CN117408095B - A method for predicting the fatigue life of asphalt at different temperatures - Google Patents

Abstract

The invention provides a method for predicting the fatigue life of asphalt at different temperatures, and relates to the technical field of asphalt testing. It includes: setting the test temperature, obtaining the actual measured data of the composite modulus and the measured fatigue life of the asphalt; establishing a prediction model, and the prediction model includes a model Parameters, divide the model parameters into two categories, namely first model parameters and second model parameters; iteratively optimize the first model parameters according to the first optimization method to obtain the optimized first model parameters; iteratively optimize according to the second optimization method The second model parameters are used to obtain the optimized second model parameters; the prediction model is updated using the optimized first model parameters and the optimized second model parameters to obtain the final prediction model, and the final prediction model is used to obtain the prediction of fatigue life. result. The prediction model of the present invention is simple in form and more practical. The model parameters are optimized in batches through two optimization methods, which can increase the accuracy of model prediction.

Description

A method for predicting the fatigue life of asphalt at different temperatures

Technical field

The invention relates to the technical field of asphalt testing, and specifically relates to a method for predicting the fatigue life of asphalt at different temperatures.

Background technique

Fatigue life is an important indicator for measuring the durability of asphalt asphalt pavement. By accurately predicting the fatigue life of asphalt, the selection and design of pavement materials, construction technology, and subsequent pavement maintenance and management can be optimized, thereby providing scientific decision-making support.

In this context, the development and application of asphalt fatigue life prediction models are particularly important. The asphalt fatigue life prediction model can guide engineers to conduct more scientific pavement design and material selection, thereby improving the durability and economic benefits of the pavement structure. At the same time, with the help of the fatigue life prediction model, engineers can achieve effective maintenance timing judgment and life prediction to ensure road safety and functional continuity.

At present, laboratories mostly use dynamic shear rheology tests to directly test the fatigue life of asphalt. However, testing the fatigue life at different temperatures requires a lot of test costs and time. Therefore, developing a prediction model that can effectively predict the fatigue life of asphalt under different environmental conditions is an important research direction in the current field of pavement engineering.

The Chinese invention patent with application number 202310163323.1 discloses a method for quickly obtaining the asphalt strain-fatigue life curve. It uses the strain-fatigue life curve to predict the fatigue life of asphalt under different strain levels and solves the fatigue resistance performance of asphalt. The test test time is long, so its accuracy and precision are limited, and it cannot cope with the test of asphalt fatigue resistance at different test temperatures.

Contents of the invention

In view of this, the present invention provides a method for predicting the fatigue life of asphalt at different temperatures. The prediction model of the present invention is simple in form and more practical. The model parameters are optimized in batches through two optimization methods, which can increase the number of model predictions. accuracy.

The technical purpose of the present invention is achieved as follows:

The invention provides a method for predicting the fatigue life of asphalt at different temperatures, which includes the following steps:

S1 Set the experimental temperature, including test temperature and predicted temperature, obtain the composite modulus of asphalt at the experimental temperature, obtain the measured data of the composite modulus of asphalt, conduct a linear amplitude scanning test on the asphalt at the test temperature, and obtain the measured data of the fatigue life of the asphalt ;

S2 establishes a prediction model. The prediction model includes model parameters, and the model parameters are divided into two categories, namely first model parameters and second model parameters;

In step S2, the representation function of the prediction model is:

;

In the formula, For fatigue life prediction data,/> is the first model parameter, which is used to describe the composite modulus of asphalt at different temperatures,/> is the second model parameter, which is used to describe the fatigue life of asphalt at different temperatures,/> For external strain,/> To reduce the frequency;

S3 Select the reference temperature among the experimental temperatures, use the measured data of the composite modulus, and based on the time-temperature equivalence principle, obtain the master curve of the composite modulus at the reference temperature and the reduction frequency at other experimental temperatures. According to the first optimization method, based on the composite modulus Iteratively optimize the first model parameters by measuring the measured data, the master composite modulus curve at the reference temperature and the reduction frequency at other experimental temperatures, and obtain the optimized first model parameters. The first optimization method is based on the principle of the least squares method. Optimization;

