is necessary to build a rebar mechanical performance prediction model to achieve the analysis of rebar mechanical performance variability and strength prediction. The main purpose of building the large-scale rebar mechanical performance prediction model is to establish a mapping relationship between rebar characteristic data and mechanical performance indicators by conducting in-depth correlation analysis of multi-dimensional features in rebar inspection data, thereby realizing the prediction and analysis of rebar strength indicators.

[0156] 1. Model Input and Output Definition

[0157] In this invention, the steel bar mechanical property prediction model takes steel bar characteristic data as input and steel bar mechanical property labels as output.

[0158] Assume the characteristic matrix of the reinforcing steel is represented as:

[0159] ;

[0160] in: Let represent the feature vector of the nth rebar sample.

[0161] The output of the steel reinforcement mechanical property prediction model is the predicted steel reinforcement performance:

[0162] ;

[0163] in: To predict yield strength; To predict tensile strength; To predict elongation after fracture; the prediction result is the model output value, which corresponds to the true label.

[0164] 2. Construction of a structural model for predicting the mechanical properties of reinforcing steel bars

[0165] To fully uncover the potential patterns in rebar testing data, this invention employs a deep neural network structure to construct a rebar mechanical property prediction model. This model consists of an input layer, multiple feature mapping layers, and an output layer.

[0166] During the model calculation process, a nonlinear mapping function is used to convert the steel reinforcement characteristic data into mechanical performance indicators. The calculation expression is as follows: ;

[0167] in: is the input feature vector for the reinforcing bars; is the model weight matrix; is the bias parameter; It is a non-linear activation function; This represents the hidden layer after feature mapping.

[0168] Through multi-layer feature mapping, deep feature information in rebar inspection data can be extracted step by step.

[0169] 3. Mechanical property prediction calculation

[0170] After feature mapping is completed, the predicted mechanical properties of the reinforcing steel are calculated through the output layer, and its expression is as follows: ;

[0171] in: This is the output layer weight matrix; These are the output layer bias parameters; This is the vector for the predicted mechanical properties of the steel reinforcement sample.

[0172] The above calculations can yield the predicted results of the steel bar's yield strength, tensile strength, and elongation after fracture.

[0173] 4. Model Training and Parameter Optimization

[0174] To improve the model's prediction accuracy, it is necessary to train and optimize the model parameters using rebar inspection data. During model training, the model parameters are updated by comparing the differences between the model's prediction results and the actual inspection results.

[0175] In this invention, mean squared error is used as the objective function for model training, and its calculation method is as follows: ;

[0176] Where: 𝐽 represents the model loss function; Labels indicating the true mechanical properties of steel reinforcement samples; This represents the model prediction result; 𝑛 represents the sample size.

[0177] By continuously iterating and optimizing the model parameters, the loss function is gradually reduced, thereby improving the model's predictive ability.

[0178] 5. Predictive output of steel reinforcement mechanical properties

[0179] After model training is complete, the trained model can be used to predict the mechanical properties of new steel reinforcement test samples. For a new steel reinforcement sample, its feature vector is input into the model to obtain the predicted mechanical properties of the steel reinforcement.

[0180] To evaluate the consistency between the predicted results and the actual detection results, the prediction error can be calculated: ;

[0181] in: This indicates the prediction error for the steel reinforcement sample.

[0182] By analyzing the prediction error, the predictive performance of the model can be evaluated.

[0183] 6. Model Updates and Continuous Optimization

[0184] As the rebar detection system accumulates new detection data, this new data can be added to the training dataset to continuously train and update the model, thereby continuously improving the model's prediction accuracy.

[0185] Through continuous data updates and model optimization, the steel bar mechanical property prediction model can gradually adapt to the performance changes of different batches of steel bar materials, thereby improving the accuracy and reliability of steel bar material quality assessment.

[0186] 7. Technical Function of this Section

[0187] By constructing a large-scale model for predicting the mechanical properties of reinforcing steel bars, the following can be achieved:

[0188] In-depth analysis of the multidimensional characteristics of rebar testing data, prediction of rebar mechanical performance indicators, analysis of rebar performance change trends, and provision of data foundation for subsequent rebar performance variability assessment.

[0189] This provides data support for the quality testing of engineering materials and the assessment of structural safety.

