Asphalt mixture performance prediction method and system

Abstract

The invention relates to the technical field of road engineering, in particular to an asphalt mixture performance prediction method and system, and the method comprises the steps: collecting physical and chemical characteristic data of an asphalt mixture, and constructing a multi-source data set; performing time synchronization on the multi-source data, and performing data association on the multi-source data to obtain a unified feature data set; related performance features are extracted from the feature data set, feature optimization is carried out, and an optimized feature set is obtained; analyzing data in the optimized feature set based on a generalized linear model, and establishing an asphalt mixture performance prediction model; and inputting to-be-predicted asphalt mixture data into the asphalt mixture performance prediction model to obtain an asphalt mixture performance prediction result. According to the method, through multi-source data acquisition, optimization feature selection and generalized linear model analysis, the prediction precision and reliability are remarkably improved, the operation process is simplified, the cost is reduced, and real-time performance prediction is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of road engineering, and particularly relates to a method and system for predicting the performance of asphalt mixtures. Background Art

[0002] With the development of modern transportation, asphalt mixtures, as the most commonly used pavement materials in road construction, their performance directly affects the durability and safety of roads. In order to ensure the long-term service performance of roads, it is particularly important to accurately predict the performance of asphalt mixtures.

[0003] Currently, the main methods for predicting the performance of asphalt mixtures include laboratory tests and empirical formula methods. Laboratory test methods measure the high-temperature stability, low-temperature cracking resistance, and fatigue life of asphalt mixtures through rutting tests, bending tests, and four-point bending fatigue tests respectively. Although this method is accurate, the process is complex, time-consuming, and costly, and it is not suitable for large-scale applications; the empirical formula method is based on historical data and empirical formulas for prediction. Although it is simple and fast, its prediction accuracy is limited by the amount of data and the applicable range of the formula, and it cannot fully consider the performance changes under complex working conditions.

[0004] The information disclosed in this background art section is only intended to deepen the understanding of the overall background art of the present disclosure, and should not be regarded as an admission or any form of suggestion that this information constitutes the prior art known to those skilled in the art. Summary of the Invention

[0005] The present invention provides a method and system for predicting the performance of asphalt mixtures, thus effectively solving the problems pointed out in the background art.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is: A method for predicting the performance of asphalt mixtures, comprising: Collecting physical and chemical characteristic data of asphalt mixtures and constructing a multi-source data set; Performing time synchronization on the multi-source data and performing data association on the multi-source data to obtain a unified feature data set; Extracting relevant performance features from the feature data set and performing feature optimization to obtain an optimized feature set; Analyzing the data in the optimized feature set based on a generalized linear model and establishing an asphalt mixture performance prediction model; Inputting the data of the asphalt mixture to be predicted into the asphalt mixture performance prediction model to obtain the asphalt mixture performance prediction result.

[0007] Further, performing time synchronization on the multi-source data includes: Select a data source as the time reference and perform unified time calibration for all data sources; Add accurate timestamps to the data collected by each sensor. If there are deviations in the timestamps of different data sources, use interpolation methods to adjust the data to be aligned under the same time reference; Define a time window for each timestamp, and the data points within the time window are considered synchronized.

[0008] Furthermore, perform data association on the multi-source data, including: Identify and match the corresponding features in different data sources in terms of their relevance in time and space; Use data fusion algorithms to preprocess the features from different data sources, and the preprocessing includes but is not limited to eliminating noise and redundant information.

[0009] Furthermore, extract relevant performance features from the feature data set, including: Define features related to the performance according to the physical and chemical properties of the asphalt mixture; Calculate performance features from the unified feature data set and combine the calculated performance features into a preliminary feature set.

[0010] Furthermore, use the extreme gradient boosting algorithm to evaluate the importance of the extracted features and obtain the performance pre-contribution values of each feature.

[0011] Furthermore, perform feature selection on each feature, including: Standardize all features in the feature data set and the target performance; Determine the regularization parameter through cross-validation and use the LASSO regression model to train the standardized feature data to obtain feature coefficients; According to the training results of the LASSO regression model, select the features with non-zero coefficients, and these features are the optimized key features.

