A method for predicting the adhesion of a modified asphalt to aggregate

Abstract

The application discloses a method for predicting the adhesion performance of modified asphalt and aggregate. First, an interface model of modified asphalt and aggregate with different mixing amounts is constructed by using MterialsStudio three-dimensional modeling software, and molecular dynamics simulation is carried out at different temperatures to calculate the interface adhesion energy. Then, based on the interface adhesion energy simulation data, the sample data is constructed with the asphalt system, aggregate, modifier, modifier mixing amount and temperature as input variables and the interface adhesion energy as the output variable. Then, the sample data is converted and normalized, the correlation coefficient between each input variable is calculated, and the input variables with strong correlation are reduced and updated. Then, the sample data is divided into a training set and a test set, a modified asphalt and aggregate adhesion energy prediction model based on a support vector machine is built by selecting a Gaussian kernel function, the prediction model is trained and parameter optimization is carried out to determine the optimal regression function. Finally, the average absolute error MAE and the determination coefficient R 2 The accuracy of the prediction model is tested.

Description

Technical Field

[0001] This invention relates to the field of materials performance technology, and in particular to a method for predicting the adhesion performance of modified asphalt to aggregates. Background Technology

[0002] Currently, evaluation methods for asphalt-aggregate interfacial adhesion can be divided into two categories: qualitative and quantitative analysis. Qualitative analysis methods mainly include the boiling water method and the water immersion method; quantitative analysis methods mainly include photoelectric colorimetry and solvent elution method.

[0003] 1. The boiling water method is relatively widely used in my country and is suitable for base asphalt, modified asphalt, and asphalt with added anti-stripping agents that has undergone heat treatment. The specific steps are: place the aggregate in an oven at 105℃±5℃, heat the asphalt, immerse the heated aggregate particles in the asphalt for 45 seconds, remove them, and place them on a test rack for 15 minutes. After cooling, immerse the aggregate particles in boiling water, and observe the asphalt detachment after 3 minutes to determine the adhesion between the asphalt and aggregate. However, this method relies on qualitative, subjective judgment of adhesion and lacks quantitative indicators.

[0004] 2. The water immersion method involves immersing pre-coated asphalt aggregate in a constant-temperature water bath at 80°C for 30 minutes, and then judging the percentage of asphalt film peeling off. However, the water immersion method has the drawback of being influenced by subjective factors, which significantly impact the evaluation of the asphalt peeling rate.

[0005] 3. Photoelectric colorimetric method refers to placing a certain amount of asphalt mixture into a phenol-saffron bio-dye of known concentration, soaking it at 60℃ for 2 hours, taking it out and placing it in a test tube, and measuring the absorbance with a spectrophotometer. The residual concentration of the dye is obtained from the relationship curve between concentration and absorbance, thereby calculating the amount of adsorption of the original aggregate and the amount of adsorption after the aggregate is detached, thus obtaining the detachment rate.

[0006] 4. The solvent stripping method involves mixing asphalt with toluene to form a 2% concentration solution. This solution is then tested in a 30°C constant-temperature aggregate test on asphalt of the same volume but different aggregate types. This method measures the adhesion between asphalt and aggregate, and the adhesion is expressed as the amount of asphalt that adheres and the rate of separation.

[0007] Although photoelectric colorimetry and solvent stripping methods can quantitatively determine the stripping rate and use the stripping rate to determine the parameters of the adhesion between asphalt and aggregate, the operation is more complicated, and the results are usually significantly biased because saffron in the solution adheres to the asphalt. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a method for predicting the adhesion performance of modified asphalt to aggregates that is simpler to operate and has better predictive effect.

[0009] To solve the above-mentioned technical problems, the present invention adopts the following technical method: a method for predicting the adhesion performance between modified asphalt and aggregates, comprising:

[0010] Step 1: Use Materials Studio 3D modeling software to construct amorphous unit cell models of various modifiers, molecular monomer structure models of various matrix asphalt, and import unit cell models of various different crystals and construct crystal plane models for them.

[0011] Step 2: Use Materials Studio 3D modeling software to construct amorphous unit cell models of modified asphalt with different dosages, and then perform geometric optimization, relaxation, and annealing to select the amorphous unit cell model of modified asphalt with different dosages that has the lowest energy.

