A Reverse Design Method for Modified Asphalt Based on Ensemble Learning

Abstract

The present invention discloses a reverse design method for modified asphalt based on ensemble learning. The method prepares SBS modified asphalt and rubber powder modified asphalt with different dosages based on a high-speed shearing mechanism; performs frequency scanning on the modified asphalt using a dynamic shear rheometer; establishes a data set with a one-to-one correspondence relationship based on modifier dosage - frequency - temperature - complex modulus - phase angle; divides the data set into a training set and a test set; trains a reverse design model for modified asphalt based on the training set; and evaluates the accuracy of the model based on the test set. This method reveals the correlation between modifier dosage and the performance of modified asphalt, realizes the purpose of quickly obtaining the corresponding modifier dosage based on the performance requirements of modified asphalt under specific test conditions, greatly shortens the time cycle, saves natural resources, and is of great significance for improving the overall intelligent level of asphalt materials.

Description

Technical Field

[0001] The present invention belongs to the field of asphalt materials, and relates to a reverse design method for asphalt materials, in particular to a reverse design method for modified asphalt based on ensemble learning. Background Art

[0002] Asphalt is an important road construction material, and the performance of asphalt directly affects the service life of roads. Adding modifiers to asphalt to prepare modified asphalt is a common method to improve the performance of asphalt. Styrene-butadiene SBS copolymer and rubber powder are the most widely used modifiers at present. Determining the dosage of the modifier is the key to preparing modified asphalt. At present, the method for determining the dosage of the modifier mainly relies on the empirical trial-and-error method, that is, continuously changing the dosage and conducting tests until the performance requirements are met. This method greatly increases the test time and wastes natural resources. If a relationship between the dosage of the modifier and the performance of modified asphalt can be established, it will be an effective way to solve the existing problems and provide strong support for the development of high-performance asphalt materials.

[0003] Machine learning can utilize a large amount of data to automatically discover patterns and rules in the data, thereby achieving autonomous learning and self-adaptive adjustment, and realizing automated decision-making and prediction. At present, machine learning methods have made breakthrough progress in many fields, including speech recognition, image recognition, natural language processing, medical diagnosis, financial prediction, etc. With the continuous development of technology and the continuous optimization of algorithms, people have currently carried out research on the relationship between the dosage of the modifier and the performance of modified asphalt using machine learning methods. However, they are all focused on the research of forward prediction models that predict the performance of modified asphalt from the dosage of the modifier and test conditions, and have not fundamentally changed the various drawbacks brought by the trial-and-error method. Therefore, establishing a reverse design model of modified asphalt that infers the dosage of the modifier from the performance of modified asphalt has become an urgent problem to be solved.

[0004] Ensemble learning is an improved strategy for traditional machine learning. Traditional machine learning methods usually use a single model or algorithm to solve problems, such as logistic regression, decision trees, etc. The basic idea of ensemble learning is to combine multiple different learning algorithms or models into a powerful model, thereby improving the overall performance and better adapting to complex problems in the real world. Random forest is a powerful ensemble learning algorithm based on decision trees. It can directly process high-dimensional data without feature selection and extraction, and has a fast training speed. Therefore, random forest has great potential in establishing a reverse design model of modified asphalt. Summary of the Invention

[0005] Aiming at the problems of long time cycle, large resource waste and high pollution emissions faced by the traditional trial-and-error method for preparing modified asphalt, the present invention provides a reverse design method for modified asphalt based on ensemble learning. This method reveals the correlation between the dosage of modifiers and the properties of modified asphalt, and realizes the purpose of quickly obtaining the corresponding modifier dosage based on the performance requirements of modified asphalt under specific test conditions, greatly shortening the time cycle, saving natural resources, and having great significance for improving the overall intelligent level of asphalt materials.