[0012] Furthermore, analyze the data in the optimized feature set based on the generalized linear model and establish an asphalt mixture performance prediction model, including: Standardize the data in the optimized feature set and define the target variable, where the target variable is the performance index of a specific asphalt mixture; Select the link function and error distribution according to the type of the defined target variable; Use the generalized linear model to train the standardized feature data and adopt maximum likelihood estimation to obtain the parameter values with the minimum error between the model prediction value and the actual observed value; The predictive performance of the model is evaluated through k-fold cross-validation, and the mean squared error is calculated as the evaluation metric; According to the cross-validation results, the regularization parameter is adjusted, and L2 regularization is applied to prevent overfitting, and finally a stable performance prediction model for the asphalt mixture is obtained.

[0013] Furthermore, the link function is a linear link function, and the error distribution is a normal distribution.

[0014] An asphalt mixture performance prediction system, the system includes: A multi-source data acquisition module that acquires physical and chemical characteristic data of the asphalt mixture and constructs a multi-source data set; A multi-source data fusion module that synchronizes the multi-source data in time and correlates the multi-source data to obtain a unified feature data set; A performance feature extraction module that extracts relevant performance features from the feature data set and optimizes the features to obtain an optimized feature set; A prediction model construction module that analyzes the data in the optimized feature set based on the generalized linear model and establishes an asphalt mixture performance prediction model; A prediction result output module that inputs the data of the asphalt mixture to be predicted into the asphalt mixture performance prediction model to obtain the asphalt mixture performance prediction result.

[0015] Furthermore, the prediction model construction module includes: A target variable definition unit that standardizes the data in the optimized feature set and defines the target variable, and the target variable is the performance index of a specific asphalt mixture; A model basic setting unit that selects a link function and an error distribution according to the type of the defined target variable; A feature data training unit that trains the standardized feature data using the generalized linear model and adopts maximum likelihood estimation to obtain the parameter value with the minimum error between the model prediction value and the actual observed value; A model performance evaluation unit that evaluates the predictive performance of the model through k-fold cross-validation and calculates the mean squared error as the evaluation metric; A model optimization and adjustment unit that adjusts the regularization parameter according to the cross-validation results and applies L2 regularization to prevent overfitting, and finally obtains a stable performance prediction model for the asphalt mixture.

[0016] Through the technical solution of the present invention, the following technical effects can be achieved: Through multi-source data acquisition, optimized feature selection and generalized linear model analysis, the present invention significantly improves the prediction accuracy and reliability, simplifies the operation process, reduces the cost, and realizes real-time performance prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a schematic flowchart of a method for predicting the performance of asphalt mixture; Figure 2 It is a schematic flowchart of time synchronization for multi-source data; Figure 3 It is a schematic flowchart of data association for multi-source data; Figure 4 It is a schematic flowchart of extracting relevant performance characteristics from the feature data set; Figure 5 It is a schematic flowchart of feature selection for each feature; Figure 6 It is a schematic flowchart of establishing an asphalt mixture performance prediction model. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0021] Embodiment 1 As Figure 1 shown, the present invention provides a method for predicting the performance of asphalt mixture, and the method includes: S1: Collect the physical and chemical characteristic data of the asphalt mixture and construct a multi-source data set; Specifically, physical and chemical property data of asphalt mixtures are collected through various sensors and devices. These data include, but are not limited to, density, asphalt content, aggregate gradation, temperature, and humidity. The selected sensors and devices are density sensors, asphalt content analyzers, aggregate gradation analysis devices, temperature sensors, and humidity sensors. Batch data collection is carried out in the laboratory or on-site. Data of multiple samples are collected each time, and a data logger is used to record the data of each sensor in real time, including the timestamp and measurement value. All the collected data will be stored in a database, and each record contains a timestamp, sensor ID, and measurement value.