[0012] Step 3: Construct interface models of modified asphalt and aggregate with different dosages using the amorphous cell models of the modified asphalt with the lowest energy obtained in Step 2 and the various crystal plane models obtained in Step 1. Then, perform molecular dynamics simulations at different temperatures to calculate the interfacial adhesion energy.

[0013] Step 4: Based on the interfacial adhesion energy simulation data obtained in Step 3, construct sample data with asphalt system, aggregate, modifier, modifier dosage, and temperature as input variables and interfacial adhesion energy as output variable.

[0014] Step 5: Transform and normalize the sample data obtained in Step 4, then calculate the correlation coefficient between each input variable and perform dimensionality reduction and update on the strongly correlated input variables;

[0015] Step 6: Divide the sample data processed in Step 5 into training set and test set, select Gaussian kernel function to build a modified asphalt and aggregate adhesion energy prediction model based on support vector machine, train the modified asphalt and aggregate adhesion energy prediction model based on training set data, and use genetic algorithm to optimize prediction model parameters and determine the optimal regression function.

[0016] Step 7: Verify the modified asphalt-aggregate adhesion energy prediction model trained in Step 6 based on the test set data, using the mean absolute error (MAE) and coefficient of determination (R²). 2 Test the accuracy of the prediction model.

[0017] Furthermore, in step 1, the matrix asphalt molecular system includes AAA-1, AAK-1, and AAM-1; the modifier includes SBS, PE, biomass oil TG, and rubber powder; and the crystals include CaO, MgO, SiO2, Fe2O3, and Al2O3.

[0018] Furthermore, in step 2, the modifier dosage ranges from 0% to 20%.

[0019] Furthermore, in step 3, when performing molecular dynamics simulations, the temperature range is set to 273K to 373K, the selected parameter is Energy, and the interfacial adhesion energy between modified asphalt and aggregate with different dosages is calculated using the corresponding script and mean value method.

[0020] Furthermore, in step 5, when transforming the sample data: first, the non-numerical variables are converted into numerical variables, and then the numerical variables are segmented and discretized.

[0021] Furthermore, in step 5, the normalization formula is as follows:

[0022]

[0023] In the formula: Z represents the original value of the sample data; Z * Z represents the normalized value of the sample data. max and Z min These are the maximum and minimum values ​​in the sample data, respectively.

[0024] Furthermore, in step 5, the Pearson correlation coefficient is used to reflect the correlation between input variables. If the Pearson correlation coefficient between any two input variables is greater than 0.8, then the two input variables are considered to have a strong correlation. Then, principal component analysis is used to reduce the dimensionality and update the input variables with strong correlation.

[0025] Furthermore, in step 6, let x be the input variable and y be the output variable; the sample data processed in step 5 is divided into a training set at a ratio of 8:2. With test set n is the number of samples in the training set, i = 1, 2, ..., n; x i and y i Let x be the i-th input variable and the i-th output variable in the training set, respectively, and N be the number of samples in the test set, j = 1, 2, ..., N; j and y j Let $j$ be the j-th input variable and $j$ be the output variable in the test set, respectively. Then, the following method is used to build, train, and optimize the prediction model for the adhesion energy of modified asphalt and aggregate:

[0026] 1) The function for constructing a support vector machine is:

[0027]

[0028] 2) The constraints for determining the support vector machine function are:

[0029]

[0030] In the formula: ω is the weight vector; ω Tξ is the transpose of the weight vector ω; C is the penalty factor; i φ is the slack variable; φ() is the support vector machine spatial classification function; b is the displacement term;

[0031] 3) The classification function φ(x) i +b) Map to a high-dimensional space and construct the Gaussian kernel function as follows:

[0032]

[0033] In the formula: K() is the Gaussian kernel function; σ is the smoothness parameter;

[0034] 4) The support vector machine regression function is constructed as follows:

[0035]

[0036] In the formula: f() is the support vector machine regression function; α i , It is a Lagrange multiplier;

[0037] 5) Initialize and encode the parameters C and σ to construct the first generation genetic population. Then, perform crossover and mutation genetic operations on the current generation population to generate the next generation population. Repeat this step to continuously optimize the parameters C and σ of the support vector machine until these two parameters meet the conditions or reach the maximum number of iterations.

