Method for early warning of performance of asphalt pavement with crumb rubber based on swelling-dissolution state of crumb rubber

Abstract

The application discloses a rubber powder asphalt pavement performance early warning method based on a rubber powder swelling-dissolution state, rubber powder modified asphalt preparation and asphalt mortar material attribute parameter acquisition, rubber powder asphalt performance index and rubber powder particle swelling-dissolution state change parameter acquisition based on macro-micro test, numerical processing of the obtained parameters, quantitative characterization of the swelling-dissolution state of the rubber powder particles, construction of a prediction model of the performance of the rubber powder modified asphalt and the swelling-dissolution state by using a neural network algorithm, modeling and microscopic parameter assignment of the heterogeneous rubber powder modified asphalt mortar in the discrete element software, obtaining the internal swelling-dissolution state of the surface layer rubber powder asphalt by using the neural network prediction model, realizing fine modeling in the discrete element and deducing the permanent deformation behavior of the pavement structure under repeated loads, and realizing the early warning of the performance of the rubber powder modified asphalt pavement structure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of neural network algorithm and asphalt pavement structure performance early warning, in particular to a crumb rubber asphalt pavement performance early warning method based on crumb rubber swelling-dissolution state. BACKGROUND

[0002] At present, the number of waste rubber tires in China is increasing year by year, which has attracted more and more attention of researchers. How to realize waste utilization has become a research hotspot. In the field of road engineering, crumb rubber modified asphalt is an effective way to realize the utilization of rubber waste. Compared with ordinary asphalt, crumb rubber modified asphalt has more excellent performance. However, at present, the use of crumb rubber modified asphalt is mostly based on isotropic theory, without micro-analysis of crumb rubber modified asphalt, ignoring the influence of rubber particle solubility, swelling degree, volume expansion rate and mass attenuation rate, etc. in it, so as to effectively reflect its performance. At present, intelligentization of road engineering is a research hotspot and future development direction. With the development of machine learning and neural network, it is possible to use it for pavement structure performance prediction. However, the current technology is not fully combined. Based on this, the present application focuses on the crumb rubber swelling-dissolution state of crumb rubber modified asphalt pavement structure performance early warning, uses neural network to develop a neural network prediction model of crumb rubber modified asphalt performance and crumb rubber distribution state, which can effectively realize the prediction of crumb rubber solubility, swelling degree, volume expansion rate and mass attenuation rate, etc. and take asphalt performance index as the pavement structure performance state evaluation to early warn the pavement structure performance. SUMMARY

[0003] The technical problem solved by the present application is that the prior art does not micro-analyze crumb rubber modified asphalt, ignores the influence of rubber particle solubility, swelling degree, volume expansion rate and mass attenuation rate, etc. in it, so as to effectively reflect its performance, etc. The present application provides a crumb rubber asphalt pavement performance early warning method based on crumb rubber swelling-dissolution state.

[0004] Technical scheme

[0005] The crumb rubber asphalt pavement performance early warning method based on crumb rubber swelling-dissolution state comprises the following steps:

[0006] S1, crumb rubber modified asphalt preparation and asphalt crumb rubber material attribute parameter acquisition;

[0007] S2, obtaining crumb rubber modified asphalt performance index and crumb rubber particle swelling-dissolution state change parameter based on macro-microscopic test;

[0008] S3, numerical value of crumb rubber modified asphalt material attribute parameter and crumb rubber particle swelling-dissolution state change parameter;

[0009] S4, constructing a prediction model of the performance of crumb rubber modified asphalt and the swelling-dissolution state of crumb rubber based on a neural network;

[0010] S5, modeling of crumb rubber modified asphalt mortar and assignment of mesoscopic parameters;

[0011] S6, a performance early warning method of crumb rubber modified asphalt pavement based on the swelling-dissolution state of crumb rubber by using a neural network algorithm.

[0012] As a preferred technical solution of the present application: in S1, first, orthogonal crumb rubber swelling-dissolution tests are carried out by using a wet method under the conditions of a crumb rubber content of 5-35%, a treatment temperature of 140-195°C, and a treatment time of 0.5-3.0h, and 180 groups of crumb rubber modified asphalt are prepared; second, crumb rubber modified asphalt is solidified in epoxy resin to form resin samples, and 180 groups of crumb rubber modified asphalt mortar samples are prepared; and finally, nanoindentation tests are carried out on all samples at a microscale to extract the elastic modulus and viscoelasticity indexes of the crumb rubber phase and the asphalt phase.