[0190] (v) Assessment of the variability of the mechanical properties of steel bars

[0191] After completing the construction of the large model for predicting the mechanical properties of steel bars and obtaining the prediction results, it is necessary to analyze the fluctuation of steel bar material properties among different samples in order to evaluate the stability of steel bar material properties.

[0192] The main purpose of assessing the variability of steel reinforcement mechanical properties is to analyze the predicted results of steel reinforcement mechanical properties through statistical analysis methods, thereby identifying the performance variation patterns of steel reinforcement materials in different production batches or test samples.

[0193] 1. Obtaining the prediction results of mechanical properties

[0194] Using the steel bar mechanical property prediction model constructed in step (iv), the mechanical properties of the steel bar test samples are predicted, thus obtaining a dataset of predicted mechanical properties for the steel bar samples. For the i-th steel bar sample, the prediction result can be expressed as: ;

[0195] in: This indicates the predicted yield strength of the steel reinforcement sample; Indicates the predicted tensile strength of the steel reinforcement sample; This represents the predicted elongation after fracture of the steel reinforcement sample.

[0196] The above method can be used to obtain the predicted mechanical properties data of multiple steel bar samples.

[0197] 2. Calculation of average batch mechanical properties

[0198] During the quality inspection of reinforcing steel bars, steel bars from the same production batch typically exhibit similar material properties. Therefore, statistical analysis of the mechanical properties of steel bars from the same batch is necessary during variability assessment.

[0199] Suppose a certain batch of steel bars contains n test samples, its average mechanical properties can be expressed as: ;

[0200] in: This indicates the average yield strength of the batch of steel bars; 𝑚 indicates the number of steel bar samples tested in the batch.

[0201] By calculating the batch average, the overall mechanical performance level of the steel reinforcement material in that batch can be reflected.

[0202] 3. Calculation of mechanical property dispersion

[0203] To describe the performance fluctuations of steel reinforcement materials within the same batch, it is necessary to calculate the dispersion of the mechanical properties of the steel reinforcement.

[0204] In this invention, the performance fluctuation is represented by calculating the standard deviation of the steel reinforcement properties. The calculation method is as follows: ;

[0205] in: This represents the standard deviation of the yield strength of this batch of steel bars.

[0206] The larger the standard deviation, the more significant the fluctuation in the performance of the steel bars between samples.

[0207] 4. Calculation of the coefficient of variation of steel reinforcement properties

[0208] To further evaluate the stability of steel reinforcement performance, the coefficient of variation of its mechanical properties can be calculated. The coefficient of variation reflects the proportional relationship between fluctuations in steel reinforcement performance and the average level; it is calculated as follows: ;

[0209] in: This represents the coefficient of variation of the mechanical properties of the steel bars in this batch.

[0210] The larger the coefficient of variation, the worse the stability of the steel reinforcement performance; the smaller the coefficient of variation, the more stable the steel reinforcement material performance.

[0211] 5. Batch performance stability evaluation

[0212] The performance stability of steel reinforcement materials can be evaluated based on the calculated coefficient of variation. In practical applications, the threshold of the coefficient of variation can be set according to the quality management requirements of engineering materials.

[0213] When the coefficient of variation of the reinforcing steel exceeds the set threshold, it indicates that there is a significant risk of performance fluctuation in the batch of reinforcing steel, and further testing or quality verification of the batch of reinforcing steel is required.

[0214] The above methods can be used to identify batches of steel bars with large performance fluctuations, thus providing a reference for the quality management of engineering materials.

[0215] 6. Output of variability assessment results

[0216] After completing the variability analysis of steel reinforcement properties, the variability assessment results of the steel reinforcement material can be output. The assessment results typically include:

[0217] Batch average strength index, strength standard deviation, performance coefficient of variation, and stability evaluation results.

[0218] The above evaluation results can comprehensively reflect the performance changes of steel reinforcement materials among different test samples, and provide basic data for subsequent steel reinforcement strength prediction and quality risk assessment.

[0219] 7. Technical Functions of This Section

[0220] By evaluating the variability of the mechanical properties of steel bars, the following can be achieved:

[0221] ① Quantitative analysis of the performance fluctuations of steel reinforcement materials;

[0222] ② Identify the performance differences between different batches of steel reinforcement materials;

[0223] ③ Provide a basis for evaluating the quality stability of reinforcing steel materials;

[0224] ④ Provide basic data for subsequent early warning of steel reinforcement quality risks.