[0006] The object of the present invention is achieved by the following technical solutions:

[0007] A reverse design method for modified asphalt based on ensemble learning, comprising the following steps:

[0008] Step 1: Select an asphalt as the base asphalt, select a styrene-butadiene SBS copolymer and a rubber powder as modifiers, determine the dosages of the styrene-butadiene SBS copolymer and the rubber powder, and use a high-speed shear mixer to prepare SBS modified asphalt and rubber powder modified asphalt respectively;

[0009] Step 2: Determine different application frequencies and test temperatures, and use a dynamic shear rheometer to perform frequency sweep tests on the modified asphalt to measure the complex modulus and phase angle of the asphalt to describe the viscoelastic properties of the modified asphalt;

[0010] Step 3: Establish a data set with a one-to-one correspondence relationship based on modifier dosage - frequency - temperature - complex modulus - phase angle, remove abnormal data from the data set, and perform mean value processing on the data in the data set to eliminate the influence of dimensions;

[0011] Step 4: Call a data splitting tool to divide the data set established in Step 3 into a training set and a test set. The data in the training set is used to train the model, and the data in the test set is used to test the accuracy of the model;

[0012] Step 5: Select an appropriate number of decision trees to build a reverse design model for modified asphalt based on the random forest algorithm, use frequency - temperature - complex modulus - phase angle as the model input, and modifier dosage as the model output. Train the model based on the training set data divided in Step 4, and optimize the model parameters based on the grid search method and the cross-validation method;

[0013] Step 6: Based on the test set data divided in Step 4, test the reverse design model for modified asphalt established in Step 5, and evaluate the accuracy of the model with the correlation coefficient r.

[0014] Compared with the prior art, the present invention has the following advantages:

[0015] The present invention establishes a reverse design model for modified asphalt, achieving the purpose of directly designing the dosage of modifiers based on the performance requirements of asphalt under specific test conditions. Fundamentally, it solves the blindness of preparing modified asphalt by the traditional trial-and-error method, greatly shortening the test time, saving natural resources, and making contributions to further improving the intelligent level of asphalt materials. Description of the Drawings

[0016] Figure 1 It is a flowchart of the reverse design method for modified asphalt of the present invention;

[0017] Figure 2 It is a verification diagram of SBS modified asphalt for the reverse design method of modified asphalt of the present invention;

[0018] Figure 3 It is a verification diagram of rubber powder modified asphalt for the reverse design method of modified asphalt of the present invention. Detailed Embodiments

[0019] The technical solutions of the present invention will be further described below in conjunction with the drawings, but are not limited thereto. Any modification or equivalent replacement of the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention shall be covered by the protection scope of the present invention.

[0020] The present invention provides a reverse design method for modified asphalt based on ensemble learning. The method prepares SBS modified asphalt and rubber powder modified asphalt with different dosages based on a high-speed shearing mechanism; performs frequency scanning on the modified asphalt using a dynamic shear rheometer; establishes a dataset with a one-to-one correspondence relationship based on modifier dosage - frequency - temperature - complex modulus - phase angle; divides the dataset into a training set and a test set; trains a modified asphalt reverse design model based on the training set; and evaluates the accuracy of the model based on the test set. As Figure 1 described, it specifically includes the following steps:

[0021] Step 1: Select a 70# asphalt as the base asphalt, select a styrene-butadiene SBS copolymer and a 60-mesh rubber powder as modifiers, control the SBS copolymer dosage to be 3%, 5%, 7%, and the rubber powder dosage to be 15%, 20%, 25%, and use a high-speed shearer to prepare SBS modified asphalt and rubber powder modified asphalt respectively.

[0022] In this step, the preparation step of the modified asphalt includes: First, calculate the contents of the modifier, matrix asphalt and stabilizer, where the stabilizer is sulfur and the dosage is 0.12‰; Second, heat the asphalt to 180-200°C, put in the modifier and shear at 500-1000 r / min for 5-15 minutes; Third, shear at a speed of 2000-4000 r / min for 30-90 min, and add the stabilizer and shear at 500-1000 r / min for 10-20 min; Finally, place the modified asphalt in an oven at 150-180°C and keep it warm for 1-3 hours.