[0013] As a preferred technical solution of the present application: the specific steps of carrying out nanoindentation tests on the crumb rubber phase in different crumb rubber modified asphalt mortar samples to extract the elastic modulus and viscoelasticity indexes of the crumb rubber phase and the asphalt phase in S1 at a microscale are as follows: first, nanoindentation tests are carried out on the crumb rubber phase in different crumb rubber modified asphalt mortar samples, and the nanoindentation crumb rubber load-depth curves of crumb rubber particles under all different swelling-dissolution states are obtained; the unloading section of the curve is fitted by using the Oliver-Pharr model, and the elastic modulus of all crumb rubber particles is calculated according to formulae (1)-(2); second, nanoindentation tests are carried out on the asphalt phase, the elastic modulus of the asphalt phase is calculated according to formulae (1)-(2), and the creep process of asphalt, i.e. the loading section of the test, is obtained under the condition that the selected asphalt constitutive model is the Burgers viscoelasticity model; and the viscoelasticity parameters of asphalt particles, i.e. the spring unit modulus E1 and the viscosity pot unit viscosity τ1 in the Maxwell model and the spring unit modulus E2 and the viscosity pot unit viscosity τ2 in the Kelvin model, are calculated according to formulae (3)-(4).

[0014]

[0015]

[0016] In the formula, E r is the complex elastic modulus of the indenter and the tested sample, S is the slope at the top of the unloading section curve, β is the geometric correction coefficient of the indenter, A is the contact area, E s is the elastic modulus of the tested sample, v s is the Poisson's ratio of the tested sample, E i is the elastic modulus of the indenter, v i is the Poisson's ratio of the indenter.

[0017]

[0018]

[0019] In the formula: h(t) is the indentation depth, P is the indentation load, θ is the half-opening angle of the indenter, t is the time, E1 is the spring unit modulus in the Maxwell model, τ1 is the viscosity of the dashpot unit in the Maxwell model, E2 is the spring unit modulus in the Kelvin model, and τ2 is the viscosity of the dashpot unit in the Kelvin model.

[0020] As a preferred technical solution of the present application: in S2, first, according to the PG grading test specification, the flash point, viscosity, rheological properties G1* / sinσ, mass change L T , rheological properties of PAV after rheological G2* / sinσ, modulus stiffness S, failure strain of each group of asphalt are determined; secondly, for the prepared crumb rubber modified asphalt, due to the difference in preparation process, the existence state of crumb rubber particles in the asphalt is different, the swelling-dissolution state of the crumb rubber particles is analyzed by using reflux elution method, toluene is used as eluent to completely elute the asphalt, and the scanning experiment is carried out under X-Ray CT to obtain the scanning graph of each group of samples; the scanning graph is converted into a gray scale graph by using Matlab, the histogram equalization function is used to enhance the contrast of the image, and the watershed segmentation method is used to completely segment the image of asphalt and crumb rubber particles; then, the volume expansion rate V R and the mass attenuation rate T of the crumb rubber particles in all different swelling-dissolution states are measured and calculated according to formula (5)-(6);

[0021]

[0022]

[0023] In the formula: V R is the volume expansion rate of the crumb rubber particles, R1 is the average diameter of the swelled crumb rubber, R0 is the average diameter of the unswelled crumb rubber; T is the mass attenuation rate, M is the mass of the crumb rubber modified asphalt sample before elution, β is the crumb rubber content, and m is the mass of the crumb rubber after elution.

[0024] As a preferred technical solution of the present application: in S3, in all samples, first, for the dual-phase crumb rubber modified asphalt material, the phase elastic modulus X obeys the Gaussian model, that is, X~N(μ,σ 2 ); the frequency histogram of the elastic modulus of the crumb rubber particles and the frequency histogram of the elastic modulus of the asphalt molecules are obtained; the probability density function of the mixed Gaussian model is calculated according to formula (7)-(8) to obtain the elastic modulus mixed Gaussian model of all crumb rubber modified asphalts; secondly, the swelling-dissolution state change parameter Y of the crumb rubber particles also obeys the Gaussian model, that is, Y~N(μ,σ2 ) ; obtaining the volume expansion rate V of the rubber powder particles R and the frequency histogram of the mass attenuation rate T, calculating the probability density function of the mixed Gaussian model according to formula (7)-(8), obtaining the swelling characteristic parameters of the rubber powder particles in the rubber powder modified asphalt sample corresponding to the group number, and realizing the numerical value of the material attribute parameters of the rubber powder modified asphalt and the swelling-dissolution state change parameter of the rubber powder particles;

[0025]

[0026] Wherein: alpha k is the probability of sample data belonging to the kth sub-model, alpha k >=0, phi (x|N k ) is a single Gaussian model probability function, P (x|N) is the probability density function of the mixed Gaussian model;

[0027]

[0028] is the kth sub-model, x is sample data, mu k is the data mean (expectation), sigma k is the data standard deviation.