S4 Iteratively optimizes the second model parameters based on the fatigue life measured data according to the second optimization method, and obtains the optimized second model parameters, where the second optimization method is an optimization method based on the principle of difference-by-difference method;

S5 uses the optimized first model parameters and the optimized second model parameters to update the prediction model to obtain the final prediction model. Use the final prediction model to predict the asphalt at the predicted temperature to obtain the prediction results of fatigue life.

Based on the above technical solution, preferably, step S3 includes:

S31 Select the reference temperature among the experimental temperatures and obtain the measured data of the composite modulus at the reference temperature, including data on the changes of the loss modulus and storage modulus with the loading frequency;

S32 processes and analyzes the measured data of the composite modulus at the reference temperature, and obtains the fitting curve of the loss modulus and storage modulus changing with the loading frequency, which is the master curve of the composite modulus at the reference temperature;

S33 According to the principle of time-temperature equivalence, convert the composite modulus master curve at the reference temperature to other experimental temperatures to obtain equivalent data, and obtain the reduced frequencies at other experimental temperatures;

S34 Based on the measured data of composite modulus, the composite modulus master curve at the reference temperature and the reduction frequency at other experimental temperatures, the first optimization method is used to iteratively optimize the first model parameters to obtain the optimized first model parameters.

Based on the above technical solution, preferably, step S34 includes:

S341 defines the first objective function and initializes the first model parameters;

S342 Calculate the first objective function value under the current first model parameters based on the measured data of composite modulus;

S343 Calculate the sensitivity matrix of the first objective function to the first model parameter, and calculate the adjustment parameter of the first model parameter based on the sensitivity matrix and the gradient of the first objective function value;

S344 superimposes the calculated adjustment parameters onto the current first model parameters to update the current first model parameters;

S345 uses the updated first model parameters to calculate the new first objective function value;

S346 Determine whether the new first objective function value converges:

If it converges, stop the iteration and output the first model parameter obtained by the final optimization as the optimized first model parameter;

If it does not converge, adjust the parameters of the sensitivity matrix according to the change in the first objective function value, and go to step S343.

Based on the above technical solution, preferably, the first objective function is:

;

In the formula, is the first objective function value,/> is the predicted data of composite modulus,/> is the measured data of composite modulus, and N is the number of data points.

Based on the above technical solution, preferably, step S343 includes:

For each first model parameter, calculate the partial derivative of the first objective function on the single first model parameter to obtain the sensitivity matrix J ;

Use the gradient operator to calculate the gradient of the first objective function ;

According to the sensitivity matrix J and gradient Calculate adjustment parameters:

;

In the formula, is the adjustment parameter, J is the sensitivity matrix,/> is the transpose of the sensitivity matrix,/> is the matrix parameter, I is the identity matrix,/> is the gradient.

Based on the above technical solution, preferably, step S4 includes:

S41 Define the second objective function , and initialize the second model parameters;

S42 sets the maximum number of iterations and adjustment step size;

S43 Select a second model parameter as the target parameter;

S44 Add the current value of the target parameter to the adjustment step to obtain the adjustment value of the target parameter;

S45 Based on the measured data of fatigue life, use the current value of the target parameter and the adjusted value of the target parameter to calculate the corresponding second objective function value. and/> ;

S46 comparison and/> The size of , determine the current value and adjustment step size of the target parameter for the next iteration, and go to step S44 until the convergence condition is reached or the maximum number of iterations is reached;

S47 Repeat steps S43-S46 to optimize each second model parameter to obtain optimized second model parameters.

Based on the above technical solution, preferably, the second objective function is:

;

In the formula, is the second objective function value,/> It is the measured data of fatigue life,/> is the average value of fatigue life measured data,/> For fatigue life prediction data,/> is the average value of the fatigue life prediction data.

