Method for predicting asphalt pavement compactness based on optimal index of random forest model

Abstract

The application discloses a kind of optimal index prediction asphalt pavement compactness method based on random forest model, comprising: collecting continuous vibration signal, temperature information and compactness information under continuous vibration;The stiffness and damping coefficient of asphalt concrete pavement under vibration are calculated under different temperatures and different compactness in the continuous vibration of vibrating road roller, and the algebraic relationship between stiffness and damping parameters and temperature, compactness under continuous vibration is established;The continuous vibration signal of vibrating road roller under different types of working conditions is calculated, and the continuous vibration signal under different types of working conditions is converted into compactness quality evaluation index;The correlation coefficient between compactness quality evaluation index and corresponding compactness is calculated, and the optimal index under corresponding working condition is found;Random forest model is trained and tested, and the optimal index prediction model of compactness under different working conditions is obtained, and the optimal index found by the model is used to predict compactness.The application can effectively realize intelligent monitoring to asphalt pavement compactness.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent monitoring of pavement compaction degree, in particular to a method for predicting asphalt pavement compaction degree based on optimal indicators of a random forest model. BACKGROUND

[0002] The asphalt pavement has a wide range of applications due to a short construction period, good flatness, good driving comfort, small driving noise and convenient maintenance. In the construction process of the asphalt pavement, compaction is the last process, and the compaction affects the compaction degree of the asphalt pavement. In the process of compaction, the compaction degree is one of the strictly controlled indicators, and has an important influence on the construction quality, service performance and service life of the asphalt pavement, and determines the water damage resistance, anti-rutting ability and durability of the asphalt pavement to some extent. Therefore, real-time monitoring of the pavement compaction degree has attracted attention and attention of road researchers.

[0003] The traditional method for detecting the compaction degree has many deficiencies. For example, when the sand filling method and the core sampling method are used to detect the compaction degree, the randomness is large and the detection result has hysteresis, and an accurate compaction degree measurement result cannot be obtained in real time. In the field of intelligent compaction, most intelligent compaction products on the market still rely on the number of rolling times or a single compaction degree indicator to evaluate the compaction degree. The former can roughly estimate the compaction degree based on the number of rolling times, but since there are many factors affecting the compaction degree, this method has a large error and is not suitable for use. The latter uses the same compaction degree indicator for all compaction conditions, which also produces many errors and even incorrect prediction conclusions, causing great losses in real life and production. At present, the machine learning method is used to intelligently predict the compaction degree. This method improves the deficiencies of the traditional method, but it needs a large number of field construction tests to obtain a large amount of data for database establishment, which wastes a large amount of manpower and material resources.

[0004] Therefore, it is a necessary means to realize the intelligent detection technology of low cost, high efficiency and high accuracy to use simulation combined with machine learning, consider multiple compaction degree indicators, develop an efficient algorithm to analyze and data mine these huge data sets, and finally find the optimal compaction degree indicator under different conditions to predict the compaction degree. SUMMARY

[0005] The present application overcomes the shortcomings and deficiencies of the prior art, and provides a method for predicting asphalt pavement compaction degree based on optimal indicators of a random forest model, which can effectively realize intelligent monitoring of the asphalt pavement compaction degree, and has lower cost, higher efficiency and higher accuracy than the existing intelligent compaction technology.

[0006] To achieve the above object, the technical scheme provided by the present application is as follows: a method for predicting asphalt pavement compactness based on optimal indexes of a random forest model, comprising the following steps:

[0007] 1) receiving continuous vibration signals collected by a vibratory roller with different mechanical parameters under different rolling passes, temperature information under continuous vibration, and compactness information under continuous vibration;

[0008] 2) calculating the stiffness and damping coefficient of a vibrated asphalt concrete pavement under different temperatures and different compactnesses in the continuous vibration of the vibratory roller according to the received continuous vibration signals through a two-degree-of-freedom viscoelastic model;

[0009] 3) respectively establishing an algebraic relationship between the stiffness and damping parameters of the vibrated asphalt concrete pavement and the temperature and compactness under continuous vibration;