[0029] As a preferred technical solution of the present application: in the S4, the performance indicators of the rubber powder modified asphalt are taken as inputs (m layers), and the corresponding state data of the rubber powder modified asphalt (the swelling characteristic parameters of the rubber powder particles) are taken as outputs (n layers); a three-layer BP network with one hidden layer is used to establish a prediction model, first, the training set of the model is built: the neural network toolbox in MATLAB is used to train the network, the training sample data is input into the network, the Gaussian mixture function according to formula (9) is taken as the excitation function of the hidden layer neurons, the logarithmic function of the Gaussian mixture model according to formula (10) is taken as the excitation function of the output layer neurons, the network training function and the performance function are determined, and the number of hidden layer neurons is initially set; secondly, the network parameters are set, after the parameters are set, the network is trained; finally, after the training is completed, the experimental data can be input to check the accuracy of the model in predicting the distribution of the rubber modified asphalt, or the performance indicators (flash point, viscosity, G1* / sin sigma, mass change L T , G2* / sin sigma, modulus stiffness S, and failure strain) of the rubber powder modified asphalt can be input to predict the distribution state thereof;

[0030]

[0031]

[0032] Wherein: xj For the jth observation, α k The probability that the observation belongs to the kth component model, α k ≥ 0, φ(x|θ k ) is the kth component Gaussian model probability function, P(x|θ) is the probability density function of the Gaussian mixture model, θ k The probability that each component model occurs in the mixture model, log L(θ) is the logarithmic function of the Gaussian mixture model.

[0033] As a preferred technical solution of the application: in S5, when assigning the mesoscopic mechanics parameters, the macroscopic performance parameters of all particles calculated by the nanoindentation test in S1 should be first compared, and the macroscopic performance parameter and mesoscopic mechanics parameter conversion calculation is carried out according to formula (11)-(12) and linear contact model, Burgers model

[0034] t = A / L (11)

[0035] E = 2(1 + v)G (12)

[0036] In the formula: A is the contact area, L is the length of the viscoelastic beam, that is, the distance between the centers of adjacent particle units, E is the elastic modulus, v is the Poisson's ratio, and G is the shear modulus.

[0037] As a preferred technical solution of the present application: in S5, first, in the PFC software, for different content of crumb rubber modified asphalt, the particles are arranged in a regular arrangement, the particle size is kept consistent, all particles are generated, combined with the random generation of crumb rubber particles, the pointer is traversed in a loop, then the non-uniform coefficient distributed elastic modulus mixed Gaussian model (7)-(8) established in step S3 is set, the random number x in (0-1) is set, P is the distribution probability, and mu and sigma are the elastic modulus parameters, different color indexes are given to the crumb rubber and asphalt particles, the particle grouping and visualization are realized, until the crumb rubber reaches the specified content, and the initial model of the non-homogeneous crumb rubber modified asphalt is constructed; secondly, the nested loop is set, all crumb rubber particle pointers are searched, then the non-uniform coefficient distributed elastic modulus mixed Gaussian model (7)-(8) established in step S3 is set, the random number x` in (0-1) is set, P is the distribution probability, and mu and sigma are the swelling characteristic parameters of the crumb rubber particles, the crumb rubber modified asphalt model is established according to the swelling-dissolution state of the crumb rubber; the contact pointer of all particles is read, the particle type of the contact connection is judged according to the color index of the crumb rubber and asphalt particles; the linear contact model is used for the contact between the crumb rubber particles, and the Burgers model is used for the contact between the asphalt particles and the crumb rubber particles to give the constitutive model, and the macroscopic performance parameters and the microscopic mechanical parameters of each particle are calculated according to formula (11)-(12) and the linear contact model and the Burgers model, and the microscopic parameters are distinguished and valued in the software, and finally the numerical simulation of the non-homogeneous crumb rubber modified asphalt in different swelling-dissolution states is realized.

[0038] As a preferred technical solution of the present application: in S6, a multi-scale coupling analysis method is used to establish a macroscopic and microscopic coupled pavement structure model of discrete elements (non-homogeneous crumb rubber modified asphalt surface layer)-finite elements (pavement structure) according to a set gradation; in different pavement bearing modes, the performance of the actual pavement structure asphalt sample is measured, the internal swelling-dissolution state of the surface layer crumb rubber modified asphalt is obtained by using the neural network prediction model of the crumb rubber modified asphalt performance and the mortar swelling-dissolution state, fine modeling is realized in the discrete element, the permanent deformation behavior of the pavement structure under repeated load is deduced, and the performance of the crumb rubber modified asphalt pavement structure is prewarned.