Based on the above technical solution, preferably, in step S46, compare and/> The size of , determines the new current value of the target parameters of the next iteration, including:

like , then the adjustment value of the target parameter is used as the current value of the target parameter for the next iteration, and the adjustment step size remains unchanged;

like , then keep the current value of the target parameter unchanged, and subtract a step difference term from the adjustment step size as the new adjustment step size for the next iteration.

Based on the above technical solution, preferably, step S1 also includes:

The fatigue failure criterion is defined based on the composite modulus and phase angle, and the measured fatigue life data is calculated based on the results of the linear amplitude sweep test and the fatigue failure criterion;

Among them, the fatigue failure criterion is that the fatigue damage reaches 35% of the initial fatigue factor, the fatigue damage is the occurrence of cracks or fractures in the asphalt sample, and the initial fatigue factor is ,/> is the complex shear modulus,/> is the phase angle.

The method of the present invention has the following beneficial effects compared with the existing technology:

(1) The present invention obtains actual measured data of asphalt fatigue life at the test temperature through a linear amplitude scanning test, and establishes a prediction model of asphalt fatigue life at different temperatures based on linear viscoelasticity theory. This model is more practical than the existing technology. Strong, the model form is simpler and more accurate, which can save a lot of test costs and time, and is of great significance to the selection, design, construction technology optimization of pavement materials, as well as later pavement maintenance and management;

(2) The asphalt fatigue life model of the present invention has a total of 6 model parameters: ,/> , where,/> Used to describe the composite modulus of asphalt at different temperatures,/> Used to describe the fatigue life of asphalt at different temperatures. The presentation of this model can predict the fatigue life of asphalt at different temperatures;

(3) The present invention uses two optimization methods to optimize the first model parameters and the second model parameters respectively. According to the respective data characteristics of the composite modulus and fatigue life, corresponding optimization methods are set up to obtain the appropriate first model to the greatest extent. model parameters and second model parameters to increase the prediction accuracy of the prediction model.

Description of the drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without exerting creative efforts.

Figure 1 is a method flow chart according to an embodiment of the present invention;

Figure 2 is the main curve of the composite modulus at the reference temperature in the first embodiment of the present invention;

Figure 3 shows the correlation between the measured fatigue life and the predicted fatigue life of the first embodiment of the present invention;

Figure 4 is the correlation between the measured fatigue life and the predicted fatigue life of the second embodiment of the present invention;

Figure 5 shows the correlation between the measured fatigue life and the predicted fatigue life of the third embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

As shown in Figure 1, the present invention provides a method for predicting the fatigue life of asphalt at different temperatures, which includes the following steps:

S1 Set the experimental temperature, including test temperature and predicted temperature, obtain the composite modulus of asphalt at the experimental temperature, obtain the measured data of the composite modulus of asphalt, conduct a linear amplitude scanning test on the asphalt at the test temperature, and obtain the measured data of the fatigue life of the asphalt ;

S2 establishes a prediction model. The prediction model includes model parameters, and the model parameters are divided into two categories, namely first model parameters and second model parameters;

S3 Select the reference temperature among the experimental temperatures, use the measured data of the composite modulus, and based on the time-temperature equivalence principle, obtain the master curve of the composite modulus at the reference temperature and the reduction frequency at other experimental temperatures. According to the first optimization method, based on the composite modulus Iteratively optimize the first model parameters by measuring the measured data, the master composite modulus curve at the reference temperature and the reduction frequency at other experimental temperatures, and obtain the optimized first model parameters. The first optimization method is based on the principle of the least squares method. Optimization;

S4 Iteratively optimizes the second model parameters based on the fatigue life measured data according to the second optimization method, and obtains the optimized second model parameters, where the second optimization method is an optimization method based on the principle of difference-by-difference method;

S5 uses the optimized first model parameters and the optimized second model parameters to update the prediction model to obtain the final prediction model. Use the final prediction model to predict the asphalt at the predicted temperature to obtain the prediction results of fatigue life.