[0023] Step 2: Use a dynamic shear rheometer to perform a frequency sweep test on the modified asphalt to measure the complex modulus and phase angle of the asphalt to describe the viscoelastic properties of the modified asphalt. The applied frequencies are 60 Hz, 45 Hz, 30 Hz, 10 Hz, 5 Hz, 1 Hz, 0.5 Hz, 0.1 Hz, 0.05 Hz, 0.01 Hz, and the test temperatures are 10°C, 20°C, 30°C, 40°C, 50°C, 60°C.

[0024] Step 3: Establish a data set with a one-to-one correspondence relationship based on modifier dosage - frequency - temperature - complex modulus - phase angle, remove abnormal data from the data set, and perform mean processing on the data in the data set to eliminate the influence of dimensions.

[0025] In this step, the formula for mean processing is:

[0026]

[0027] where x i ' is the value after removing the dimension of the data, x i is the original numerical value of the data, is the mean value of the data.

[0028] Step 4: Call the data splitting tool to divide the data set established in Step 3 into a training set and a test set according to a ratio of 9:1. The data in the training set is used to train the model, and the data in the test set is used to test the accuracy of the model.

[0029] Step 5: Select an appropriate number of decision trees to build a reverse design model of modified asphalt based on the random forest algorithm. Use frequency - temperature - complex modulus - phase angle as the model input and modifier dosage as the model output. Train the model based on the training set data divided in Step 4, and optimize the model parameters based on the grid search method and cross - validation method.

[0030] In this step, the random forest constructs multiple decision trees by randomly selecting subsets of data and subsets of features, and then combines the results of these decision trees through voting to obtain the final prediction result.

[0031] Step 6: Based on the test set data divided in Step 4, the reverse design model of modified asphalt established in Step 5 is tested, and the accuracy of the model is evaluated by the correlation coefficient r.

[0032] In this step, the formula for the correlation coefficient r is:

[0033]

[0034] where x is the original value of the true data, is the average value of the true data, y is the original value of the predicted data, is the average value of the predicted data. The value of the correlation coefficient ranges from -1 to 1. -1 indicates a negative correlation between the two variables, 0 indicates no correlation between the two variables, and 1 indicates a positive correlation between the two variables.

[0035] Example:

[0036] This example provides a reverse design method for modified asphalt based on ensemble learning. The specific operation process is as follows:

[0037] Step 1: Select a 70# asphalt as the base asphalt, and select a styrene-butadiene SBS copolymer and a 60-mesh rubber powder as modifiers; control the SBS copolymer content to be 3%, 5%, 7%, and the rubber powder content to be 15%, 20%, 25%, and use a high-speed shear mixer to prepare SBS modified asphalt and rubber powder modified asphalt respectively. The preparation steps of the modified asphalt are as follows: First, calculate the contents of the modifier, base asphalt and stabilizer. The stabilizer is sulfur, and the content is 0.12‰; Second, heat the asphalt to 180°C, put in the modifier and shear at 500 r / min for 10 minutes; Third, shear at a speed of 3000 r / min for 60 min, and add the stabilizer and shear at 500 r / min for 15 min; Finally, keep the modified asphalt in an oven at 160°C for 2 hours.

[0038] Step 2: Use a dynamic shear rheometer to perform frequency sweep tests on the modified asphalt to measure the complex modulus and phase angle of the asphalt to describe the viscoelastic properties of the modified asphalt. The applied frequencies are 60 Hz, 45 Hz, 30 Hz, 10 Hz, 5 Hz, 1 Hz, 0.5 Hz, 0.1 Hz, 0.05 Hz, 0.01 Hz, and the test temperatures are 10°C, 20°C, 30°C, 40°C, 50°C, 60°C, that is, each asphalt at each content corresponds to 60 data points.

[0039] Step 3: Based on the data set established with a one-to-one correspondence between modifier content - frequency - temperature - complex modulus - phase angle, remove the abnormal data in the data set, and perform mean value processing on the data in the data set to eliminate the influence of dimensions.

[0040] Step 4: Call the data splitting tool to divide the dataset established in Step 3 into a training set and a test set at a ratio of 9:1. The training set data is used to train the model, and the test set data is used to test the accuracy of the model.