[0039] Advantages:

[0040] Compared with the prior art, the present application has the following advantages:

[0041] (1) The volume expansion rate and the mass attenuation rate of the crumb rubber particles are used to characterize the swelling-dissolution behavior state of the crumb rubber particles in the crumb rubber modified asphalt;

[0042] (2) The BP neural network algorithm is applied to construct a prediction model of the performance and swelling-dissolution state of rubber powder modified asphalt based on the dataset training, which has strong generalization ability.

[0043] (3) Based on the mixed Gaussian model with non-uniform coefficient distribution and the particle random generation algorithm in discrete element method, the model of non-uniform rubber powder modified asphalt is constructed.

[0044] (4) Through macroscopic experiments, microscopic experiments and particle micromechanical state assignment, multi-scale performance early warning is achieved at three scales: particle-rubber powder modified asphalt-pavement. Attached Figure Description

[0045] Figure 1 This is a flowchart of the application;

[0046] Figure 2 This is a schematic diagram of the neural network prediction model of this application. Detailed Implementation

[0047] The invention will now be further described with reference to the accompanying drawings.

[0048] Example 1

[0049] like Figure 1 As shown, the performance early warning method for rubber powder asphalt pavement based on the swelling-dissolution state of rubber powder includes the following steps: S1, preparation of rubber powder modified asphalt and acquisition of asphalt mastic material property parameters;

[0050] First, a wet process was used, under different preparation conditions of different adhesive powder dosages, processing temperatures, and processing times:

[0051] ① The amounts of adhesive powder added are 5%, 10%, 15%, 20%, 25%, and 35%, respectively;

[0052] ② The processing temperatures were 140, 150, 160, 175, and 195℃, respectively;

[0053] ③ The processing times were 0.5, 1.0, 1.5, 2.0, 2.5, and 3.0 hours, respectively;

[0054] Orthogonal swelling-dissolution tests were conducted to prepare 180 groups of rubber-modified asphalt. Next, the rubber-modified asphalt was cured in epoxy resin to form resin samples, and 180 groups of rubber-modified asphalt mortar samples were prepared.

[0055] The nano-indentation test is performed on the rubber powder phase in different rubber powder modified asphalt mortar samples to obtain the nano-indentation rubber powder load-depth curves of the rubber powder particles in all different swelling-dissolution states, the unloading section of the curve is fitted by using the Oliver-Pharr model, and the elastic modulus of all the rubber powder particles is calculated according to formula (1)-(2); secondly, the nano-indentation test is performed on the asphalt phase, the elastic modulus of the asphalt phase is calculated according to formula (1)-(2), and the creep process of the asphalt, i.e. the loading section of the test, is obtained under the condition that the selected asphalt constitutive model is the Burgers viscoelastic model, and the viscoelastic parameters of the asphalt particles, i.e. the spring unit modulus E1 and the viscous pot unit viscosity τ1 in the Maxwell model and the spring unit modulus E2 and the viscous pot unit viscosity τ2 in the Kelvin model, are calculated according to formula (3)-(4)

[0056]

[0057]

[0058] In the formula: E r is the complex elastic modulus of the indenter and the tested sample, S is the slope at the top of the unloading section curve, β is the geometric correction coefficient of the indenter, A is the contact area, E s is the elastic modulus of the tested sample, v s is the Poisson's ratio of the tested sample, E i is the elastic modulus of the indenter, v i is the Poisson's ratio of the indenter).

[0059]

[0060]

[0061] In the formula: h(t) is the indentation depth, P is the indentation load, θ is the half-opening angle of the indenter, t is the time, E1 is the spring unit modulus in the Maxwell model, τ1 is the viscous pot unit viscosity in the Maxwell model, E2 is the spring unit modulus in the Kelvin model, and τ2 is the viscous pot unit viscosity in the Kelvin model.

[0062] S2, obtaining the performance index of the rubber powder modified asphalt and the swelling-dissolution state change parameter of the rubber powder particles based on the macro-micro test;

[0063] Firstly, according to the PG grading test specification, the flash point of each group of asphalt is determined by using the Cleveland open cup method; the viscosity of each group of asphalt is determined by using the Brookfield viscometer method; the rheological property G1* / sinσ of each group of asphalt is determined by using the dynamic shear rheometer method; and the mass change L T; the PAV post-rheological property of each group of asphalt is determined by using the PAV post-rheological property test of asphalt; the modulus stiffness S of each group of asphalt is determined by using the bending beam rheometer method; the failure strain of each group of asphalt is determined by using the direct tensile method; secondly, for the 180 groups of crumb rubber modified asphalt, due to the different preparation processes, the existence state of the crumb rubber particles in the asphalt is different, the swelling-dissolution state of the crumb rubber particles is analyzed by using the reflux elution method, the asphalt is completely eluted by using toluene as the eluent, and the scanning experiment is carried out under the X-Ray CT to obtain the scanning graph of each group of samples; the scanning graph is converted into a gray scale graph by using Matlab, the histogram equalization function is used to enhance the contrast of the image, and the watershed segmentation method is used to completely segment the image of the asphalt and the crumb rubber particles; then, the volume expansion rate V R and the mass attenuation rate T of the crumb rubber particles in all different swelling-dissolution states are measured and calculated according to formulae (5)-(6).