[0022] To ensure the accuracy and reliability of the data, the data is first cleaned to remove outliers and noise. Statistical methods (such as the 3σ rule) are used to identify and delete obvious abnormal data points. Then, data at different scales are normalized and converted to the same scale range (usually [0, 1] or [-1, 1]) for subsequent feature extraction and modeling. Subsequently, data from different sensors and devices are fused to generate a data set containing all physical and chemical properties.

[0023] S2: Perform time synchronization on the multi-source data and perform data association on the multi-source data to obtain a unified feature data set; Specifically, by performing time synchronization and data association on the multi-source data, it is ensured that the data collected by different sensors are aligned under the same time reference, the inconsistency problem caused by time deviation is eliminated, and data from different sources are integrated through feature matching and data fusion technologies to remove noise and redundant information, thereby generating a complete and unified feature data set. This process significantly improves the accuracy and reliability of the data, ensuring the consistency and high quality of the basic data for subsequent analysis and modeling.

[0024] S3: Extract relevant performance features from the feature data set and perform feature optimization to obtain an optimized feature set; Specifically, key features related to the performance of asphalt mixtures are extracted from the feature data set, and feature optimization techniques are used to select and optimize these features to remove redundant information and noise, thereby obtaining a concise and effective optimized feature set. The purpose of this step is to improve the data representation ability and the prediction performance of the model, ensure that only the most representative features are used, enhance the generalization ability and robustness of the model, and finally construct a more accurate and efficient asphalt mixture performance prediction model.

[0025] S4: Analyze the data in the optimized feature set based on the generalized linear model and establish an asphalt mixture performance prediction model; Specifically, by using the Generalized Linear Model (GLM) to analyze and model the data in the optimized feature set, the relationship between the features and the performance indicators of asphalt mixtures can be effectively captured, and an accurate performance prediction model can be established. By leveraging the flexibility and strong interpretability of GLM, various types of data distributions and relationships can be processed to generate a robust and efficient prediction model. This not only improves the prediction accuracy and reliability of the model but also makes the prediction results more interpretable and practical.

[0026] S5: Input the asphalt mixture data to be predicted into the asphalt mixture performance prediction model to obtain the asphalt mixture performance prediction result.

[0027] Through multi-source data collection, optimized feature selection, and generalized linear model analysis, the present invention significantly improves the prediction accuracy and reliability, simplifies the operation process, reduces costs, and realizes real-time performance prediction.

[0028] As a preference of the above embodiment, time synchronization of multi-source data includes: A10: Select one data source as the time reference and perform unified time calibration on all data sources; A20: Add accurate timestamps to the data collected by each sensor. If there are deviations in the timestamps of different data sources, use interpolation methods to adjust the data to be aligned under the same time reference; A30: Define a time window for each timestamp, and the data points within the time window are regarded as synchronized.

[0029] Specifically, first, select a reliable and stable data source from multiple data sources as the time reference. For example, the data source of the density sensor can be selected, and it is calibrated with a unified time using an external synchronous clock (such as GPS time) to ensure the accuracy and consistency of its timestamps. In this way, the data of the density sensor becomes the reference for all other data sources. Next, add accurate timestamps to the data collected by each sensor to ensure that each piece of data records the corresponding collection time. To ensure the consistency of timestamps of different data sources, check whether there are deviations in the timestamps of each data source. If deviations are found between timestamps, interpolation methods are used to adjust the data to align it under the same time reference. For example, linear interpolation methods can be used. The specific steps include: assuming that the data points between two timestamps are not under the same time reference, divide their time difference into several equally spaced small segments, perform linear interpolation calculations on the data points within these small segments, and adjust the timestamps of each data point to align it to the same time reference. For example, if the timestamp of the asphalt content analyzer is 2 milliseconds slower than that of the density sensor, divide the timestamps of the data points of the asphalt content analyzer into several equally spaced small segments and perform linear interpolation calculations to align it to the time reference of the density sensor. Finally, define a reasonable time window for each timestamp to ensure that the data points within this time window are considered synchronous. The size of the time window should be determined according to the actual application requirements and data collection frequency. A preferred setting is to define the time window as ±5 milliseconds, which means that the data points within this time window are considered synchronous. For example, if the timestamp recorded by the density sensor is 1000 milliseconds, the data points between 995 and 1005 milliseconds are all considered synchronous data. This time window setting ensures the synchrony and validity of the data, thereby improving the accuracy of multi-source data fusion. Through the above steps, time synchronization of multi-source data is performed to ensure that different data sources are aligned under the same time reference and are considered synchronous data within a reasonable time window. This process lays the foundation for subsequent data association and fusion, ensuring the accuracy and consistency of the data and providing high-quality data input for the performance prediction model.