[0010] 4) establishing eight different types of working conditions, i.e., high-temperature high-frequency high-amplitude, low-temperature high-frequency high-amplitude, high-temperature low-frequency high-amplitude, high-temperature high-frequency low-amplitude, low-temperature low-frequency high-amplitude, low-temperature high-frequency low-amplitude, high-temperature low-frequency low-amplitude, and low-temperature low-frequency low-amplitude, according to the different temperatures and mechanical parameters of the vibratory roller, and calculating the continuous vibration signals of the vibratory roller under the eight different types of working conditions through the algebraic relationship of step 3) and the two-degree-of-freedom viscoelastic model;

[0011] 5) converting the continuous vibration signals under different types of working conditions into compactness quality evaluation indexes;

[0012] 6) calculating the correlation coefficient between the compactness quality evaluation indexes and the corresponding compactness, and finding the compactness quality evaluation index corresponding to the maximum correlation coefficient under different types of working conditions as the optimal index under the corresponding working condition;

[0013] 7) taking the category of the compactness quality evaluation index as a feature value, labeling the mechanical parameters of the vibratory roller and the physical parameters of the asphalt concrete, combining the extracted feature value and the label into a database, randomly dividing the data in the database into a training set and a test set, inputting the training set into a random forest model for machine learning, detecting the accuracy of the trained random forest model with the test set, obtaining the compactness optimal index prediction model under different types of working conditions, and predicting the compactness with the optimal index found by the model, thereby effectively improving the accuracy of predicting the compactness.

[0014] Further, in step 1), the mechanical parameters of the vibratory roller refer to the vibration frequency and vibration amplitude of the vibratory roller vibratory wheel, the continuous vibration signal refers to the continuous acceleration data measured by the acceleration sensor installed at the wheel shaft of the vibratory roller vibratory wheel, the temperature data measured by the infrared temperature sensor installed at the wheel shaft position on both sides of the vibratory wheel, and the compaction degree data obtained by the real-time measurement of the nuclear density gauge.

[0015] Further, in step 2), according to the vibration mechanics theory, the vibratory roller and the asphalt concrete pavement system are simplified and simulated into a two-degree-of-freedom viscoelastic model, the upper frame m1 is connected with the vibratory wheel m2 through a spring damping system, the stiffness of the spring is k1, and the damping coefficient of the damper is c1, and the vibrated asphalt concrete m3 is also connected with the ground through a spring damping system, the stiffness of the spring is k2, and the damping coefficient of the damper is c2; the two-degree-of-freedom viscoelastic model is simulated by using the simulink software, and the stiffness k2 and the damping coefficient c2 of the spring damping system between the vibrated asphalt concrete and the ground can be calculated according to the obtained acceleration data at different temperatures and compaction degrees;

[0016] The vibration mechanics equations of the two-degree-of-freedom viscoelastic model are shown in the following formulas (1), (2), (3), (4), formula (5) is the specific expansion of the mass matrix M, formula (6) is the specific expansion of the damping matrix C, and formula (7) is the specific expansion of the stiffness matrix K obtained by simulating the two-degree-of-freedom viscoelastic model by using the simulink software:

[0017]

[0018]

[0019] F0=M e ω 2 (3)

[0020]

[0021] Acceleration matrix:

[0022] Velocity matrix:

[0023] Displacement matrix:

[0024] Mass matrix:

[0025] Damping matrix:

[0026] Stiffness matrix:

[0027] wherein F0 is the exciting force; ω is the exciting angular frequency; M e is the static eccentric moment of the eccentric block; and x1 represent the acceleration, velocity and displacement of the upper frame, respectively; and x2 represent the acceleration, velocity and displacement of the center of the vibrating wheel, respectively; and X represent the acceleration matrix, velocity matrix and displacement matrix, respectively; F(t) is a function of the load varying with time.

[0028] Further, in step 3), an exponential function is used to respectively fit the algebraic relationship between the stiffness k2 and the damping coefficient c2 of the spring-damper system between the vibrated asphalt concrete and the ground and the temperature and the compaction degree under continuous vibration.

[0029] Further, in step 4), the specific steps for calculating the continuous vibration signal are as follows:

[0030] 4.1) According to the consideration of the actual situation of the compaction operation, 120-170℃ is defined as a high temperature condition, 80-120℃ is defined as a low temperature condition; 40-60Hz is defined as a high frequency condition, 20-40Hz is defined as a low frequency condition; 5-8mm is defined as a high amplitude condition, and 3-5mm is defined as a low amplitude condition;

[0031] 4.2) The set temperature and compaction degree are input into the algebraic expression obtained in step 3) to obtain the stiffness k2 and the damping coefficient c2 of the spring-damper system, and in this process, one type of working condition is set as n groups of data, and a total of 8n groups of data are obtained;

[0032] 4.3) The obtained stiffness k2 and damping coefficient c2 of the spring-damper system are input into the two-degree-of-freedom viscoelastic model established by simulink, and the acceleration signal is calculated to obtain the continuous vibration acceleration signal of the vibratory roller under 8 different types of working conditions.