[0064]

[0065]

[0066] In the formula: V R is the volume expansion rate of the crumb rubber particles, R1 is the average diameter of the swelled crumb rubber, R0 is the average diameter of the unswelled crumb rubber; T is the mass attenuation rate, M is the mass of the crumb rubber modified asphalt sample before elution, β is the crumb rubber content, and m is the mass of the crumb rubber after elution.

[0067] S3, numerical value of material attribute parameters and swelling-dissolution state change parameter of crumb rubber particles of crumb rubber modified asphalt

[0068] Among the 180 groups of samples, firstly, for the two-phase crumb rubber modified asphalt material, the phase elastic modulus X obeys the Gaussian model, that is, X ~ N(μ, σ 2 ); the frequency histogram of the elastic modulus of the crumb rubber particles and the frequency histogram of the elastic modulus of the asphalt molecules are obtained; the probability density function of the mixed Gaussian model is calculated according to formulae (7)-(8), and the mixed Gaussian model of the elastic modulus of all crumb rubber modified asphalts is obtained; secondly, the swelling-dissolution state change parameter Y of the crumb rubber particles also obeys the Gaussian model, that is, Y ~ N(μ, σ 2 ); the frequency histogram of the volume expansion rate V R and the mass attenuation rate T of the crumb rubber particles is obtained, the probability density function of the mixed Gaussian model is calculated according to formulae (7)-(8), and the mixed Gaussian model of the swelling characteristic parameters of the crumb rubber particles in the 180 groups of crumb rubber modified asphalt samples is obtained; the numerical value of the material attribute parameters and the swelling-dissolution state change parameter of the crumb rubber particles of the crumb rubber modified asphalt is realized.

[0069]

[0070] wherein: a k is the probability of the sample data belonging to the kth sub-model, a k ≥ 0, φ(x|N k ) is a single Gaussian model probability function, P(x|N) is the probability density function of the mixture Gaussian model.

[0071]

[0072] is the kth sub-model, x is the sample data, μ k is the data mean (expectation), σ k is the data standard deviation.

[0073] S4, constructing a prediction model of the performance of crumb rubber modified asphalt and the swelling-dissolution state based on a neural network;

[0074] As Figure 2 shown, the performance indicators of crumb rubber modified asphalt are taken as inputs (m layers), i.e., (flash point, viscosity, G1* / sinσ, mass change L T , G2* / sinσ, modulus stiffness S, and failure strain), and the corresponding state data of crumb rubber modified asphalt (swelling characteristic parameters of crumb rubber particles) are taken as outputs (n layers); a three-layer BP network with one hidden layer is used to establish the prediction model, the number of neurons is 6, first, the training set of the model is built: the neural network toolbox in MATLAB is used for network training, the training sample data is normalized by using the permnmx function and then input into the network, the Gaussian mixture function as shown in formula (9) is used as the excitation function of the hidden layer neurons, the logarithmic function of the Gaussian mixture model as shown in formula (10) is used as the excitation function of the output layer neurons, the network training function is traingdx, the performance function is mse, and the number of hidden layer neurons is initially set to 10; secondly, the network parameters are set: the number of network iterations epochs is 5000 times, the expected error goal is 0.00000001, the learning rate l r is 0.01, after the parameters are set, the network is trained; after the training is finally completed, the experimental data can be input to check the accuracy of the model in predicting the distribution of rubber modified asphalt, or the performance indicators of crumb rubber modified asphalt (flash point, viscosity, G1* / sinσ, mass change L T , G2* / sinσ, modulus stiffness S, and failure strain) can be input to predict the distribution state thereof.

[0075]

[0076]

[0077] wherein: x j is the jth observation, α k is the probability that the observation belongs to the kth component model, α k ≥ 0, φ(x|θ k ) is the kth component Gaussian model probability function, P(x|θ) is the probability density function of the mixture Gaussian model, θ k is the probability of each component model occurring in the mixture model, log L(θ) is the log function of the Gaussian mixture model.

