Asphalt mixture performance prediction method and system and storage medium

Abstract

The invention provides an asphalt mixture performance prediction method and system and a storage medium, and belongs to the technical field of asphalt mixture performance testing. Comprising the following steps: step 1, preparing an asphalt mixture sample, and carrying out preheating and heat preservation treatment on the sample; step 2, carrying out a load test on a pavement area where the sample is located to obtain axle load times x1 and void water action times x2; 3, according to the axle load frequency x1 and the void water action frequency x2, obtaining the cumulative bearing frequency X, and fitting a relational expression between the cumulative bearing frequency X and the rut depth; 4, constructing a neural network model, optimizing fitting parameters in the relational expression of the cumulative bearing times X and the rut depth until an iteration termination condition or the maximum iteration times is reached, outputting the trained fitting parameters, and substituting the trained fitting parameters into the relational expression of the cumulative bearing times X and the rut depth; and 5, predicting the rut depth by adopting the optimized relational expression between the cumulative bearing times X and the rut depth.

Description

Technical Field

[0001] The present invention relates to the technical field of asphalt mixture performance testing, and in particular to an asphalt mixture performance prediction method, system and storage medium. Background Art

[0002] Rutting and water damage of asphalt pavement reduce its performance and service life, thus increasing the cost of repair and maintenance. Currently, the rutting test, submerged Marshall test, freeze-thaw splitting test and Hamburger rutting test are commonly used to evaluate the anti-rutting and anti-water damage performance of asphalt mixtures. Compared with other test methods, the submerged Hamburger rutting test can simultaneously evaluate the anti-rutting and anti-water damage performance of asphalt mixtures.

[0003] However, distinguishing the confounding effects of load and moisture in the submerged hamburger rutting test is a key issue in characterizing the properties of the mixture using this test. The existing solution is to conduct the hamburger rutting test under both dry and submerged conditions to separate the deformation caused by viscoplastic flow and that caused by moisture in the test. This method can separate the effect of the rutting response of the dry hamburger rutting test from the rutting response curve of the submerged hamburger rutting test, thereby determining the deformation caused by moisture. Although this method is feasible, conducting dry and submerged hamburger rutting tests doubles the testing workload. Therefore, it is urgent to develop a new analytical method for evaluating the rutting resistance and water damage resistance of the specimen. Summary of the invention

[0004] In view of this, the present invention proposes an asphalt mixture performance prediction method, system and storage medium capable of distinguishing the anti-rutting performance and anti-water damage performance of the asphalt mixture.

[0005] The technical solution of the present invention is achieved in this way:

[0006] In one aspect, the present invention provides a method for predicting asphalt mixture performance, comprising the following steps:

[0007] Step 1: Prepare asphalt mixture samples and preheat and insulate the samples;

[0008] Step 2: Carry out a load test on the pavement area where the sample is located. During the time period when the specified rutting depth is reached, record the variables including the equivalent load conversion value borne by the asphalt mixture during each load test, each load time, slope angle, daily average temperature, pore water pressure, sample length, and sample water content, and obtain the axle load times x1 and pore water action times x2;

[0009] Step 3: According to the axle load times x1 and the pore water action times x2, the cumulative load times X is obtained, and the relationship between the cumulative load times X and the rutting depth is fitted;

[0010] Step 4: Build a neural network model, optimize the fitting parameters in the relationship between the cumulative load times X and the rutting depth, until the iteration termination condition or the maximum number of iterations is reached, and output the trained fitting parameters to substitute into the relationship between the cumulative load times X and the rutting depth;

[0011] Step 5: Use the relationship between the cumulative load times X and the rutting depth after optimizing the fitting parameters, and predict the rutting depth under the corresponding axle load times and pore water action times according to the input parameters.

[0012] On the basis of the above technical scheme, preferably, the preparation of the asphalt mixture sample in step 1 is to design the mineral aggregate gradation according to the asphalt-stone ratio of 4.3-4.9%, wherein the average density of the mineral aggregate is 2.78 g / cm 3 ; Filling in a rigid metal box with a density of 88%-90% to make a sample, the sample is prepared by keeping warm at 60 degrees Celsius for more than 2 hours.

[0013] Preferably, in step 2, the equivalent load conversion value borne by the asphalt mixture, each load time, slope angle, daily average temperature, pore water pressure, sample length, and sample water content are recorded during each load test to obtain the axle load times x1 and pore water action times x2, respectively, by constructing the following relationship: Where n0 is the wave speed; p is the ground pressure in MPa; v is the vehicle speed in m / s; t1 is the duration of each load; P E is the equivalent load conversion value of each load, unit is MPa; θ is the slope angle; T is the daily average temperature; l is a non-zero constant; P N is the pore water pressure, unit is MPa; t2 is the action time of the pore water pressure; L is the length of the sample; s is the adjustment coefficient; γ% is the moisture content of the sample, and the moisture content of the sample is greater than zero.