[0033] Further, in step 5), the continuous vibration signals under different types of working conditions are converted into compaction quality evaluation indexes CMV, CCV, K S , E vib , VCV, Omega, MDP, E, wherein CMV is a compaction meter value, CCV is a compaction control value, the acceleration signal is subjected to fast Fourier transform through the powergui module of the simulink software, and then combined with the corresponding formula to calculate, as shown in the following formulas (8) and (9):

[0034]

[0035]

[0036] wherein A 0.5ΩA represents the amplitude of the 0.5th harmonic component Ω A represents the amplitude of the 1st harmonic component 1.5Ω A represents the amplitude of the 1.5th harmonic component 2Ω A represents the amplitude of the 2nd harmonic component 2.5Ω A represents the amplitude of the 2.5th harmonic component 3Ω A represents the amplitude of the 3rd harmonic component

[0037] K S and E vib are indicators of stiffness and modulus, respectively; wherein K S represents the soil stiffness coefficient, which is calculated as shown in the following formula (10):

[0038]

[0039] In the formula, is the phase lag angle between the eccentric force and the displacement of the vibratory drum; Omega is the excitation frequency of the vibratory roller; m d is the mass of the vibratory drum; e0 is the eccentric moment representing the center of the vibratory drum;

[0040] E vib is calculated as shown in the following formulas (11) and (12):

[0041]

[0042]

[0043] In the formula, b is the width of the vibratory drum; R is the radius of the vibratory drum; E vib is the vibratory modulus; v is the Poisson's ratio; and delta has no specific meaning and is only used as a parameter for calculation to avoid complexity of formula (11);

[0044] VCV represents the vibratory compaction value, which is a mechanical indicator and is a function of force; and Omega is an energy indicator, which is calculated as shown in the following formula (13):

[0045]

[0046] In the formula, m d , m f are the masses of the vibratory drum and the frame, respectively; g represents the acceleration of gravity; and T represents a working period.

[0047] E is an energy indicator, which is calculated as shown in the following formula (14):

[0048]

[0049] In the formula, E represents the total compaction power per unit volume; A is the amplitude of the vibrating wheel; w is the vertical load of the vibrating wheel; f is the exciting frequency; N is the number of rolling passes; h is the rolling layer thickness; and v' is the driving speed.

[0050] MDP is an energy index, and the specific calculation of MDP is shown in the following formula (15):

[0051]

[0052] In the formula, MDP represents the mechanical driving power; P g is the total driving power of the road roller; W is the gravity of the road roller; V is the walking speed of the road roller; a is the climbing angle; and m and c are the mechanical energy loss rates.

[0053] Further, in step 6), the correlation coefficient between the compaction quality evaluation index and the corresponding compaction degree is calculated by using the Pearson product-moment correlation coefficient, as shown in the following formula (16):

[0054]

[0055] In the formula, ρ X,Y is the Pearson product-moment correlation coefficient, X is the continuous compaction index, Y is the compaction degree measured by the coreless density, cov(X, Y) is the covariance of X and Y, σ X is the variance of X, and σ Y is the variance of Y.

[0056] The compaction quality evaluation index corresponding to the maximum correlation coefficient under different types of working conditions is found, and is defined as the optimal index under the corresponding working condition to effectively predict the compaction degree.

[0057] Further, in step 7), the categories of the compaction quality evaluation index are taken as the feature values, and after the feature values are obtained, since each group of working conditions corresponds to a group of labels, including the vibration amplitude of the road roller, the vibration frequency of the road roller, the temperature, the stiffness and damping coefficient of the vibrated asphalt concrete pavement, the continuous vibration acceleration, and the number of rolling passes, and each group of working conditions corresponds to a group of feature values, i.e., the categories of the compaction quality evaluation index; therefore, the feature values and the labels are combined together one by one, so that the required data set can be obtained; then the data set is randomly divided into a training set and a test set according to a set proportion, wherein the training set is used to provide data for the random forest model to learn, so as to obtain the required machine learning prediction model, i.e., the trained random forest model, and the test set is used to judge the accuracy of the generated machine learning prediction model, and the optimal index found by using the model is used to predict the compaction degree, so as to effectively improve the accuracy of predicting the compaction degree.