Specifically, step S1 also includes:

The fatigue failure criterion is defined based on the composite modulus and phase angle, and the measured fatigue life data is calculated based on the results of the linear amplitude sweep test and the fatigue failure criterion;

Among them, the fatigue failure criterion is that the fatigue damage reaches 35% of the initial fatigue factor, the fatigue damage is the occurrence of cracks or fractures in the asphalt sample, and the initial fatigue factor is ,/> is the complex shear modulus,/> is the phase angle.

In the embodiment of the present invention, the actual measured data of the composite modulus is the data at all experimental temperatures. It can be directly measured to obtain the measured data of the composite modulus at all temperatures, or it can be first measured to obtain the actual measured data of the composite modulus at the test temperature. data, and then select a temperature among the test temperatures as the target temperature. According to the time-temperature equivalence principle, the composite modulus measured data at the predicted temperature can be calculated based on the composite modulus at this target temperature.

Specifically, in the embodiment of the present invention, the prediction model representation function of step S2 is:

;

In the formula, For fatigue life prediction data,/> is the first model parameter, which is used to describe the composite modulus of asphalt at different temperatures,/> is the second model parameter, which is used to describe the fatigue life of asphalt at different temperatures,/> For external strain,/> To reduce the frequency.

In this embodiment, the prediction model of asphalt fatigue life has 6 model parameters: , , where,/> Used to describe the composite modulus of asphalt at different temperatures,/> Used to describe the fatigue life of asphalt at different temperatures, the presentation of this model can predict the fatigue life of asphalt at different temperatures.

Specifically, step S3 includes:

S31 Select the reference temperature among the experimental temperatures and obtain the measured data of the composite modulus at the reference temperature, including data on the changes of the loss modulus and storage modulus with the loading frequency;

S32 processes and analyzes the measured data of the composite modulus at the reference temperature, and obtains the fitting curve of the loss modulus and storage modulus changing with the loading frequency, which is the master curve of the composite modulus at the reference temperature;

S33 According to the principle of time-temperature equivalence, convert the composite modulus master curve at the reference temperature to other experimental temperatures to obtain equivalent data, and obtain the reduced frequencies at other experimental temperatures;

S34 Based on the measured data of composite modulus, the composite modulus master curve at the reference temperature and the reduction frequency at other experimental temperatures, the first optimization method is used to iteratively optimize the first model parameters to obtain the optimized first model parameters.

Among them, step S34 includes:

S341 defines the first objective function and initializes the first model parameters;

The first objective function is:

;

In the formula, is the first objective function value,/> is the predicted data of composite modulus,/> is the measured data of composite modulus, and N is the number of data points.

S342 Calculate the first objective function value under the current first model parameters based on the measured data of composite modulus;

S343 Calculate the sensitivity matrix of the first objective function to the first model parameter, and calculate the adjustment parameter of the first model parameter based on the sensitivity matrix and the gradient of the first objective function value;

For each first model parameter, calculate the partial derivative of the first objective function on the single first model parameter to obtain the sensitivity matrix J ;

Use the gradient operator to calculate the gradient of the first objective function ;

According to the sensitivity matrix J and gradient Calculate adjustment parameters:

;

In the formula, is the adjustment parameter, J is the sensitivity matrix,/> is the transpose of the sensitivity matrix,/> is the matrix parameter, I is the identity matrix,/> is the gradient.

S344 superimposes the calculated adjustment parameters onto the current first model parameters to update the current first model parameters;

S345 uses the updated first model parameters to calculate the new first objective function value;

S346 Determine whether the new first objective function value converges:

If it converges, stop the iteration and output the first model parameter obtained by the final optimization as the optimized first model parameter;

If it does not converge, adjust the parameters of the sensitivity matrix according to the change in the first objective function value, and go to step S343.