[0030] Preferably, as in the above embodiment, data association of multi-source data includes: B10: Identify and match the corresponding features in different data sources for their relevance in time and space; B20: Use data fusion algorithms to preprocess the features from different data sources. The preprocessing includes, but is not limited to, eliminating noise and redundant information.

[0031] Specifically, by analyzing the features provided by each data source, such as density, asphalt content, aggregate gradation, temperature, and humidity, find their consistency in time and space. For each data source, determine the relationship between its features and those of other data sources, and align them through timestamps to ensure that all features are compared and processed at the same time point. For example, by comparing the timestamps of the density sensor and the temperature sensor, ensure that the data collected within the same time period can be directly correlated and compared. Data fusion algorithms can use techniques such as Kalman filtering and Bayesian estimation. First, apply Kalman filtering to preliminarily fuse multi-source data and eliminate random noise. Kalman filtering is a recursive algorithm that obtains a more accurate estimated value through the weighted average of measured values and predicted values. For example, for density and temperature data, Kalman filtering can effectively smooth the data and reduce fluctuations caused by measurement errors. Subsequently, apply the Bayesian estimation method to further correlate and process the multi-source data. Bayesian estimation generates a posterior distribution by combining prior knowledge and observed data, thereby more accurately estimating unknown parameters. For example, for asphalt content and aggregate gradation data, Bayesian estimation can correct and optimize the observed data by considering prior knowledge (such as the historical data distribution) to obtain more accurate feature values.

[0032] As a preference of the above embodiment, extract relevant performance features from the feature data set, including: C10: Define features related to the performance according to the physical and chemical properties of the asphalt mixture; C20: Calculate performance features from the unified feature data set and combine the calculated performance features into a preliminary feature set.

[0033] Specifically, the key performance indicators of asphalt mixtures include high-temperature stability, low-temperature crack resistance, and fatigue life, etc. These performance indicators are affected by multiple physical and chemical characteristic parameters. Common related characteristics include but are not limited to: density, asphalt content, aggregate gradation, temperature, humidity, viscosity, and tensile strength. Next, the collected and cleaned data contains various characteristic information after multi-source data fusion. For each sample, according to the defined performance-related characteristics, the corresponding characteristic values are extracted from the data set. For example, the density, asphalt content, and temperature values of each sample are extracted from the data set. Then, according to these characteristic values, the specific performance characteristics of each sample are calculated. For example, the stability index of each sample under high-temperature conditions or the crack resistance index under low-temperature conditions can be calculated. These calculated performance characteristics are combined into a preliminary characteristic set, and each characteristic set contains all relevant performance characteristics. Through the above steps, the key characteristics related to the performance of asphalt mixtures are defined and extracted to ensure that all calculated performance characteristics are accurately and completely reflected in the preliminary characteristic set. This step provides a detailed and comprehensive characteristic data basis for subsequent feature optimization and model training, ensuring that the model can accurately predict various performance indicators of asphalt mixtures.

[0034] As a preference of the above embodiment, the extreme gradient boosting algorithm is used to evaluate the importance of the extracted features, and the performance pre-contribution values of each feature are obtained.