[0078] S5, modeling of non-homogeneous rubber powder modified asphalt mortar and microscopic parameter assignment;

[0079] In the microscopic mechanics parameter assignment, all the macroscopic performance parameters of the particles calculated by the nanoindentation test in S1 should be first referred to, and the macroscopic performance parameter and microscopic mechanics parameter conversion calculation is performed according to the formula (11)-(12) and the linear contact model formula (13), Burgers model formula (14)

[0080] t = A / L (11)

[0081] E = 2(1 + v)G (12)

[0082] In the formula, A is the contact area, L is the length of the viscoelastic beam, i.e. the distance between the centers of adjacent particle units, E is the elastic modulus, v is the Poisson's ratio, and G is the shear modulus.

[0083] Linear contact model:

[0084]

[0085] Burgers model:

[0086] K kn = E2t

[0087] C kn = τ2t

[0088] K Mn = E1t

[0089] C Mn = τ1t

[0090]

[0091]

[0092]

[0093]

[0094] where: K kn is the normal contact stiffness in the Kelvin model, C kn is the normal contact viscosity in the Kelvin model, K Mn is the normal contact stiffness in the Maxwell model, C Mn is the normal contact viscosity in the Maxwell model, K ks is the shear contact stiffness in the Kelvin model, C ks is the shear contact viscosity in the Kelvin model, K ms is the shear contact stiffness in the Maxwell model, C ms is the shear contact viscosity in the Maxwell model.

[0095] Firstly, in the PFC software, for different contents of crumb rubber modified asphalt, the particles are arranged in a regular arrangement, and the particle size is kept consistent. All particles are generated, combined with the random generation of crumb rubber particles, and then the non-uniform coefficient distribution elastic modulus mixed Gaussian model (7)-(8) established in step S3 is set. The random number x in (0-1) is set, P is the distribution probability, and μ and σ are the elastic modulus parameters. Different color indexes are given to the crumb rubber and asphalt particles to realize particle grouping and visualization until the specified content of crumb rubber is reached. The initial model of non-homogeneous crumb rubber modified asphalt is constructed; secondly, the nested loop is set, and the pointer of all crumb rubber particles is searched, and then the non-uniform coefficient distribution elastic modulus mixed Gaussian model (7)-(8) established in step S3 is set. The random number x` in (0-1) is set, P is the distribution probability, and μ and σ are the swelling characteristic parameters of crumb rubber particles. The crumb rubber modified asphalt model is established according to the swelling-dissolution state of crumb rubber; the contact pointer of all particles is read, and the color index of crumb rubber and asphalt particles is used to determine the type of particles connected by the contact; the linear contact model is used for the contact between crumb rubber particles, and the Burgers model is used for the contact between asphalt particles and crumb rubber particles to give the constitutive model. Each particle is calculated according to formula (11)-(12) and the linear contact model and the Burgers model, and the microscopic parameters are distinguished and assigned in the software. Finally, the numerical simulation of non-homogeneous crumb rubber modified asphalt in different swelling-dissolution states of all crumb rubber particles is realized;

[0096] S6, a crumb rubber modified asphalt pavement performance early warning method based on the swelling-dissolution state of crumb rubber using a neural network algorithm;

[0097] The macro-micro coupling pavement structure model of discrete element (heterogeneous crumb rubber modified asphalt surface) and finite element (pavement structure) is established by using multi-scale coupling analysis method and setting gradation. The performance of actual pavement structure asphalt sample is measured under different pavement bearing modes. The internal swelling-dissolution state of surface crumb rubber modified asphalt is obtained by using the neural network prediction model of crumb rubber modified asphalt performance and crumb rubber modified asphalt swelling-dissolution state. The fine modeling in discrete element is realized. The permanent deformation behavior of pavement structure under repeated load is deduced. The performance of crumb rubber modified asphalt pavement structure is prewarned.