In this embodiment, the first optimization method is an optimization method based on the principle of least squares. The least squares method is a commonly used method in statistics and mathematics and is used to fit a set of data points to a function model. It is based on the following principle: for a given data point, the least squares method determines the best-fitting model parameters by minimizing the sum of squared residuals from the data point to the model predicted value. In other words, it looks for a functional model that minimizes the sum of squares of the residuals from the data points to the model predicted values. The purpose of this is to fit the model to best describe the relationship between the data points. The first optimization method in this embodiment is based on the principle of the least squares method, introducing a sensitivity matrix, and continuously iteratively optimizing the first model parameters based on gradient descent and the sensitivity matrix.

In this embodiment, through iterative optimization, the first model parameters can be continuously adjusted so that the prediction results are more consistent with the measured data, thereby improving the accuracy and reliability of the model. Using sensitivity matrices and gradients to calculate adjustment parameters can accelerate the adjustment process of the first model parameters and improve the efficiency and convergence speed of optimization. By setting appropriate convergence judgment conditions, the optimization process can be automated and manual intervention reduced.

Specifically, step S4 includes:

S41 Define the second objective function , and initialize the second model parameters;

The second objective function is:

;

In the formula, is the second objective function value,/> It is the measured data of fatigue life,/> is the average value of fatigue life measured data,/> For fatigue life prediction data,/> is the average value of the fatigue life prediction data.

S42 sets the maximum number of iterations and adjustment step size;

S43 Select a second model parameter as the target parameter;

S44 Add the current value of the target parameter to the adjustment step to obtain the adjustment value of the target parameter;

S45 Based on the measured data of fatigue life, use the current value of the target parameter and the adjusted value of the target parameter to calculate the corresponding second objective function value. and/> ;

S46 comparison and/> The size of , determine the current value and adjustment step size of the target parameter for the next iteration, and go to step S44 until the convergence condition is reached or the maximum number of iterations is reached;

This step specifically includes:

like , then the adjustment value of the target parameter is used as the current value of the target parameter for the next iteration, and the adjustment step size remains unchanged;

like , then keep the current value of the target parameter unchanged, and subtract a step difference term from the adjustment step size as the new adjustment step size for the next iteration.

S47 Repeat steps S43-S46 to optimize each second model parameter to obtain optimized second model parameters.

In this embodiment, the second optimization method is an optimization method based on the principle of the difference-by-difference method. The difference-by-difference method is a method for numerically approximating the solution of a differential equation. Its basic idea is to approximate the derivative in the differential equation through difference, thereby converting the differential equation into a difference equation, and then use the difference equation to perform numerical calculations. This embodiment introduces the idea of the difference-by-difference method into the optimization process of the second model parameters, that is, introduces the concepts of adjustment step size and adjustment value, and provides an update method of the adjustment step size, so as to approach the optimal value through step-by-step iteration. Optimal solution: During the iterative process, the second model parameters are constantly updated in the direction of the optimal solution. This method can increase the generalization ability of the model and speed up the convergence speed, making the optimization process more stable and improving optimization accuracy.

This invention verifies the effectiveness of the prediction model through three specific embodiments:

Example 1

In this example, a linear amplitude scan test was performed on the PAV-aged asphalt according to the AASHTO TP 101-12 standard. Four asphalts were selected for testing, namely PG 58-28#1, PG 58-28#2, PG 58-28#3, PG 58-28#4, # indicates different batches, the test temperatures are 13°C and 31°C, and the predicted temperatures are 19°C and 25°C. According to the test data, the fatigue life calculation results of the four kinds of asphalt under different applied strains are shown in Table 1.

Table 1 Fatigue life of four kinds of asphalt at different strain levels

According to the prediction model provided by the present invention, the composite modulus master curve at the reference temperature of this embodiment is obtained based on step S3, as shown in Figure 2. In Figure 2, the data points represented by triangles, circles, and diamonds are the existing data points. Experimental data of composite modulus at 19°C, 25°C, and 31°C. The data points represented by the cross are the composite modulus master curve at the reference temperature of 25°C. This master curve can obtain the reduction frequency at any temperature. .