[0035] Specifically, first, organize the preliminary feature set and the corresponding target performance indicators (such as high-temperature stability, low-temperature crack resistance performance, and fatigue life) into a training data set, ensuring that each sample contains all the extracted features and their corresponding performance indicators. During this process, standardize and preprocess the data to ensure the consistency and quality of the input data and reduce the impact of outliers on model training. Then, configure the parameters of the Extreme Gradient Boosting (XGBoost) algorithm, including the learning rate, maximum depth, and number of trees, etc. The learning rate determines the magnitude of each step update, the maximum depth controls the complexity of each tree, and the number of trees determines the complexity of the overall model. Next, use the preliminary feature set and target performance indicators to train the XGBoost model. Build a decision tree model through iterative optimization and step-by-step improvement. The model gradually adds new trees by minimizing the loss function, thereby enhancing the prediction ability of the overall model. During the training process, the model continuously adjusts the structure and weights of each tree to improve the prediction accuracy of the target performance indicators. After training, utilize the feature importance evaluation function of the XGBoost model to calculate the contribution value of each feature to the model's prediction performance. The importance of a feature is measured by its frequency of occurrence in the decision tree and its contribution to the information gain of the model. Specifically, the model records the number of times each feature appears in all decision trees and the information gain brought each time it appears, and comprehensively calculates the importance score of each feature based on this information. Finally, extract the importance scores of each feature to generate a sorted list of feature importance, showing the contribution values of each feature in performance prediction.

[0036] As a preference of the above embodiment, feature selection is performed on each feature, including: D10: Standardize all features and target performance in the feature data set; D20: Determine the regularization parameter through cross-validation and use the LASSO regression model to train the standardized feature data to obtain feature coefficients; D30: According to the training results of the LASSO regression model, select the features with non-zero coefficients, and the features are the optimized key features.

[0037] Specifically, by calculating the mean and standard deviation of each feature, each feature value is transformed into a standard normal distribution, which can eliminate the magnitude differences between features and ensure the comparability of weight calculations for each feature during model training. For example, for the feature density, calculate its mean and standard deviation, and then perform standardization processing to convert the original data into standardized data. Next, perform k-fold cross-validation, divide the dataset into k subsets, where each subset is used as the validation set in turn, and the remaining subsets are used as the training set. Repeatedly perform model training and validation to determine the optimal regularization parameter 𝜆. LASSO regression constrains feature selection and model parameters by adding an L1 regularization term to the loss function, thereby reducing the influence of unimportant features. During the training process, LASSO regression calculates the coefficients of each feature, and through the regularization term, makes the coefficients of some features approach zero. Finally, after training is completed, extract the coefficients of all features in the LASSO regression model, and screen out the features with non-zero coefficients. These features are considered to be the key features that have a significant impact on the target performance. Combine these key features to form an optimized feature set, providing high-quality input data for subsequent model training and performance prediction.

[0038] As a preference of the above embodiment, analyze the data in the optimized feature set based on the generalized linear model and establish an asphalt mixture performance prediction model, including: S41: Perform standardization processing on the data in the optimized feature set, and define the target variable, where the target variable is the performance index of a specific asphalt mixture; S42: Select the link function and error distribution according to the type of the defined target variable; S43: Use the generalized linear model to train the standardized feature data, and adopt maximum likelihood estimation to obtain the parameter values with the minimum error between the model prediction value and the actual observed value; S44: Evaluate the prediction performance of the model through k-fold cross-validation, and calculate the mean square error as the evaluation index; S45: According to the cross-validation results, adjust the regularization parameter, and apply L2 regularization to prevent overfitting, and finally obtain a stable asphalt mixture performance prediction model.

[0039] Specifically, after standardization, the target variable is defined as specific asphalt mixture performance indicators, such as high-temperature stability, low-temperature crack resistance, or fatigue life. These indicators will serve as the prediction targets of the model. The Generalized Linear Model (GLM) associates the linear predictor with the expected value of the target variable through a link function and assumes that the target variable follows a specific distribution. The Generalized Linear Model (GLM) associates the linear predictor with the expected value of the target variable through a link function and assumes that the target variable follows a specific distribution. During the training process, the best model parameters 𝛽 are found by minimizing the log-likelihood function. Specifically, the general form of GLM is , where 𝜇 is the expected value of the target variable, 𝑔 is the link function, 𝛽 is the model parameter, and the maximum likelihood estimation finds the parameter value that maximizes the log-likelihood function through an iterative optimization algorithm such as gradient descent. Cross-validation divides the dataset into k subsets, and each subset is used as the validation set in turn, while the other subsets are used as the training set for model training and validation. The mean squared error of each validation is calculated, and the average value is taken as the prediction error of the model. Finally, L2 regularization adds an L2 norm penalty term of the parameter to the model loss function to control the model complexity and avoid overfitting. The size of the regularization parameter 𝜆 is adjusted, and the best value is selected through cross-validation, so that the model has strong generalization ability and stability while maintaining good prediction performance. Through the above steps, based on the Generalized Linear Model, the data in the optimized feature set is analyzed, and an accurate and stable asphalt mixture performance prediction model is established, providing scientific performance prediction and decision-making support for road engineering.