[0225] (vi) Reinforcing steel strength prediction and quality assessment

[0226] After completing the construction of the steel reinforcement mechanical property prediction model and the assessment of steel reinforcement performance variability, it is necessary to predict and analyze the strength level of the steel reinforcement material, and to comprehensively evaluate the quality of the steel reinforcement material based on the prediction results. The main purpose of steel reinforcement strength prediction and quality assessment is to determine whether the steel reinforcement material meets the requirements for engineering use by using model prediction and index analysis on steel reinforcement test data, thereby providing a basis for decision-making in engineering material quality management.

[0227] 1. Strength prediction of steel reinforcement samples

[0228] Using the steel reinforcement mechanical property prediction model constructed in step (iv), the strength of the new steel reinforcement test samples is predicted. For the k-th steel reinforcement sample, its input feature vector can be used to calculate the predicted strength index of the steel reinforcement: ;

[0229] in: This indicates the predicted yield strength of the steel reinforcement sample; This indicates the predicted tensile strength of the steel reinforcement sample.

[0230] Through model calculations, the strength prediction results of steel reinforcement materials can be obtained before a complete test is conducted.

[0231] 2. Calculation of standard strength ratio

[0232] To evaluate whether the strength of reinforcing steel meets the requirements for engineering use, it is necessary to compare the predicted strength with the standard specified strength. In this invention, the quality of reinforcing steel is evaluated by calculating the ratio of steel strength to concrete strength. The calculation method is as follows: ;

[0233] in: Indicates the ratio of steel reinforcement strength; Indicates the predicted yield strength of the steel reinforcement; This indicates the minimum yield strength specified in the steel reinforcement standard.

[0234] When the strength ratio is greater than or equal to 1, it indicates that the predicted strength of the steel reinforcement meets the standard requirements.

[0235] 3. Construction of Comprehensive Quality Evaluation Indicators

[0236] To comprehensively evaluate the quality of reinforcing steel, this invention constructs a reinforcing steel quality evaluation index based on the predicted strength of the reinforcing steel. This index considers both the predicted strength and the performance stability of the reinforcing steel, and its calculation method is as follows: ;

[0237] in: Indicates the quality evaluation index of steel reinforcement; Indicates the ratio of steel reinforcement strength; This represents the coefficient of variation of the steel reinforcement performance; α and β are weighting coefficients, and satisfy α+β=1.

[0238] By introducing the stability factor of steel reinforcement performance, the quality of steel reinforcement materials can be evaluated more comprehensively.

[0239] 4. Batch Reinforcing Steel Quality Assessment

[0240] In actual engineering testing, steel reinforcement is usually managed in batches for quality control. Therefore, after evaluating the quality of an individual steel reinforcement sample, it is necessary to conduct an overall quality assessment of the same batch of steel reinforcement.

[0241] By statistically analyzing the quality evaluation indicators of the same batch of steel reinforcement samples, the overall quality level of that batch of steel reinforcement materials can be obtained. When the quality evaluation indicators of a batch of steel reinforcement are lower than a set threshold, it can be determined that the batch of steel reinforcement materials has a quality risk.

[0242] 5. Output of quality assessment results

[0243] After completing the reinforcement strength prediction and quality assessment, the reinforcement material quality assessment results can be output. The assessment results typically include:

[0244] Predicted yield strength of reinforcing bars, predicted tensile strength of reinforcing bars, strength ratio of reinforcing bars, quality evaluation index of reinforcing bars, and batch quality evaluation results.

[0245] The above assessment results can provide a reference for the quality management of engineering materials, and also provide data support for material testing institutions to formulate quality control measures.

[0246] 6. Technical Functions of This Section

[0247] Through methods for predicting steel reinforcement strength and assessing its quality, the following can be achieved:

[0248] ① Predict and analyze the strength level of steel reinforcement materials;

[0249] ② Determine whether the steel reinforcement material meets the engineering design standards;

[0250] ③ Conduct a comprehensive analysis of the strength level and performance stability of the reinforcing steel;

[0251] ④ Provide a basis for quality control and risk warning of engineering materials.

[0252] (vii) Model update and performance optimization

[0253] During the long-term operation of the rebar testing system, the data size of the rebar testing database will continue to expand as engineering projects progress and testing data accumulates. To ensure that the rebar mechanical property prediction model can continuously adapt to new testing data characteristics, the model needs to be dynamically updated and its performance optimized, thereby continuously improving the model's prediction accuracy and stability.