[0058] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0059] 1、The present application combines 4 types of 8 compaction quality evaluation indexes to comprehensively evaluate the compaction degree, and takes the index with the strongest correlation with the compaction degree as the optimal compaction quality evaluation index under this working condition, thereby greatly improving the accuracy of the prediction result and greatly improving the construction efficiency, and having strong commercial utilization value.

[0060] 2、The present application mainly adopts the random forest method, extracts characteristic values according to a large amount of calculation data under 8 different types of working conditions, and finally realizes that the optimal prediction index of the compaction degree can be found through different types of working condition parameters, greatly reduces the time spent on construction test, improves the efficiency, and also improves the prediction accuracy.

[0061] 3、The present application combines the vibration mechanics and the simulink simulation method, clearly and accurately simulates the vibratory road roller and the asphalt concrete pavement system, saves a lot of test cost, and improves the result precision.

[0062] 4、The compaction degree prediction model (i.e., the random forest model) of the present application considers various factors affecting the compaction process, including the physical parameters (damping, stiffness) of the asphalt concrete and the mechanical parameters (truck mass, damper stiffness, damper damping, static eccentricity, angular velocity) of the vibratory road roller, and therefore can be applied to different types of working conditions, has comprehensive characteristic values, has universality, and has strong practical significance. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 is a flowchart of the method of the present application.

[0064] Figure 2 is a simplified model diagram of the vibratory road roller-asphalt concrete vibration system.

[0065] Figure 3 is a compaction meter type index calculation principle diagram. DETAILED DESCRIPTION

[0066] The present application will be further described in detail below in combination with embodiments and drawings, but the embodiments of the present application are not limited thereto.

[0067] As shown in Figure 1 , the present embodiment discloses an optimal index prediction method for asphalt pavement compaction degree based on a random forest model, comprising the following steps:

[0068] 1) receiving continuous vibration signals collected by a vibratory roller with different mechanical parameters at different rolling passes, temperature information under continuous vibration, and compaction degree information under continuous vibration. The mechanical parameters of the vibratory roller refer to the vibration frequency and vibration amplitude of the vibratory roller vibratory wheel; the vibration frequency is found on the frequency spectrum obtained by fast Fourier transform of the acceleration signal, and the vibration amplitude is obtained by twice integration of the acceleration signal. The continuous vibration signal refers to the continuous acceleration data measured by the acceleration sensor installed at the wheel shaft of the vibratory roller, the temperature data measured by the infrared temperature sensor installed at the wheel shaft position on both sides of the vibratory wheel, and the compaction degree data obtained by real-time measurement of the nuclear density gauge.

[0069] 2) According to the received continuous vibration signal, the stiffness and damping parameters of the vibrated asphalt concrete pavement under different temperatures and different compaction degrees in the continuous vibration of the vibratory roller are calculated by a two-degree-of-freedom viscoelastic model. According to the vibration mechanics theory, the vibratory roller and the asphalt concrete pavement system are simplified and simulated into a two-degree-of-freedom viscoelastic model, as shown in Figure 2 The spring stiffness k1 and the damper damping coefficient c1 are connected between the upper frame m1 and the vibratory wheel m2, and the spring stiffness k2 and the damper damping coefficient c2 are connected between the vibrated asphalt concrete m3 and the ground. The two-degree-of-freedom viscoelastic model is simulated by simulink software, and the spring stiffness k2 and the damper damping coefficient c2 between the vibrated asphalt concrete and the pavement can be calculated according to the obtained acceleration data under different temperatures and compaction degrees. The specific vibration mechanics equations are shown in formulas (1), (2), (3), and (4), wherein: F0 is the exciting force; ω is the exciting angular frequency; M e is the static eccentric moment of the eccentric block; and x1 represent the acceleration, velocity and displacement of the upper frame, respectively; and x2 represent the acceleration, velocity and displacement of the center of the vibratory wheel, respectively. In formula (4), and X represent the acceleration matrix, velocity matrix and displacement matrix, respectively; F(t) is a function of load changing with time. Formula (5) is the specific expansion of the mass matrix M, formula (6) is the specific expansion of the damping matrix C, and formula (7) is the specific expansion of the stiffness matrix K obtained by simulating the two-degree-of-freedom viscoelastic model by simulink software.