[0040] Preferably, as in the above embodiment, the link function is a linear link function, and the error distribution is a normal distribution.

[0041] Specifically, the linear link function and the normal distribution are selected because the performance indicators of asphalt mixtures are usually continuous variables, and their distribution characteristics can be well described by the normal distribution. The linear link function is simple and direct, and can map the linear combination of feature variables to the expected value of the target variable, simplifying the calculation and interpretation of the model. The normal distribution assumption is applicable to most engineering application scenarios, can effectively handle errors and noises, and provides stable and reliable model prediction results. Therefore, this selection can improve the accuracy, interpretability, and practicality of the model.

[0042] Embodiment 2 Based on the same inventive concept as a method for predicting asphalt mixture performance in the foregoing embodiment, the present invention also provides an asphalt mixture performance prediction system, which includes: A multi-source data acquisition module that acquires physical and chemical characteristic data of asphalt mixtures and constructs a multi-source data set; The multi-source data fusion module synchronizes the time of multi-source data and correlates the multi-source data to obtain a unified set of feature data; The performance feature extraction module extracts relevant performance features from the set of feature data and optimizes the features to obtain an optimized set of features; The prediction model construction module analyzes the data in the optimized set of features based on the generalized linear model and establishes an asphalt mixture performance prediction model; The prediction result output module inputs the data of the asphalt mixture to be predicted into the asphalt mixture performance prediction model to obtain the asphalt mixture performance prediction result.

[0043] The above prediction system in the present invention can effectively implement the asphalt mixture performance prediction method, and the technical effects that can be achieved are as described in the above embodiments, which will not be elaborated here.

[0044] As a preference of the above embodiment, the prediction model construction module includes: The target variable definition unit standardizes the data in the optimized set of features and defines the target variable, where the target variable is a specific performance index of the asphalt mixture; The model basic setting unit selects the link function and error distribution according to the type of the defined target variable; The feature data training unit trains the standardized feature data using the generalized linear model and adopts maximum likelihood estimation to obtain the parameter values with the minimum error between the model prediction value and the actual observed value; The model performance evaluation unit evaluates the prediction performance of the model through k-fold cross-validation and calculates the mean square error as the evaluation index; The model optimization and adjustment unit adjusts the regularization parameter according to the cross-validation result and applies L2 regularization to prevent overfitting, and finally obtains a stable asphalt mixture performance prediction model.

[0045] Similarly, for the above optimization solutions of the system, the corresponding optimization effects of the methods in the first embodiment can also be respectively achieved, which will not be elaborated here either.

[0046] Although the present application has been described in combination with specific features and their embodiments, obviously, various modifications and combinations can be made without departing from the spirit and scope of the present application. Accordingly, the present specification and the drawings are merely exemplary descriptions of the present application defined by the appended claims, and are considered to have covered any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A method for predicting asphalt mixture performance, characterized in that: include: Collect data on the physical and chemical properties of asphalt mixtures and build multi-source data sets; Time synchronization is performed on multi-source data, and data association is performed on the multi-source data to obtain a unified feature data set; Extracting relevant performance features from the feature data set, and performing feature optimization to obtain an optimized feature set; Analyzing the data in the optimized feature set based on a generalized linear model and establishing an asphalt mixture performance prediction model; The asphalt mixture data to be predicted is input into the asphalt mixture performance prediction model to obtain the asphalt mixture performance prediction result.