[0041] Step 5: Select an appropriate number of decision trees to build a modified asphalt reverse design model based on the random forest algorithm. Use frequency-temperature-complex modulus-phase angle as the model input and modifier dosage as the model output. Train the model based on the training set data divided in Step 4, and optimize the model parameters based on the grid search method and cross-validation method.

[0042] Step 6: Based on the test set data divided in Step 4, test the modified asphalt reverse design model established in Step 5. The results are shown in Figure 2 and Figure 3 respectively. Calculate the correlation coefficients to be 0.93 and 0.92 respectively, indicating that there is an obvious correlation between the experimental values and the model design values, and also proving that the modified asphalt reverse design model established based on ensemble learning has a good application effect.

[0043] In summary, the modified asphalt reverse design method based on the present invention clarifies the correlation between the modifier dosage and the properties of modified asphalt, establishes a reverse design model for modified asphalt. At the same time, the correlation coefficients between the true values and the model design values of SBS modified asphalt and rubber powder modified asphalt are both above 0.9, indicating that the model has a good application effect. The present invention provides a new means for the preparation of modified asphalt and makes a contribution to further improving the intelligent level of asphalt materials.

Claims

1. A modified asphalt reverse design method based on ensemble learning, characterized in that The method comprises the following steps: Step 1: Select an asphalt as a base asphalt, select a styrene-butadiene SBS copolymer and a rubber powder as a modifier, determine the dosage of the styrene-butadiene SBS copolymer and the dosage of the rubber powder, and use a high-speed shearing machine to prepare SBS modified asphalt and rubber powder modified asphalt respectively. The preparation steps of the modified asphalt are as follows: first, calculate the content of the modifier, the base asphalt and the stabilizer; second, heat the asphalt to 180-200°C, add the modifier and shear at 500-1000r / min for 5-15 minutes; third, shear at a speed of 2000-4000r / min for 30-90min, and add the stabilizer and shear at 500-1000r / min for 10-20min; finally, place the modified asphalt in a 150-180°C oven for 1-3 hours; Step 2: Determine different applied frequencies and test temperatures, and use a dynamic shear rheometer to perform frequency scanning on the modified asphalt to test the complex modulus and phase angle of the asphalt to describe the viscoelastic properties of the modified asphalt; Step 3: Establish a one-to-one correspondence data set based on modifier dosage-frequency-temperature-complex modulus-phase angle, remove abnormal data in the data set, and average the data set data to eliminate the dimension effect; Step 4: Call the data segmentation tool to divide the data set created in step 3 into a training set and a test set. The training set data is used to train the model, and the test set data is used to test the accuracy of the model. Step 5: Select an appropriate number of decision trees to build a modified asphalt reverse design model based on the random forest algorithm, with frequency-temperature-complex modulus-phase angle as the model input and modifier dosage as the model output. Train the model based on the training set data divided in step 4, and optimize the model parameters based on the grid search method and cross-validation method; Step 6: Based on the test set data divided in step 4, the modified asphalt reverse design model established in step 5 is tested, and the accuracy of the model is evaluated by the correlation coefficient r.

2. The modified asphalt reverse design method based on ensemble learning according to claim 1 is characterized in that The stabilizer is sulfur.

3. The modified asphalt reverse design method based on ensemble learning according to claim 1 is characterized in that In the step 2, the averaging formula is: Among them, x i ' is the value after de-dimensionalization of the data, x i is the original value of the data, is the data mean.

4. The modified asphalt reverse design method based on ensemble learning according to claim 1 is characterized in that In the step 5, the random forest constructs multiple decision trees by randomly selecting a subset of data and a subset of features, and then combines the results of these decision trees by voting to obtain the final prediction result.

5. The modified asphalt reverse design method based on ensemble learning according to claim 1 is characterized in that In step 6, the formula of the correlation coefficient r is: Among them, x is the original value of the real data, is the average value of the real data, y is the original value of the predicted data, is the average value of the predicted data; the correlation coefficient ranges from -1 to 1, where -1 means that the two variables are negatively correlated, 0 means that the two variables are not correlated, and 1 means that the two variables are positively correlated.

Images