[0070]

[0071]

[0072]

[0073] Acceleration matrix:

[0074] Velocity matrix:

[0075] Displacement matrix:

[0076] Mass matrix:

[0077] Damping matrix:

[0078] Stiffness matrix:

[0079] 3) Establishing the algebraic relationship between the stiffness and damping parameters of the vibrated asphalt concrete pavement and the temperature and compaction degree under continuous vibration, specifically, using an exponential function to respectively fit the algebraic relationship between the stiffness k2 and damping coefficient c2 of the spring-damper system of the vibrated asphalt concrete and the ground and the temperature and compaction degree under continuous vibration.

[0080] 4) According to the different temperatures and mechanical parameters of the vibratory roller, eight different types of working conditions are established, including high-temperature high-frequency high-amplitude, low-temperature high-frequency high-amplitude, high-temperature low-frequency high-amplitude, high-temperature high-frequency low-amplitude, low-temperature low-frequency high-amplitude, low-temperature high-frequency low-amplitude, high-temperature low-frequency low-amplitude, and low-temperature low-frequency low-amplitude. The continuous vibration signals of the vibratory roller under the eight different types of working conditions are calculated through the algebraic relationship and the two-degree-of-freedom viscoelastic model. The specific steps for calculating the continuous vibration signals are as follows:

[0081] 4.1) According to the actual situation of compaction operation, 120-170℃ is defined as the high-temperature condition, 80-120℃ is defined as the low-temperature condition, 40-60Hz is defined as the high-frequency condition, 20-40Hz is defined as the low-frequency condition, 5-8mm is defined as the high-amplitude condition, and 3-5mm is defined as the low-amplitude condition.

[0082] 4.2) The set temperature and compaction degree are input into the algebraic formula obtained in step 3) to obtain the stiffness k2 and damping coefficient c2 of the spring-damper system. In this process, one type of working condition is set as n groups of data, and a total of 8n groups of data are obtained.

[0083] 4.3) The obtained stiffness k2 and damping coefficient c2 of the spring-damper system are input into the two-degree-of-freedom viscoelastic model established by simulink, and the acceleration signal can be calculated to obtain the continuous vibration acceleration signal of the vibratory roller under the eight different types of working conditions.

[0084] 5) The continuous vibration signals under different types of working conditions are converted into compaction quality evaluation indicators CMV, CCV, K S , and E vib, VCV, Omega, MDP, E. Among them, CMV is the compaction meter value, and CCV is the compaction control value. The acceleration signal is subjected to fast Fourier transform (FFT) through the powergui module of the simulink software, and then combined with the formula to obtain, as shown in equations (8), (9): In the formula, A 0.5Ω represents the vibration amplitude of the 0.5th harmonic component, A Ω represents the vibration amplitude of the 1st harmonic component, A 1.5Ω represents the vibration amplitude of the 1.5th harmonic component, A 2Ω represents the vibration amplitude of the 2nd harmonic component, A 2.5Ω represents the vibration amplitude of the 2.5th harmonic component, A 3Ω represents the vibration amplitude of the 3rd harmonic component, and the specific calculation principle is shown in equation (7). Figure 3

[0085]

[0086]

[0087] K S and E vib are stiffness / modulus indicators. Among them, K S represents the soil stiffness coefficient, and the specific calculation process is shown in equation (10): In the formula, is the phase lag angle between the eccentric force and the displacement of the vibratory roller; Ω is the excitation frequency of the vibratory roller; m d is the mass of the vibratory roller; x2 is the displacement of the vibratory roller; e0 is the eccentric moment representing the center of the vibratory roller;

[0088]

[0089] E vib is the specific calculation process of equation (11), (12): In the formula, b is the width of the vibratory roller; R is the radius of the vibratory roller; E vib is the vibration modulus; v is the Poisson's ratio; δ has no specific meaning, and is only used as a parameter to participate in operation to avoid equation (11) being too complex.