[0254] 1. New test data access

[0255] When the detection system generates new rebar detection data, the new data needs to be connected to the existing rebar detection database and processed according to the methods in steps (1) to (3) to construct the corresponding rebar feature data and mechanical performance label data.

[0256] Let the set of newly added rebar inspection samples be: ;

[0257] in: This represents the newly added set of steel reinforcement test samples; 𝑚 represents the number of newly added samples; This indicates the nth newly added steel bar sample.

[0258] By merging the newly detected samples with the original dataset, a new training dataset can be formed.

[0259] 2. Incremental Model Training

[0260] After obtaining new training data, the steel bar mechanical property prediction model needs to be incrementally trained so that the model can learn new data features.

[0261] In this invention, model parameter updates are achieved by minimizing the model prediction error. For newly added data samples, the prediction error can be expressed as: ;

[0262] in: Labels indicating the actual mechanical properties of newly added steel reinforcement samples; This indicates the model's prediction results; This indicates the prediction error.

[0263] By analyzing the prediction error, model parameters can be adjusted to reduce the prediction error.

[0264] 3. Model parameter update

[0265] During model training, the model parameters are updated using optimization algorithms. The model parameter update process can be represented as:

[0266] ;

[0267] in: This represents the model parameters at the nth iteration; represents the updated model parameters; 𝜂 represents the learning rate; ∇L represents the gradient of the loss function.

[0268] By continuously iterating and updating the model parameters, the predictive ability of the model can be gradually improved.

[0269] 4. Model Performance Evaluation

[0270] After updating the model, its predictive performance needs to be evaluated to verify the effectiveness of the updated model. In this invention, model performance can be evaluated by calculating the model's predictive accuracy, and the calculation method is as follows:

[0271] ;

[0272] Where: Acc represents the model prediction accuracy; This represents the actual mechanical properties of the steel reinforcement sample; This indicates the model's prediction results.

[0273] By calculating the prediction accuracy of the model, we can assess the performance improvement after the model update.

[0274] 5. Model Update Deployment

[0275] When the model performance evaluation results meet the preset requirements, the updated model can be deployed to the rebar detection system for subsequent rebar detection data analysis and performance prediction.

[0276] In practical applications, model updates can be performed at certain intervals, such as according to the testing batch or testing time cycle, so as to ensure that the model can continuously adapt to the changing trends of steel reinforcement material properties.

[0277] 6. Continuous Model Optimization

[0278] By continuously incorporating new test data and updating the model, the steel reinforcement mechanical property prediction model can gradually accumulate more data features, thereby improving the model's predictive ability. Simultaneously, the model update mechanism can adapt to changes in the material properties of steel reinforcement from different production batches, giving the model better generalization ability and stability.

[0279] 7. Technical Functions of This Section

[0280] Through model update and performance optimization mechanisms, the following can be achieved:

[0281] ① The model continuously learns from new rebar inspection data;

[0282] ② Improve the accuracy of predicting the mechanical properties of reinforcing bars;

[0283] ③ Improve the model's adaptability to different batches of steel reinforcement materials;

[0284] ④ Ensure the long-term stability and reliability of the rebar detection system.

[0285] The working principle of this application is as follows:

[0286] First, at the data level, multi-source information of steel bar samples is obtained through the steel bar testing data acquisition steps, including basic information of steel bars, geometric parameter data, mechanical performance test data and test environment data. Data association and structured storage are achieved through a unified numbering mechanism, thereby constructing a steel bar testing dataset.

[0287] Secondly, at the data representation level, through the steel bar feature data construction step, the original detection data is transformed into a unified feature vector representation, and data of different dimensions are normalized. At the same time, statistical features are extracted to form a multidimensional steel bar feature matrix. Meanwhile, through the steel bar mechanical property label construction step, indicators such as yield strength, tensile strength and elongation after fracture obtained from tensile tests are constructed into label vectors, thereby forming the input data and target data required for model training.

[0288] Then, at the model level, a large model for predicting the mechanical properties of steel bars is constructed. A deep neural network is used to perform nonlinear mapping on the steel bar feature data to establish the mapping relationship between steel bar features and mechanical properties. The model is trained using historical detection data and the model parameters are continuously optimized with the goal of minimizing the prediction error, thereby achieving accurate prediction of the mechanical performance indicators of steel bars.