Claims

1. A method for early warning of the performance of rubber powder asphalt pavement based on the swelling-dissolution state of rubber powder, characterized in that, Includes the following steps: S1. Preparation of rubber-modified asphalt and acquisition of asphalt mastic material property parameters: In step S1, firstly, a wet method is used with a rubber powder content of 5-35%, a treatment temperature of 140-195℃, and a treatment time of 0.5-3.0h to conduct orthogonal rubber powder swelling-dissolution tests, resulting in the preparation of 180 sets of rubber powder modified asphalt. Secondly, the rubber powder modified asphalt is cured in epoxy resin to form resin samples, preparing 180 sets of rubber powder modified asphalt mastic samples. Finally, nano-indentation tests are performed on all samples at the microscale to extract the elastic modulus and viscoelasticity of the rubber powder phase and the asphalt phase. S2. Based on macro-micro experiments, obtain the performance indicators of rubber powder modified asphalt and the parameters of rubber powder particle swelling-dissolution state changes; S3. Numericalization of the property parameters of rubber powder modified asphalt materials and the parameters of the swelling-dissolution state change of rubber powder particles; In S3, for all samples, firstly, for the dual-phase rubber-modified asphalt material, the elastic modulus of each phase is... X It follows a Gaussian model, that is ; Obtain the frequency histogram of the elastic modulus of the rubber powder particles and the frequency histogram of the elastic modulus of the asphalt phase; Calculate the probability density function of the Gaussian mixture model according to equations (7)-(8) to obtain the Gaussian mixture model of the elastic modulus of all rubber powder modified asphalt; Secondly, the swelling-dissolution state change parameters of the rubber powder particles. Y It also follows a Gaussian model, that is ; Obtain the volume expansion rate of the adhesive powder particles V R The frequency histogram of mass decay rate T is used to calculate the probability density function of the Gaussian mixture model according to equations (7)-(8), and the swelling characteristic parameters of rubber particles in the corresponding number of rubber-modified asphalt samples are obtained by the Gaussian mixture model; the material property parameters of rubber-modified asphalt and the swelling-dissolution state change parameters of rubber particles are realized. （7） in: Let be the probability density function of the Gaussian mixture model; （8） x For sample data, μ k The mean of the data. The standard deviation of the data; S4. Construct a predictive model for the performance and swelling-dissolution state of rubber powder modified asphalt based on neural networks; S5. Modeling and micro-parameter assignment of heterogeneous rubber powder modified asphalt mastic; In S5, when assigning micromechanical parameters, the macroscopic performance parameters of all particles calculated by the nanoindentation test in S1 are first compared with those of S1. The conversion calculation between macroscopic performance parameters and micromechanical parameters is performed according to equations (11)-(12) and the linear contact model and Burgers model. （11） （12） In the formula: A is the contact area. L This is the length of the viscoelastic beam, i.e., the distance between the centers of adjacent particle elements. E For elastic modulus, v Poisson's ratio, G Shear modulus; S6. A method for early warning of the performance of rubber powder modified asphalt pavement based on the swelling-dissolution state of rubber powder using a neural network algorithm.

2. The method for early warning of rubber powder asphalt pavement performance based on the swelling-dissolution state of rubber powder according to claim 1, characterized in that, The specific steps for extracting the elastic modulus and viscoelastic indices of the rubber powder phase and the asphalt phase by performing nanoindentation tests on all samples at the microscale are as follows: First, nanoindentation tests are performed on the rubber powder phase in different rubber powder modified asphalt mastic samples to obtain the nanoindentation load-depth curves of rubber powder particles in all different swelling-dissolution states. The unloading segment of the curve is fitted using the Oliver-Fal model, and the elastic modulus of all rubber powder particles is calculated according to equations (1)-(2). Second, nanoindentation tests are performed on the asphalt phase, and the elastic modulus of the asphalt phase is calculated according to equations (1)-(2). Under the selected asphalt constitutive model as the Burgers viscoelastic model, the creep process of asphalt, i.e. the loading segment of the test, is obtained, and the viscoelastic parameters of the asphalt particles are calculated according to equations (3)-(4): spring element modulus in the Maxwell model. E 1 and viscosity of the sticky pot unit The spring element modulus in the Kelvin model E 2 and viscosity of the sticky pot unit : （1） （2） In the formula: E r Let S be the composite elastic modulus of the indenter and the tested sample, and let S be the slope of the top of the unloading section curve. β Here, is the geometric correction factor for the indenter, and A is the contact area. E s The elastic modulus of the tested sample, v s For the Poisson ratio of the tested sample, E i For the elastic modulus of the indenter, v i The Poisson's ratio of the pressure head; （3） （4） In the formula: h ( t () represents the indentation depth. P max For the maximum indentation load, θ The half-open angle of the pressure head, t For time, E 1 represents the modulus of the spring element in the Maxwell model. Here is the viscosity of the sticky pot element in the Maxwell model. E 2 represents the modulus of the spring element in the Kelvin model. This represents the viscosity of the sticky pot unit in the Kelvin model.