[0090]

[0091]

[0092] VCV represents the vibratory compaction value, which is a mechanical indicator, and is a function of force. Omega is an energy indicator. The specific calculation process is shown in equation (13): In the formula, m d , m f ​respectively, the mass of the vibrating drum and the frame; g represents the acceleration of gravity; T represents a working cycle; e0is the eccentric moment representing the center of the vibrating drum; ω is the vibration angular frequency;

[0093]

[0094] E is an energy-based index. The specific calculation process is shown in equation (14): In the equation, E represents the total compaction power per unit volume; A is the amplitude of the vibrating drum; w is the vertical load of the vibrating drum; f is the excitation frequency; N is the number of compaction passes; h is the compaction layer thickness; v' is the driving speed;

[0095]

[0096] MDP is an energy-based index. The specific calculation process of MDP is shown in equation (15): In the equation, MDP represents the mechanical driving power; P g is the total driving power of the road roller; W is the gravity of the road roller; V is the walking speed of the road roller; a is the climbing angle; m, c are the mechanical energy loss rates of the road roller itself;

[0097]

[0098] 6) Calculate the correlation coefficient between the compaction quality evaluation index and the corresponding compaction degree, and find the compaction quality evaluation index corresponding to the maximum correlation coefficient under different types of working conditions. The correlation coefficient between the compaction quality evaluation index and the corresponding compaction degree is calculated using the Pearson product-moment correlation coefficient. The corresponding calculation formula (16) is as follows: In the equation, ρ X,Y is the Pearson product-moment correlation coefficient, X is the continuous compaction index, Y is the compaction degree measured by the non-nuclear density, cov(X, Y) is the covariance of X and Y, σ X is the variance of X, σ Y is the variance of Y.

[0099]

[0100] The compaction quality evaluation index corresponding to the maximum correlation coefficient under different types of working conditions is found, and it is defined as the optimal index under the corresponding working condition to effectively predict the compaction degree.

[0101] 7) The category of the compaction degree index is taken as a characteristic value, the mechanical parameters of the vibratory roller and the physical parameters of the asphalt concrete are labeled, the extracted characteristic value is combined with the label to form a database, the data in the database is randomly divided into a training set and a test set, the training set is input into a random forest model for machine learning, the accuracy of the trained random forest model is detected by the test set, and a compaction degree optimal index prediction model under different types of working conditions is obtained. Among them, the category of the compaction degree quality evaluation index is taken as a characteristic value, after obtaining the above characteristic value, since each group of working conditions corresponds to a group of labels, including the vibration amplitude of the roller, the vibration frequency of the roller, the temperature, the stiffness and damping coefficient of the vibrated asphalt concrete pavement, the continuous vibration acceleration, and the rolling times. Each group of working conditions corresponds to a group of characteristic values, i.e. the category of the compaction degree quality evaluation index. Therefore, the characteristic value and the label are combined one by one to obtain the required data set; then the data set is randomly divided into a training set and a test set according to 8:2, wherein the training set is used to provide data for the random forest model to learn, so as to obtain the required machine learning prediction model (i.e. the trained random forest model), and the test set is used to judge the accuracy of the generated machine learning prediction model. And use the optimal index found by this model to predict the compaction degree, thereby effectively improving the accuracy of predicting the compaction degree.

[0102] The above embodiments are the preferred embodiments of the present application, but the embodiments of the present application are not limited by the above embodiments, and any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present application are equivalent replacement methods, which are all included in the protection scope of the present application.