[0038] Furthermore, in step 7:

[0039] 1) The formula for calculating the Mean Absolute Error (MAE) is:

[0040]

[0041] In the formula: f(x) j To input variable x in the test set j Substitute the predicted value into the support vector machine regression function;

[0042] 2) Determining coefficient R 2 The calculation formula is:

[0043]

[0044] In the formula: This is the mean of all output variables in the test set.

[0045] To address the problems of subjective judgment, lack of quantitative indicators, significant bias in results, and high cost in traditional evaluation methods for the adhesion performance of asphalt and aggregate, this invention provides a method for predicting the adhesion performance of modified asphalt and aggregate. This method uses Materials Studio 3D modeling software to simulate various modified asphalt-aggregate interface models under different modifier dosages and temperature conditions, thereby simulating the corresponding adhesion energy. Using this as a dataset, and constructing input variables with base asphalt, aggregate, modifier, modifier dosage, and temperature, and the adhesion energy as the output variable, a prediction model for the adhesion performance of modified asphalt and aggregate based on the support vector machine algorithm is established. This method can predict the adhesion energy of different modified asphalt-aggregate interfaces under different modifier dosages and temperature conditions within the relevant range (for example, randomly selecting AAK-1 as the base asphalt, SiO2 as the aggregate, SBS as the modifier, 13% as the modifier dosage, and 323K as the temperature as the input variables can yield the adhesion energy of the corresponding interface). Compared with traditional interfacial adhesion energy testing experiments, it reduces a lot of financial and material waste, is simpler to operate, has reliable quantitative evaluation indicators, and has better prediction results. Attached Figure Description

[0046] Figure 1 This is a flowchart of the method for predicting the adhesion performance between modified asphalt and aggregates involved in this invention;

[0047] Figure 2 This is a schematic diagram of the amorphous unit cell model of the modifier in an embodiment of the present invention (in the figure, (a), (b), (c), and (d) are schematic diagrams of the amorphous unit cell models of rubber powder, SBS, biomass oil TG, and PE, respectively).

[0048] Figure 3 The following is a schematic diagram of the molecular monomer structure model of the matrix asphalt in the embodiments of the present invention (in the figure, (a) is a schematic diagram of the molecular monomer structure model of the three molecular monomers of the asphaltene component; (b) is a schematic diagram of the molecular monomer structure model of the two molecular monomers of the aromatic component; (c) is a schematic diagram of the molecular monomer structure model of the five molecular monomers of the resin component; (d) is a schematic diagram of the molecular monomer structure model of the two molecular monomers of the saturated component).

[0049] Figure 4 The following are schematic diagrams of amorphous unit cell models of modified asphalt with different dosages in the AAA-1 system according to embodiments of the present invention (in the figure, (a) is a schematic diagram of amorphous unit cell model of SBS modified asphalt with 2% dosage; (b) is a schematic diagram of amorphous unit cell model of rubber powder modified asphalt with 2% dosage; (c) is a schematic diagram of amorphous unit cell model of biomass oil TG modified asphalt with 3% dosage; and (d) is a schematic diagram of amorphous unit cell model of PE modified asphalt with 4% dosage).

[0050] Figure 5The following are schematic diagrams of the interface models of modified asphalt and aggregate with different dosages in the embodiments of the present invention (in the figure, (a) is a schematic diagram of the interface model of PE modified asphalt and aggregate; (b) is a schematic diagram of the interface model of SBS modified asphalt and aggregate; (c) is a schematic diagram of the interface model of biomass oil TG modified asphalt and aggregate; (d) is a schematic diagram of the interface model of rubber powder modified asphalt and aggregate).

[0051] Figure 6 This is a flowchart illustrating the optimization of prediction model parameters using a genetic algorithm in this invention;

[0052] Figure 7 This is a comparison chart of prediction results in an embodiment of the present invention. Detailed Implementation

[0053] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0054] like Figure 1 As shown, a method for predicting the adhesion properties of modified asphalt to aggregates includes:

[0055] Step 1: Use Materials Studio 3D modeling software to construct amorphous cell models of various modifiers, molecular monomer structure models of various matrix asphalts, and cell models of various different crystals (i.e. aggregates), and construct crystal plane models for them.