3. The method for early warning of rubber powder asphalt pavement performance based on the swelling-dissolution state of rubber powder according to claim 1, characterized in that: In step S2, firstly, according to the PG grading test specifications, the flash point, viscosity, and rheological properties of each group of asphalt are determined. Quality change Rheological properties after PAV 1. Modulus, stiffness S, and failure strain; 2. For the prepared rubber powder modified asphalt, due to different preparation processes, the rubber powder particles exist in different states inside the asphalt. The swelling-dissolution state of the rubber powder particles is analyzed by reflux elution. Toluene is used as the eluent to completely elute the asphalt. The samples are placed under X-Ray CT for scanning experiments to obtain the scanning images of each group of samples. The scanning images are converted into grayscale images using Matlab. The histogram equalization function is used to enhance the contrast of the images. The watershed segmentation method is used to completely separate the images of asphalt and rubber powder particles. Then, the volume expansion rate of the rubber powder particles under all different swelling-dissolution states is measured and calculated according to formulas (5)-(6). V R and mass decay rate T; （5） （6） In the formula: V R The volume expansion rate of the adhesive powder particles. R 1 represents the average diameter of the swollen adhesive powder. R 0 represents the average diameter of the unswollen adhesive powder; T is the mass decay rate. M The mass of the asphalt sample modified with rubber powder before elution. β This refers to the amount of adhesive powder added. m The quality of the eluted adhesive powder.

4. The method for early warning of rubber powder asphalt pavement performance based on the swelling-dissolution state of rubber powder according to claim 1, characterized in that... In S4, the various performance indicators of rubber-modified asphalt are used as inputs, and the corresponding state data of rubber-modified asphalt are used as outputs. A prediction model is established using a three-layer multi-input single-output BP network with one hidden layer. First, the training set of the model is built: the neural network toolbox in MATLAB is used to train the network. The prediction model first inputs the training sample data into the network, and uses the Gaussian mixture function as the activation function of the hidden layer neurons according to Equation (9), and the logarithmic function of the Gaussian mixture model as the activation function of the output layer neurons according to Equation (10). The network training function and performance function are determined, and the number of hidden layer neurons is initially set. Second, the network parameters are set. After the parameters are set, the network is trained. Finally, after the training is completed, the accuracy of the model's prediction of the distribution of rubber-modified asphalt is checked by inputting experimental data, and the distribution state is predicted by inputting various performance indicators of rubber-modified asphalt. （9） （10） in: x j For the first j One observation data, Let be the probability density function of the Gaussian mixture model. The probability of each sub-model occurring in the mixture model. The logarithmic function of a Gaussian mixture model.

5. The method for early warning of rubber powder asphalt pavement performance based on the swelling-dissolution state of rubber powder according to claim 1, characterized in that, In step S5, firstly, within the PFC software, for asphalt modified with different amounts of rubber powder, a regular arrangement of particles is adopted, and the particle size is kept consistent to generate all particles. Combined with the random generation of rubber powder particles, the pointer is used to traverse and loop. Then, according to the non-uniform coefficient distribution elastic modulus mixed Gaussian model (7)-(8) established in step S3, a random number between 0 and 1 is set. To determine the elastic modulus parameter, different color indices are assigned to the rubber powder and asphalt particles respectively to achieve particle grouping and visualization until the rubber powder content is reached, thus constructing an initial model of heterogeneous rubber powder modified asphalt. Secondly, nested loops are set to traverse and search all rubber powder particle pointers, and then, according to the non-uniform coefficient distribution elastic modulus mixed Gaussian model (7)-(8) established in step S3, random numbers within 0-1 are set. The swelling characteristic parameters of the rubber powder particles are used to realize the establishment of the rubber powder modified asphalt model based on the swelling-dissolution state of the rubber powder; each contact pointer is read for all particles, and the particle type of this contact connection is determined according to the color index of the rubber powder and asphalt particles; the contact between rubber powder particles adopts the linear contact model, and the contact between asphalt particles and between asphalt particles and rubber powder particles adopts the Burgers model to give the constitutive model. For each particle, the macroscopic performance parameters and microscopic mechanical parameters are converted and calculated according to Equation (11)-(12) and the linear contact model and Burgers model, and the microscopic parameters are assigned in the software. Finally, the numerical simulation of non-homogeneous rubber powder modified asphalt under different swelling-dissolution states of all rubber powder particles is realized.

6. The method for early warning of rubber powder asphalt pavement performance based on the swelling-dissolution state of rubber powder according to claim 1, characterized in that, In S6, a multi-scale coupling analysis method is adopted to establish a macro-micro coupled pavement structure model according to a set gradation. Under different pavement load modes, the performance of actual pavement structure asphalt samples is measured. Using a neural network prediction model of the performance of rubber-modified asphalt and the swelling-dissolution state of the rubber paste, the internal swelling-dissolution state of the surface layer rubber-modified asphalt is obtained. Refined modeling is achieved in discrete element method, and the permanent deformation behavior of the pavement structure under repeated load is deduced, so as to realize early warning of the performance of rubber-modified asphalt pavement structure.

Images