Claims

1. A method for predicting the compaction degree of asphalt pavement based on an optimal index of a random forest model, characterized in that, The method comprises the following steps: 1) receiving continuous vibration signals, temperature information under continuous vibration, and compaction degree information under continuous vibration of a vibratory roller with different mechanical parameters under different compaction passes; 2) calculating the stiffness and damping coefficient of the vibrated asphalt concrete pavement under different temperatures and different compaction degrees in the continuous vibration of the vibratory roller by a two-degree-of-freedom viscoelastic model according to the received continuous vibration signals; 3) using an exponential function to respectively fit the algebraic relationship between the stiffness k2 and the damping coefficient c2 of the spring-damper system between the vibrated asphalt concrete and the ground and the temperature and the compaction degree under continuous vibration; 4) establishing eight different types of working conditions, i.e., high-temperature high-frequency high-amplitude, low-temperature high-frequency high-amplitude, high-temperature low-frequency high-amplitude, high-temperature high-frequency low-amplitude, low-temperature low-frequency high-amplitude, low-temperature high-frequency low-amplitude, high-temperature low-frequency low-amplitude, and low-temperature low-frequency low-amplitude, according to different temperatures and mechanical parameters of the vibratory roller, and calculating the continuous vibration signals of the vibratory roller under the eight different types of working conditions by the algebraic relationship of step 3) and the two-degree-of-freedom viscoelastic model; The specific steps for calculating the continuous vibration signals are as follows: 4.1) according to the consideration of the actual situation of the compaction operation, 120-170℃ is defined as the high-temperature condition, 80-120℃ is defined as the low-temperature condition, 40-60Hz is defined as the high-frequency condition, 20-40Hz is defined as the low-frequency condition, 5-8mm is defined as the high-amplitude condition, and 3-5mm is defined as the low-amplitude condition; 4.2) inputting the set temperature and compaction degree into the algebraic expression obtained in step 3) to obtain the stiffness k2 and the damping coefficient c2 of the spring-damper system, in which process one type of working condition is set as n groups of data, and a total of 8n groups of data are obtained; 4.3) inputting the obtained stiffness k2 and damping coefficient c2 of the spring-damper system into the two-degree-of-freedom viscoelastic model established by simulink to calculate the acceleration signal, so as to obtain the continuous vibration acceleration signals of the vibratory roller under the eight different types of working conditions; 5) converting the continuous vibration signals under different types of working conditions into compaction degree quality evaluation indexes; 6) calculating the correlation coefficient between the compaction degree quality evaluation indexes and the corresponding compaction degrees, and finding the compaction degree quality evaluation index corresponding to the maximum correlation coefficient under different types of working conditions as the optimal index under the corresponding working condition; 7) taking the category of the compaction degree quality evaluation index as a characteristic value, labeling the mechanical parameters of the vibratory roller and the physical parameters of the asphalt concrete, combining the extracted characteristic value and the label into a database, randomly dividing the data in the database into a training set and a test set, inputting the training set into a random forest model for machine learning, detecting the accuracy of the trained random forest model with the test set, obtaining the compaction degree optimal index prediction model under different types of working conditions, and predicting the compaction degree with the optimal index found by the model, thereby effectively improving the accuracy of predicting the compaction degree.

2. The method for predicting the compaction degree of asphalt pavement based on the optimal index of the random forest model according to claim 1, characterized in that: In step 1), the mechanical parameters of the vibratory roller refer to the vibration frequency and amplitude of the vibratory roller vibratory wheel, the continuous vibration signal refers to the continuous acceleration data measured by the acceleration sensor installed at the wheel shaft of the vibratory roller vibratory wheel, the temperature data measured by the infrared temperature sensor installed at the wheel shaft position on both sides of the vibratory wheel, and the compaction degree data obtained by the real-time measurement of the nuclear density gauge.

3. The method for predicting the compaction degree of asphalt pavement based on the optimal index of the random forest model according to claim 2, characterized in that: In step 2), according to the vibration mechanics theory, the vibratory roller and the asphalt concrete pavement system are simplified and simulated into a two-degree-of-freedom viscoelastic model, the upper frame m1 is connected with the vibratory wheel m2 through a spring damping system, the stiffness of the spring is k1, and the damping coefficient of the damper is c1, and the vibrated asphalt concrete m3 is also connected with the ground through a spring damping system, the stiffness of the spring is k2, and the damping coefficient of the damper is c2; the two-degree-of-freedom viscoelastic model is simulated by using the simulink software, and the stiffness k2 and the damping coefficient c2 of the spring damping system between the vibrated asphalt concrete and the ground can be calculated according to the obtained acceleration data at different temperatures and compaction degrees; The vibration mechanics equations of the two-degree-of-freedom viscoelastic model are shown in the following formulas (1), (2), (3), (4), the specific expansion of the mass matrix M is shown in formula (5), the specific expansion of the damping matrix C is shown in formula (6), and the specific expansion of the stiffness matrix K obtained by simulating the two-degree-of-freedom viscoelastic model by using the simulink software is shown in formula (7): F0 = M e ω 2 (3) acceleration matrix: Velocity matrix: Displacement matrix: Mass matrix: Damping matrix: Stiffness matrix: where F0is the exciting force; ω is the exciting angular frequency; M e is the static eccentric moment of the eccentric block; and x1respectively represent the acceleration, velocity and displacement of the upper frame; and x2respectively represent the acceleration, velocity and displacement of the center of the vibrating wheel; and X respectively represent the acceleration matrix, velocity matrix and displacement matrix; F(t) is a function of the load varying with time.