[0056] 101. When constructing the amorphous unit cell model of the modifier:

[0057] First, models for four modifiers—SBS, PE, biomass oil TG, and rubber powder—were constructed separately. Specifically, for the modifier SBS: repeating unit structures of styrene and 1,3-butadiene were constructed to determine the degree of polymerization of styrene-butadiene rubber (SBS), and then a copolymer SBS model was constructed. For PE: repeating unit structures of polyethylene were constructed and their degree of polymerization was determined to obtain the PE model. For biomass oil TG: TG was used to replace biomass oil to construct a molecular model of TG. For rubber powder: SBR was used to replace styrene and 1,3-butadiene, repeating unit structures of styrene and 1,3-butadiene were constructed to determine the degree of polymerization of styrene-butadiene rubber (SBR), and then a copolymer styrene-butadiene rubber (SBR) model was constructed.

[0058] The Forcite module was then used to apply force fields and charges to the four modifiers, and their geometry was optimized, with a maximum iteration count of 5000. During the optimization process, the COMPASS II force field was used to distribute the charge, and the SMART algorithm was employed for energy reduction, resulting in configurations with lower energy, such as... Figure 2 As shown.

[0059] 102. When constructing the molecular monomer structure model of matrix bitumen:

[0060] First, based on the content of each component in each system of the matrix asphalt in Table 1, the single-component structure of each component is constructed.

[0061] Table 1. Component distribution of the three base asphalt systems

[0062]

[0063] The Forcite module is then used to apply force fields and charges to the monolithic structure obtained in the previous step, and geometric optimization is performed on it, with a maximum iteration count of 5000. During the optimization process, the COMPASS II force field is used to distribute the charge, and the SMART algorithm is employed for energy reduction, resulting in a lower-energy configuration, such as... Figure 3 As shown.

[0064] 103. When constructing a crystal plane model:

[0065] First, find the cell structures of CaO, MgO, SiO2, Fe2O3, and Al2O3 in the inorganic crystal software (Findit) and import them into the Materials Studio 3D modeling software.

[0066] Next, the Cleave Surface tool is used to cut the five cell structures obtained in the previous step along the (001)(001)(100)(104)(001) planes, while setting the thickness parameter to 2.

[0067] Then, use the Build Vacuum Slab tool to set the vacuum layer to 10 on the crystal plane obtained in the previous step.

[0068] Finally, the Supercell tool was used to define U=5 and V=5 to expand the cell of the obtained crystal plane.

[0069] Step 2

[0070] First, select any geometrically optimized matrix asphalt molecular monomer structure in any system obtained in step 1 and any geometrically optimized modifier amorphous cell model obtained in step 1 for combination, and control its dosage by adjusting the combination amount of modifier. Specifically, the modifier dosage range is 0% to 20%.

[0071] Then, using the Amorphous Cell module in the Materials Studio 3D modeling software, amorphous unit cell models of modified asphalt with different dosages were constructed. Some schematic diagrams are shown below. Figure 4As shown (the base asphalt is AAA-1 as an example).

[0072] Finally, the amorphous unit cell models of modified asphalt with different dosages were geometrically optimized. The maximum number of iterations was set to 20,000. Then, the configuration with lower energy after geometric optimization was selected for annealing. The NPT ensemble was selected for annealing. The temperature was increased from 300K to 800K and then cooled to room temperature. The annealing cycle was performed for 5 times.

[0073] Step 3

[0074] First, select any amorphous unit cell model of modified asphalt with different admixtures after annealing in step 2, and any crystal plane model obtained in step 1. Using the Build Layer tool, define the crystal plane model as Layer 1 and the amorphous unit cell models of modified asphalt with different admixtures as Layer 2, and set the Vacuum parameter to 30. In the Matching toolbar, select Average to obtain the interface models of modified asphalt and aggregates with different admixtures, such as... Figure 5 As shown.