[0289] After obtaining the predicted results of the mechanical properties of steel bars, at the analytical level, through the evaluation of the variability of the mechanical properties of steel bars, the predicted results are statistically analyzed, and the batch average, standard deviation and coefficient of variation are calculated, thereby characterizing the performance fluctuation of steel bar materials in different samples and batches, and realizing the quantitative evaluation of the stability of steel bar materials.

[0290] Furthermore, at the application level, through the steps of steel bar strength prediction and quality assessment, the predicted steel bar strength index is compared with the standard strength, and a comprehensive quality evaluation index is constructed by combining the variability assessment results, thereby determining whether the steel bar material meets the requirements of engineering application and identifying potential quality risks.

[0291] Finally, at the system optimization level, through model update and performance optimization mechanisms, newly added detection data is continuously introduced into the model training process. Through incremental learning and parameter updates, the model's predictive ability is continuously improved, enabling the model to adapt to the performance changes of steel reinforcement materials in different production batches and achieve long-term stable operation of the system.

[0292] In summary, this application constructs a closed-loop intelligent analysis system for the mechanical properties of reinforcing steel bars through a technical path of "data acquisition—feature construction—model prediction—variability assessment—quality evaluation—model update," thereby achieving efficient prediction and systematic evaluation of the performance of reinforcing steel bars and improving the efficiency of reinforcing steel bar testing data utilization and the scientificity and accuracy of engineering material quality assessment.

[0293] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel bars based on a large model, characterized in that, Includes the following steps: S1. Rebar Testing Data Acquisition: Collect basic information, geometric parameter data, mechanical property test data, and test environment data of rebar samples, and construct a structured rebar testing dataset; S2. Construction of rebar feature data: The rebar detection data is processed by feature extraction and organization to construct a rebar feature vector, and then normalized and statistically extracted to form a rebar feature matrix; S3. Construction of Mechanical Property Labels for Steel Reinforcing Bars: Based on the tensile test results of steel reinforcing bars, the yield strength, tensile strength and elongation after fracture are extracted to construct a label vector and label matrix for the mechanical properties of steel reinforcing bars. S4. Construction of a large-scale model for predicting the mechanical properties of reinforcing bars: The deep learning model is trained using the aforementioned feature matrix and mechanical property label matrix of reinforcing bars to establish the mapping relationship between reinforcing bar features and mechanical properties, and the prediction results of the mechanical properties of reinforcing bars are output. S5. Evaluation of the variability of steel bar mechanical properties: Based on the predicted results of steel bar mechanical properties, calculate the statistical characteristic parameters of steel bar mechanical properties, including the mean, standard deviation and coefficient of variation, to evaluate the fluctuation of steel bar performance; S6. Reinforcing bar strength prediction and quality assessment: The predicted strength of the reinforcing bar is compared with the standard strength, and a comprehensive quality evaluation index is constructed by combining the coefficient of variation to assess the quality of the reinforcing bar material; S7. Model Update and Performance Optimization: Introduce new rebar detection data into the model training process and update model parameters through incremental learning to improve model prediction accuracy and adaptability.

2. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: In the rebar testing data acquisition step, basic rebar information is collected and managed through a laboratory information management system, and a unique testing identification code is generated for each rebar sample to achieve full-process association and traceability of rebar testing data. The basic rebar information includes at least the rebar specification and model, rebar manufacturer, rebar production batch, rebar arrival time, project name, testing commissioning unit, and testing number. Through the aforementioned unique identification code, the basic rebar information is uniformly associated with geometric parameter data, mechanical performance test data, and test environment data to form a complete rebar testing record. This ensures the consistency and traceability of testing data during acquisition, storage, and subsequent analysis, and provides a reliable data foundation for subsequent rebar feature data construction and model training.

3. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: The geometric parameter data acquisition includes obtaining the diameter, length, and surface morphology features of the reinforcing bars. The diameter is determined by taking multiple measurements at different locations along the axial direction of the reinforcing bar and averaging them to reduce the impact of local wear or non-uniformity on the measurement results. The length of the reinforcing bar is obtained using a laser rangefinder or measuring tool. The surface morphology information of the reinforcing bar is obtained by acquiring image data of the reinforcing bar surface through an image acquisition device to characterize the surface corrosion, cracks, or defects. The geometric parameter data is input into the detection system database through automatic acquisition or manual entry and is associated with the reinforcing bar detection number for storage, thereby realizing the structured expression of the geometric feature data of the reinforcing bars and providing basic data support for subsequent feature extraction and model analysis.

4. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: The test environment data includes test temperature, test humidity, test time, and test equipment number. This test environment data is collected and recorded synchronously in the testing database during the steel reinforcement mechanical property test. Since environmental conditions affect the test results of the steel reinforcement mechanical properties, introducing the test environment data as a feature variable can reflect the impact of environmental factors on the test results during subsequent feature construction. The test environment data is stored in association with the steel reinforcement test number and, together with the steel reinforcement geometric parameter data and mechanical property test data, constitutes a complete test data record, thereby improving the accuracy and reliability of the steel reinforcement test data analysis.

5. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: In the rebar feature data construction step, the geometric features, production information features, inspection environment features, and time features in the rebar inspection data are uniformly organized and rebar feature vectors are constructed. During feature processing, the min-max normalization method is used to standardize each feature variable, mapping features of different dimensions to a unified numerical range to eliminate the impact of differences in feature scales on model training. At the same time, statistical feature information is extracted by calculating the statistical mean and standard deviation of the feature data to reflect the overall distribution of the rebar inspection data. Based on this, the feature vectors of all rebar samples are combined to form a rebar feature matrix, which serves as the input data for the subsequent rebar mechanical property prediction model.

6. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: The large-scale model for predicting the mechanical properties of reinforcing bars is a deep neural network model. The model includes an input layer, at least one hidden layer, and an output layer. The input layer receives the feature matrix data of the reinforcing bars. The hidden layer performs multi-layer mapping on the input features through a weight matrix and a nonlinear activation function to extract deep feature information from the reinforcing bar detection data. The output layer outputs the predicted mechanical properties of the reinforcing bars. The model models the complex relationship between the feature data of the reinforcing bars and the mechanical performance indicators through multi-layer nonlinear mapping, thereby improving the model's ability to express and predict multi-source data.

7. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: During the model training process, the mechanical properties label of steel bars is used as the target output. The model parameters are optimized by minimizing the error between the model prediction result and the true label. The error is expressed as mean square error as the loss function, and the model parameters are continuously updated through iterative training so that the model prediction result gradually approaches the true value. During the training process, the model is trained in batches on historical steel bar detection data, enabling it to learn the mapping relationship between steel bar feature data and mechanical properties, thereby improving the model's generalization ability and prediction accuracy.

8. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: In the step of evaluating the variability of the mechanical properties of steel bars, the predicted mechanical properties data of steel bar samples from the same production batch are statistically analyzed to calculate the batch average, standard deviation and coefficient of variation, wherein the coefficient of variation is the ratio of the standard deviation to the average, and is used to characterize the degree of fluctuation in the performance of steel bar materials. By comparing the coefficient of variation with a preset threshold, the stability of the steel reinforcement material performance is determined. When the coefficient of variation exceeds the threshold, it is determined that the batch of steel reinforcement has a risk of performance fluctuation, thereby realizing a quantitative assessment of the performance variability of steel reinforcement materials.

9. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: In the steel reinforcement strength prediction and quality assessment step, the ratio of the predicted yield strength of the steel reinforcement to the standard specified strength is calculated, and a comprehensive quality evaluation index is constructed by combining the steel reinforcement performance variation coefficient. The evaluation index is obtained by weighting the strength index and the stability index. The quality of steel reinforcement samples and batches of steel reinforcement materials is evaluated by this comprehensive index. When the evaluation index is lower than the preset threshold, it is determined that there is a quality risk in the steel reinforcement material, thereby realizing a comprehensive assessment of the quality of steel reinforcement materials.

10. The method for evaluating the variability of mechanical properties and predicting the strength of reinforcing steel based on a large model as described in claim 1, characterized in that: In the model update and performance optimization steps, the newly added rebar detection data is fused with the original training data to construct a new training dataset, and the rebar mechanical property prediction model is updated using incremental training. During the model update process, the model parameters are iteratively optimized using the gradient descent algorithm to reduce prediction errors. By periodically introducing new data for model training, the model can continuously learn the performance change characteristics of different batches of rebar materials, thereby improving the model's prediction accuracy and adaptability to changes in data distribution.

Images