In this embodiment, the first model parameters obtained by the first optimization method are As shown in Table 2 below:

Table 2 The first model parameters of the fatigue life prediction model of four asphalts

In this embodiment, the second model parameters obtained by the second optimization method are As shown in Table 3 below:

Table 3 Second model parameters of the fatigue life prediction model of four asphalts

The first model parameter and the second model parameter of the optimized prediction model and the reduction frequency at the temperature to be predicted (19℃, 25℃) Substituting into the representation function of the prediction model, the fatigue life at the predicted temperature is obtained.

In this example, in order to verify the accuracy of the prediction model, the fatigue life of PG 58-28#1, PG 58-28#2, PG 58-28#3, and PG 58-28#4 was measured at 19°C and 25°C. . The correlation between the measured fatigue life and the predicted fatigue life of the four asphalts at 19°C and 25°C is shown in Figure 3. The data points represented by circles in Figure 3 are model prediction values at two measured temperatures, while the data points represented by triangles are measured results from two intermediate temperatures. The simulated prediction values are in good agreement with the measured values. The model can Accurately predict fatigue life at different temperatures.

Example 2

In this example, a linear amplitude scan test was performed on the PAV-aged asphalt according to the AASHTO TP 101-12 standard. Four asphalts were selected for testing, namely PG 64-22#1, PG 64-22#2, PG 64-22#3, PG 64-22#4, # indicates different batches, the test temperatures are 19°C and 37°C, and the predicted temperatures are 25°C and 31°C. According to the test data, the fatigue life calculation results of the four kinds of asphalt under different applied strains are shown in Table 4.

Table 4 Fatigue life of four kinds of asphalt at different strain levels

Using the existing experimental data of composite modulus and based on the time-temperature equivalence principle, the composite modulus master curve at the reference temperature and the reduction frequency at any temperature are obtained .

In this embodiment, the first model parameters obtained by the first optimization method are As shown in Table 5 below:

Table 5 Fatigue life prediction model parameters of four kinds of asphalt

In this embodiment, the second model parameters obtained by the second optimization method are As shown in Table 6 below:

Table 6 Fatigue life prediction model parameters of four kinds of asphalt

The first model parameter and the second model parameter of the optimized prediction model and the reduction frequency at the temperature to be predicted (25℃, 31℃) Substituting into the representation function of the prediction model, the fatigue life at the predicted temperature is obtained.

In order to verify the accuracy of the model, actual fatigue life measurements at 25°C and 31°C are required for PG 64-22#1, PG 64-22#2, PG 64-22#3, and PG 64-22#4. The correlation between the measured fatigue life and the predicted fatigue life of the four asphalts at 25°C and 31°C is shown in Figure 4. The data points represented by circles in Figure 4 are model prediction values at two measured temperatures, while the data points represented by triangles are measured results from two intermediate temperatures. The simulated prediction values are in good agreement with the measured values. The model can Accurately predict fatigue life at different temperatures.

Example 3

In this example, a linear amplitude scan test was performed on the PAV-aged asphalt according to the AASHTO TP 101-12 standard. Four asphalts were selected for testing, namely PG 70-22#1, PG 70-22#2, PG 70-22#R1, PG70-22#R2, # indicates different batches, R indicates after modification, the test temperatures are 25°C and 43°C, and the predicted temperatures are 31°C and 37°C. According to the test data, the calculation results of the asphalt fatigue life of the four kinds of asphalt under different applied strains are shown in Table 7.

Table 7 Fatigue life of four asphalts at different strain levels

Using the existing experimental data of composite modulus and based on the time-temperature equivalence principle, the composite modulus master curve at the reference temperature and the reduction frequency at any temperature are obtained .