[0075] Then, dynamic simulations were performed using the Forcite module. Before the simulation, Layer 1 was fixed using the Constraints function in Modify. The simulation time was 200 ps, ​​the NVT ensemble was selected, and different temperatures were set within the range of 273 K to 373 K. The accuracy was set to Medium, and the cutoff radius was [value missing]. The force field remains COMPASS II. The van der Waals nonbonding interaction and electrostatic nonbonding interaction are solved using the atom-based and Ewald methods, respectively. The charge is set to use current, and the step size is 1 fs, resulting in an xtd file containing 201 frames.

[0076] Finally, the script calculates all the adhesion energies of the xtd file obtained in the previous step, and then takes the values ​​at the convergence point to calculate the final adhesion energy of the interface using the mean method. The formula for calculating the interface adhesion energy is:

[0077] E inter =(E p +E q -E p-q (8)

[0078] E p-q It is the energy that Layer 2 and Layer 1 reach in equilibrium after being adhered together; E p It is the energy of Layer 2 when it reaches equilibrium after dynamics; E q It is the energy of Layer 1 in dynamic equilibrium; E interThis represents the interaction energy between the interface of Layer2 and Layer1, i.e., the adhesion energy.

[0079] Step 4: Based on the interfacial adhesion energy simulation data obtained in Step 3, randomly select 50 sets of data (see Table 2 below) and construct sample data with asphalt system, aggregate, modifier, modifier dosage, and temperature as input variables and interfacial adhesion energy as output variable.

[0080] Table 2 Sample Data

[0081]

[0082]

[0083]

[0084] Step 5

[0085] First, the three non-numerical variables—base asphalt system, aggregate, and modifier—in the sample data obtained in step 4 are converted into 0-1 numerical variables using one-hot encoding technology, and finally transformed into sparse matrices as follows:

[0086]

[0087]

[0088]

[0089] The two numerical variables, modifier dosage and temperature, are then discretized into segments: 0–5% is one segment, 5–10% is one segment, 10–15% is one segment, 15–20% is one segment; 273k–298k is one segment, 298k–323k is one segment, 323k–358k is one segment, and 358k–393k is one segment.

[0090] Next, normalization is performed according to the following formula (1):

[0091]

[0092] In the formula: Z represents the original value of the sample data; Z * Z represents the normalized value of the sample data. max and Z min These are the maximum and minimum values ​​in the sample data, respectively.

[0093] Then, the Pearson correlation coefficient is used to reflect the correlation between input variables. If the Pearson correlation coefficient between any two input variables is greater than 0.8, then the two input variables are considered to be strongly correlated.

[0094] Finally, principal component analysis was used to reduce the dimensionality and update the strongly correlated input variables. In Table 3, the Pearson correlation coefficients between different input variables are all less than 0.8, indicating that the different input variables are independent of each other, so no dimensionality reduction was performed.

[0095] Table 3. Pearson correlation coefficient matrix among different input variables

[0096]

[0097] Step 6: Let x be the input variable and y be the output variable; divide the sample data processed in Step 5 into a training set at an 8:2 ratio. With test set n is the number of samples in the training set, i = 1, 2, ..., n; x i and y i Let x be the i-th input variable and the i-th output variable in the training set, respectively, and N be the number of samples in the test set, j = 1, 2, ..., N; j and y j Let $j$ be the j-th input variable and $j$ be the output variable in the test set, respectively. Then, the following method is used to build, train, and optimize the prediction model for the adhesion energy of modified asphalt and aggregate:

[0098] 1) The function for constructing a Support Vector Machine (SVM) is as follows:

[0099]

[0100] 2) The constraints for determining the support vector machine function are:

[0101]

[0102] In the formula: ω is the weight vector; ω T ξ is the transpose of the weight vector ω; C is the penalty factor; i φ is the slack variable; φ() is the spatial classification function of the support vector machine; b is the displacement term.

[0103] 3) The classification function φ(x) i +b) Map to a high-dimensional space and construct the Gaussian kernel function as follows:

[0104]

[0105] In the formula: K() is the Gaussian kernel function; σ is the smoothness parameter.

[0106] 4) The support vector machine regression function is constructed as follows:

[0107]

[0108] In the formula: f() is the support vector machine regression function; α i , It is a Lagrange multiplier.