4. The method for predicting the compaction degree of asphalt pavement based on the optimal index of the random forest model according to claim 3, characterized in that: In step 5), the continuous vibration signals under different types of working conditions are converted into compaction quality evaluation indexes CMV, CCV, K S vib VCV, Omega, MDP, E, wherein CMV is a compaction meter value, CCV is a compaction control value, the acceleration signal is subjected to fast Fourier transform through a powergui module of simulink software, and then combined with corresponding formulas to obtain, as shown in the following formulas (8) and (9):​ wherein A 0.5Ω represents the amplitude of the 0.5th harmonic component, A Ω represents the amplitude of the 1st harmonic component, A 1.5Ω represents the amplitude of the 1.5th harmonic component, A 2Ω represents the amplitude of the 2nd harmonic component, A 2.5Ω represents the amplitude of the 2.5th harmonic component, A 3Ω represents the amplitude of the 3rd harmonic component; K S and E vib are respectively the indicators of stiffness and modulus; wherein K S represents the soil stiffness coefficient, which is calculated as shown in equation (10): where φ is the phase lag angle between the eccentric force and the displacement of the vibratory drum; Ω is the exciting frequency of the vibratory roller; m d is the mass of the vibratory drum; e0is the eccentric moment representing the center of the vibratory drum; E vib The specific calculation is shown in the following equations (11) and (12): where b is the width of the vibrating wheel; R is the radius of the vibrating wheel; E vib is the modulus of vibration; v is the Poisson's ratio; δ has no specific meaning and is only a parameter involved in the calculation to avoid the complexity of equation (11); VCV represents the vibratory compaction value, which is a function of force, and Omega is an energy index, which is calculated as shown in formula (13): where m d , m f are the mass of the vibrating wheel and the vehicle body, respectively; g is the acceleration of gravity; and T is one working cycle. E is an energy index, which is calculated as shown in formula (14): In the formula, E represents the total compaction power per unit volume; A is the amplitude of the vibratory wheel; w is the vertical load of the vibratory wheel; f is the excitation frequency; N is the number of rolling passes; h is the rolling layer thickness; and v' is the driving speed; MDP is an energy index, and the specific calculation of MDP is shown in formula (15): In the formula, MDP represents the mechanical driving power; P g is the total driving power of the road roller; W is the weight of the road roller; V is the walking speed of the road roller; a is the climbing angle; and m and c are the mechanical energy loss rates of the road roller itself.

5. The method for predicting the compaction degree of asphalt pavement based on the optimal index of the random forest model according to claim 4, characterized in that: In step 6), the correlation coefficient between the compaction quality evaluation index and the corresponding compaction degree is calculated by using the Pearson product-moment correlation coefficient, as shown in formula (16): where p X′,Y is the Pearson's product-moment correlation coefficient, X' is the continuous compaction indicator, Y is the compaction degree measured by the non-nuclear density, cov(X', Y) is the covariance of X' and Y, σ X′ is the variance of X', and σ Y is the variance of Y. The compaction quality evaluation index corresponding to the maximum correlation coefficient under different types of working conditions is found, and it is defined as the optimal index under the corresponding working condition to effectively predict the compaction degree.

6. The method for predicting the compaction degree of asphalt pavement based on the optimal index of the random forest model according to claim 5, characterized in that: In step 7), the categories of the compaction quality evaluation indicators are taken as feature values, and after the feature values are obtained, since each group of working conditions corresponds to a group of labels, including the vibration amplitude of the road roller, the vibration frequency of the road roller, the temperature, the stiffness and damping coefficient of the vibrated asphalt concrete pavement, the continuous vibration acceleration and the rolling times; and each group of working conditions corresponds to a group of feature values, i.e., the categories of the compaction quality evaluation indicators; therefore, the feature values and the labels are combined together one by one, and the required data set can be obtained; then the data set is randomly divided into a training set and a test set according to a set proportion, wherein the training set provides data for the random forest model to learn, so as to obtain the required machine learning prediction model, i.e., the trained random forest model, and the test set is used to judge the accuracy of the generated machine learning prediction model, and the optimal indicator found by the model is used to predict the compaction degree, thereby effectively improving the accuracy of the predicted compaction degree.

Images