In this embodiment, the first model parameters obtained by the first optimization method are As shown in Table 8 below:

Table 8 Fatigue life prediction model parameters of four kinds of asphalt

In this embodiment, the second model parameters obtained by the second optimization method are As shown in Table 9 below:

Table 9 Fatigue life prediction model parameters of four kinds of asphalt

The first and second model parameters of the optimized prediction model and the reduction frequency at the temperatures to be predicted (31°C, 37°C) Substituting into the representation function of the prediction model, the fatigue life at the predicted temperature is obtained.

In order to verify the accuracy of the model, actual fatigue life measurements at 31°C and 37°C are required for PG 70-22#1, PG 70-22#2, PG 70-22#R1, and PG 70-22#R2. The correlation between the measured fatigue life and the predicted fatigue life of the four asphalts at 31°C and 37°C is shown in Figure 5. The data points represented by circles in Figure 5 are model prediction values at two measured temperatures, while the data points represented by triangles are measured results from two intermediate temperatures. The simulated prediction values are in good agreement with the measured values. The model can Accurately predict fatigue life at different temperatures.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of the present invention. within the scope of protection.

Claims (9)

1. The method for predicting the fatigue life of asphalt at different temperatures is characterized by comprising the following steps:

s1, setting an experimental temperature, wherein the experimental temperature comprises a test temperature and a prediction temperature, acquiring composite modulus of asphalt at the experimental temperature, obtaining composite modulus actual measurement data of the asphalt, and performing a linear amplitude scanning test on the asphalt at the test temperature to obtain fatigue life actual measurement data of the asphalt;

s2, establishing a prediction model, wherein the prediction model comprises model parameters, and dividing the model parameters into two types, namely a first model parameter and a second model parameter;

in step S2, the representation function of the prediction model is:

；

in the method, in the process of the invention,for fatigue life prediction data, < >>Is a first model parameter describing the complex modulus of asphalt at different temperatures, +.>Is a second model parameter describing fatigue life of asphalt at different temperatures, +.>For external strain->To reduce the frequency;

s3, selecting a reference temperature from experimental temperatures, obtaining a composite modulus main curve at the reference temperature and a reduced frequency at other experimental temperatures based on a time-temperature equivalent principle by utilizing composite modulus measured data, and iteratively optimizing first model parameters according to a first optimization method based on the composite modulus measured data, the composite modulus main curve at the reference temperature and the reduced frequency at other experimental temperatures to obtain optimized first model parameters, wherein the first optimization method is an optimization method based on a least square method principle;

s4, iteratively optimizing second model parameters based on fatigue life actual measurement data according to a second optimization method to obtain optimized second model parameters, wherein the second optimization method is an optimization method based on a difference-by-difference method principle;

and S5, updating the prediction model by using the optimized first model parameter and the optimized second model parameter to obtain a final prediction model, and predicting asphalt at a prediction temperature by using the final prediction model to obtain a prediction result of fatigue life.

2. A method for predicting fatigue life of asphalt at different temperatures as defined in claim 1, wherein step S3 comprises:

s31, selecting a reference temperature from experimental temperatures, and acquiring composite modulus measured data at the reference temperature, wherein the data comprise the change of loss modulus and storage modulus along with loading frequency;

s32, processing and analyzing the composite modulus measured data at the reference temperature to obtain a fitting curve of the loss modulus and the storage modulus along with the change of the loading frequency, namely a composite modulus main curve at the reference temperature;

s33, according to a time-temperature equivalent principle, converting a composite modulus main curve at a reference temperature into other experimental temperatures to acquire equivalent data, and obtaining a reduced frequency at the other experimental temperatures;

and S34, carrying out iterative optimization on the first model parameters by adopting a first optimization method based on the measured data of the composite modulus, the main curve of the composite modulus at the reference temperature and the reduction frequency at other experimental temperatures to obtain the optimized first model parameters.