[0109] 5) The C and σ values ​​in the Gaussian radial basis function are tuned using a genetic algorithm (GA). The tuning process is as follows: Figure 6 As shown in Table 4, the specific settings are initialized and encoded to construct the first generation genetic population. Then, the current generation population is generated by performing crossover and mutation genetic operations on the current generation population operators. This step is repeated to continuously optimize the C and σ of the support vector machine until these two parameters meet the conditions or reach the maximum number of iterations.

[0110] Table 4 Model Settings

[0111]

[0112] The optimal parameters were finally determined to be C = 25.3066 and σ = 0.2610. The prediction was performed using these optimal parameters, and the prediction results are as follows: Figure 7 As shown. Figure 7 In this paper, SVM represents the modified asphalt-aggregate adhesion energy prediction model based on support vector machine without genetic algorithm optimization, and GA-SVM represents the modified asphalt-aggregate adhesion energy prediction model based on support vector machine with genetic algorithm optimization involved in this invention. Figure 7 It can be seen that the predicted values ​​obtained by GA-SVM are closer to the true values ​​than those obtained by SVM.

[0113] Step 7: Verify the modified asphalt-aggregate adhesion energy prediction model trained in Step 6 based on the test set data, using the mean absolute error (MAE) and coefficient of determination (R²). 2 Test the accuracy of the prediction model. Among them:

[0114] 1) The formula for calculating the Mean Absolute Error (MAE) is:

[0115]

[0116] In the formula: f(x) j To input variable x in the test set j Substitute the predicted value into the support vector machine regression function;

[0117] 2) Determining coefficient R 2 The calculation formula is:

[0118]

[0119] In the formula: This is the mean of all output variables in the test set.

[0120] The specific evaluation results of the prediction model are shown in Table 5 below:

[0121] Table 5 Evaluation Results of the Prediction Model

[0122]

[0123] The prediction evaluation results in Table 5 above further show that GA-SVM's predicted values ​​are closer to the true values ​​than SVM's, and GA-SVM's final MAE is 6.8380%, R0. 2 0.9998 (R) 2 The closer the value is to 1, the stronger the explanatory power of the input variable on the output variable, and the better the model fits the data. This well confirms that the interface adhesion energy predicted by GA-SVM fits the true value very well, and the prediction effect of this invention is good.

[0124] The above embodiments are preferred implementations of the present invention. In addition, the present invention can be implemented in other ways. Any obvious substitutions without departing from the concept of the present technical solution are within the protection scope of the present invention.

[0125] To facilitate understanding by those skilled in the art of the improvements of this invention over the prior art, some of the accompanying drawings and descriptions have been simplified, and for clarity, some other elements have been omitted from this application. Those skilled in the art should realize that these omitted elements may also constitute the content of this invention.

Claims

1. A method for predicting the adhesion properties of modified asphalt to aggregates, characterized in that, include: Step 1: Use Material Studio 3D modeling software to construct amorphous unit cell models of various modifiers, molecular monomer structure models of various matrix asphalt, and import unit cell models of various different crystals and construct crystal plane models for them. Step 2: Use Material Studio 3D modeling software to construct amorphous unit cell models of modified asphalt with different dosages, and then perform geometric optimization, relaxation, and annealing to select the amorphous unit cell model of modified asphalt with different dosages that has the lowest energy. Step 3: Construct interface models of modified asphalt and aggregate with different dosages using the amorphous cell models of the modified asphalt with the lowest energy obtained in Step 2 and the various crystal plane models obtained in Step 1. Then, perform molecular dynamics simulations at different temperatures to calculate the interfacial adhesion energy. Step 4: Based on the interfacial adhesion energy simulation data obtained in Step 3, construct sample data with asphalt system, aggregate, modifier, modifier dosage, and temperature as input variables and interfacial adhesion energy as output variable. Step 5: Transform and normalize the sample data obtained in Step 4, then calculate the correlation coefficient between each input variable and perform dimensionality reduction and update on the strongly correlated input variables; Step 6: Divide the sample data processed in Step 5 into training set and test set, select Gaussian kernel function to build a modified asphalt and aggregate adhesion energy prediction model based on support vector machine, train the modified asphalt and aggregate adhesion energy prediction model based on training set data, and use genetic algorithm to optimize prediction model parameters and determine the optimal regression function. Step 7: Verify the modified asphalt-aggregate adhesion energy prediction model trained in Step 6 based on the test set data, using the mean absolute error (MAE) and coefficient of determination (R²). 2 Test the accuracy of the prediction model.