2. The asphalt mixture performance prediction method according to claim 1, characterized in that: Time synchronization of multi-source data, including: Select one data source as the time base and perform unified time calibration on all data sources; Add accurate timestamps to the data collected by each sensor. If there is a deviation in the timestamps of different data sources, use interpolation methods to align the data under the same time base. A time window is defined for each timestamp, within which data points are considered synchronized.

3. The asphalt mixture performance prediction method according to claim 1, characterized in that: Performing data association on the multi-source data includes: Identify and match corresponding features in different data sources, and their correlation in time and space; The features from different data sources are preprocessed using a data fusion algorithm, and the preprocessing includes but is not limited to eliminating noise and redundant information.

4. The asphalt mixture performance prediction method according to claim 1, characterized in that: Extracting relevant performance features from the feature data set includes: Define the characteristics related to the performance of asphalt mixtures based on their physical and chemical properties; Performance features are calculated from the unified feature data set, and the calculated performance features are combined into a preliminary feature set.

5. The asphalt mixture performance prediction method according to claim 1, characterized in that: The extreme gradient boosting algorithm is used to evaluate the importance of the extracted features and obtain the performance pre-contribution value of each feature.

6. The asphalt mixture performance prediction method according to claim 1, characterized in that: Perform feature selection on each feature, including: Standardizing all features and target performance in the feature data set; The regularization parameter is determined by cross-validation, and the LASSO regression model is used to train the standardized feature data to obtain the feature coefficient; According to the training result of the LASSO regression model, a feature whose coefficient is not zero is selected, and the feature is the optimized key feature.

7. The asphalt mixture performance prediction method according to claim 1, characterized in that: The data in the optimized feature set are analyzed based on a generalized linear model, and an asphalt mixture performance prediction model is established, including: Standardizing the data in the optimization feature set and defining a target variable, wherein the target variable is a specific performance index of the asphalt mixture; Select the link function and error distribution according to the type of target variable defined; The standardized feature data are trained using a generalized linear model, and the maximum likelihood estimation is used to obtain the parameter value with the minimum error between the model prediction value and the actual observation value; The prediction performance of the model was evaluated by k-fold cross validation, and the mean square error was calculated as the evaluation indicator; According to the cross-validation results, the regularization parameters are adjusted, and L2 regularization is applied to prevent overfitting, and finally a stable asphalt mixture performance prediction model is obtained.

8. The asphalt mixture performance prediction method according to claim 7, characterized in that: The link function is a linear link function, and the error distribution is a normal distribution.

9. An asphalt mixture performance prediction system, characterized in that: The system comprises: Multivariate data collection module collects data on the physical and chemical properties of asphalt mixtures and builds a multi-source data set; A multi-source data fusion module performs time synchronization on multi-source data and performs data association on the multi-source data to obtain a unified feature data set; A performance feature extraction module extracts relevant performance features from the feature data set and performs feature optimization to obtain an optimized feature set; A prediction model building module analyzes the data in the optimization feature set based on a generalized linear model and establishes an asphalt mixture performance prediction model; The prediction result output module inputs the asphalt mixture data to be predicted into the asphalt mixture performance prediction model to obtain the asphalt mixture performance prediction result.

10. The asphalt mixture performance prediction system according to claim 9, characterized in that: The prediction model building module includes: A target variable definition unit is used to standardize the data in the optimization feature set and define a target variable, wherein the target variable is a specific performance index of the asphalt mixture; A model basic setting unit selects a link function and an error distribution according to the type of the target variable defined; The feature data training unit uses a generalized linear model to train the standardized feature data and adopts maximum likelihood estimation to obtain the parameter value with the minimum error between the model prediction value and the actual observation value; Model performance evaluation unit, which evaluates the prediction performance of the model through k-fold cross validation and calculates the mean square error as the evaluation index; The model optimization and adjustment unit adjusts the regularization parameters according to the cross-validation results, and applies L2 regularization to prevent overfitting, and finally obtains a stable asphalt mixture performance prediction model.

Images