3. A method for predicting fatigue life of asphalt at different temperatures as defined in claim 2, wherein step S34 comprises:

s341 defines a first objective function and initializes a first model parameter;

s342, calculating a value of a first objective function under the current first model parameter according to the actual measurement data of the composite modulus;

s343, calculating a sensitive matrix of the first objective function to the first model parameter, and calculating an adjustment parameter of the first model parameter according to the sensitive matrix and the gradient of the value of the first objective function;

s344, superposing the calculated adjustment parameters to the current first model parameters to update the current first model parameters;

s345 calculates a new value of the first objective function by using the updated first model parameters;

s346 judges whether or not the value of the new first objective function converges:

if the first model parameters are converged, stopping iteration, and outputting the first model parameters obtained through final optimization as optimized first model parameters;

if not, the parameters of the sensitive matrix are adjusted according to the change condition of the value of the first objective function, and the step S343 is proceeded.

4. A method for predicting fatigue life of asphalt at different temperatures as defined in claim 3, wherein the first objective function is:

；

in the method, in the process of the invention,for the value of the first objective function, +.>Is the predicted data of complex modulus, +.>Is the measured data of the composite modulus, and N is the number of data points.

5. The method of claim 4, wherein step S343 comprises:

for each first model parameter, calculating the partial derivative of the first objective function on the single first model parameter to obtain a sensitive matrixJ；

Computing gradients of a first objective function using gradient operators；

Based on a sensitivity matrixJAnd gradientCalculating an adjustment parameter:

；

in the method, in the process of the invention,in order to adjust the parameters of the device,Jis a sensitive matrix->Transpose of sensitive matrix +.>As a parameter of the matrix,Iis a matrix of units which is a matrix of units,is a gradient.

6. A method for predicting fatigue life of asphalt at different temperatures as defined in claim 1, wherein step S4 comprises:

s41 defining a second objective functionInitializing second model parameters;

s42, setting the maximum iteration times and the adjustment step length;

s43, selecting a second model parameter as a target parameter;

s44, superposing the current value of the target parameter with an adjustment step length to obtain an adjustment value of the target parameter;

s45, according to the actual measurement data of the fatigue life, calculating corresponding second objective function values by using the current value of the objective parameter and the adjustment value of the objective parameterAnd->；

S46 comparisonAnd->To determine the size of the next iterationThe current value of the target parameter and the adjustment step length are transferred to the step S44 until reaching the convergence condition or reaching the maximum iteration times;

s47, repeating the steps S43-S46, and optimizing each second model parameter to obtain optimized second model parameters.

7. A method of predicting fatigue life of asphalt at different temperatures as recited in claim 6, wherein the second objective function is:

；

in the method, in the process of the invention,for the second objective function value, < >>For fatigue life measured data, < >>For the average of fatigue life measured data, +.>For fatigue life prediction data, < >>Is the average of the fatigue life prediction data.

8. The method for predicting fatigue life of asphalt at different temperatures as defined in claim 6, wherein in step S46, the comparison is madeAnd->To determine the size of the next stepNew current values of the target parameters of the generation, including:

if it isTaking the adjustment value of the target parameter as the current value of the target parameter of the next iteration, and keeping the adjustment step length unchanged;

if it isAnd keeping the current value of the target parameter unchanged, and subtracting a step difference item from the adjustment step length to serve as a new adjustment step length of the next iteration.

9. The method for predicting fatigue life of asphalt at different temperatures as defined in claim 6, wherein step S1 further comprises:

defining a fatigue failure criterion according to the composite modulus and the phase angle, and calculating to obtain fatigue life actual measurement data according to the result of the linear amplitude scanning test and the fatigue failure criterion;

wherein the fatigue failure criterion is that the fatigue failure reaches 35% of the initial fatigue factor, the fatigue failure is that the asphalt sample has cracks or breaks, and the initial fatigue factor is that，/>Is complex shear modulus>Is the phase angle.