2. The method for predicting the adhesion performance of modified asphalt to aggregates according to claim 1, characterized in that: In step 1, the matrix asphalt molecular system includes AAA-1, AAK-1, and AAM-1; the modifiers include SBS, PE, biomass oil TG, and rubber powder; and the crystals include CaO, MgO, SiO2, Fe2O3, and Al2O3.

3. The method for predicting the adhesion performance of modified asphalt to aggregates according to claim 2, characterized in that: In step 2, the modifier dosage ranges from 0% to 20%.

4. The method for predicting the adhesion performance between modified asphalt and aggregates according to claim 3, characterized in that: In step 3, when performing molecular dynamics simulation, the temperature range is set to 273K to 373K, the selected parameter is Energy, and the interfacial adhesion energy between modified asphalt and aggregate with different dosages is calculated using the corresponding script and mean value method.

5. The method for predicting the adhesion properties of modified asphalt to aggregates according to claim 4, characterized in that: In step 5, when transforming the sample data, the non-numerical variables are first converted into numerical variables, and then the numerical variables are segmented and discretized.

6. The method for predicting the adhesion properties of modified asphalt to aggregates according to claim 4 or 5, characterized in that: In step 5, the normalization formula is as follows: In the formula: Z represents the original value of the sample data; Z * Z represents the standardized value of the sample data. max and Z min These are the maximum and minimum values ​​in the sample data, respectively.

7. The method for predicting the adhesion properties of modified asphalt to aggregates according to claim 6, characterized in that: In step 5, the Pearson correlation coefficient is used to reflect the correlation between input variables. If the Pearson correlation coefficient between any two input variables is greater than 0.8, then the two input variables are considered to have a strong correlation. Then, principal component analysis is used to reduce the dimensionality and update the input variables with strong correlation.

8. The method for predicting the adhesion properties of modified asphalt to aggregates according to claim 7, characterized in that: In step 6, let x be the input variable and y be the output variable; the sample data processed in step 5 is divided into a training set in an 8:2 ratio. With test set n is the number of samples in the training set, i = 1, 2, ..., n; x i and y i Let x be the i-th input variable and the i-th output variable in the training set, respectively, and N be the number of samples in the test set, j = 1, 2, ..., N; j and y j Let j be the j-th input variable and the j-th output variable in the test set, respectively. Then, the following method is used to build, train, and optimize the modified asphalt-aggregate adhesion energy prediction model: 1) The function for constructing a support vector machine is: 2) The constraints for determining the support vector machine function are: In the formula: ω is the weight vector; ω T ξ is the transpose of the weight vector ω; C is the penalty factor; i φ is the slack variable; φ() is the support vector machine spatial classification function; b is the displacement term; 3) The classification function φ(x) i +b) Map to a high-dimensional space and construct the Gaussian kernel function as follows: In the formula: K() is the Gaussian kernel function; σ is the smoothness parameter; 4) The support vector machine regression function is constructed as follows: In the formula: f() is the support vector machine regression function; α i , It is a Lagrange multiplier; 5) Initialize and encode the parameters C and σ to construct the first generation genetic population. Then, perform crossover and mutation genetic operations on the current generation population to generate the next generation population. Repeat this step to continuously optimize the parameters C and σ of the support vector machine until these two parameters meet the conditions or reach the maximum number of iterations.

9. The method for predicting the adhesion properties of modified asphalt to aggregates according to claim 8, characterized in that: In step 7: 1) The formula for calculating the Mean Absolute Error (MAE) is: In the formula: f(x) j To input variable x into the test set j Substitute the predicted value into the support vector machine regression function; 2) Determine the coefficient R 2 The calculation formula is: In the formula: This is the mean of all output variables in the test set.

Images