[0014] Further preferably, the value of the adjustment coefficient s is related to the average monthly precipitation at the location of the sample. If the average monthly precipitation is lower than 30, the value of s is (0.8, 1); if the average monthly precipitation is higher than 30, the value of s is [1, 1.2).

[0015] Further preferably, in step 3, the cumulative load times X is obtained according to the axial load times x1 and the pore water action times x2, by setting X=λ1Σx1+λ2Σx2, wherein λ1 and λ2 are weights, and the values ​​of λ1 and λ2 are both (0, 1).

[0016] More preferably, the relationship between the cumulative load times X and the rutting depth is fitted by setting the rutting depth to f(X), and then: Where k1 is the first threshold; k2 is the second threshold; a, b, c, d, E and F iN are fitting parameters, i=1, 2, 3, 4, 5, 6.

[0017] Further preferably, the first threshold k1 corresponds to the cumulative load times of 94% density of the asphalt mixture; the second threshold k2 corresponds to the cumulative load times of 96% density of the asphalt mixture; the calculation formula of the density G is: Where m0 is the mass of asphalt mixture; A is the oil-stone ratio; V1 is the volume of asphalt before mixing; V2 is the volume of mineral aggregate before mixing; ρ0 is the theoretical maximum density of asphalt mixture.

[0018] On the basis of the above technical solution, preferably, constructing the neural network model in step 4 is to construct a neural network with at least three hidden layer structures, and the activation function of the neural network model is a sigmoidal function.

[0019] On the other hand, the present invention also provides an asphalt mixture performance prediction system, comprising:

[0020] The sample forming module is used to prepare asphalt mixture samples and preheat and insulate the samples;

[0021] The load test module is used to perform a load test on the pavement area where the specimen is located until the specified rutting depth is reached;

[0022] The data recording module is used to record the equivalent load conversion value of the asphalt mixture, each load time, slope angle, daily average temperature, pore water pressure, sample length, and sample water content during each load test within the time period of reaching the specified rutting depth; obtain the axle load times x1 and pore water action times x2;

[0023] The depth fitting module is used to obtain the axle load times x1 and the pore water action times x2 according to the contents of each variable obtained by the data recording module, and further obtain the cumulative load times X; fit the relationship between the cumulative load times X and the rutting depth;

[0024] Training module: construct a neural network model, optimize the fitting parameters in the relationship between the cumulative load times X and the rutting depth, until the iteration termination condition or the maximum number of iterations is reached, and output the trained fitting parameters to substitute into the relationship between the cumulative load times X and the rutting depth;

[0025] The performance prediction module uses the relationship between the cumulative load times X and the rutting depth after optimizing the fitting parameters, and predicts the rutting depth under the corresponding axle load times and pore water action times according to the input parameters.

[0026] In a third aspect, the present invention further provides a computer-readable storage medium, when a computer program stored in the storage medium is run on a computer, the computer is enabled to execute the above-mentioned asphalt mixture performance prediction method.

[0027] The asphalt mixture performance prediction method, system and storage medium provided by the present invention have the following beneficial effects compared with the prior art:

[0028] (1) The present application converts and accumulates the axle load times and the pore water action times respectively to obtain the cumulative load times corresponding to the standard test state, thereby improving the scope of application of the present method. Regardless of how the environment changes, it can always be converted to the standard test state, thereby providing a reference benchmark for subsequent rutting depth prediction, without the need to conduct separate load tests and pore water action tests;

[0029] (2) The relationship between the cumulative number of loads X and the rutting depth is provided, taking into account the initial state, partial compaction state, and fully compacted state of the asphalt mixture, and setting the threshold value corresponding to the density state, thereby limiting the rutting depth relationship in different intervals, making the rutting conversion relationship more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0031] Figure 1 The present invention is a flowchart of the steps of an asphalt mixture performance prediction method, system and storage medium. DETAILED DESCRIPTION

[0032] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] like Figure 1 As shown, the present invention provides a method for predicting asphalt mixture performance, comprising the following steps:

[0034] Step 1: Prepare asphalt mixture samples and preheat and insulate the samples.

[0035] In this step, asphalt mixture samples were prepared, and the grading of the mineral material was designed according to the asphalt-stone ratio of 4.3-4.9%, wherein the average density of the mineral material was 2.78 g / cm 3 ; Asphalt can be base asphalt or modified asphalt. The preparation of the sample is to pour the asphalt mixture into a rigid metal box and fill it. The density is 88%-90% to make the sample. The sample is kept warm at 60 degrees Celsius for more than 2 hours. There is no need to fully compact the sample. A foundation pit can be dug in the road surface in advance, and the sample can be filled into the foundation pit and kept flush with the road surface. The surface of the sample can be covered with a protective layer that isolates air or water, which can reduce the aging degree of the sample and improve the strength and stability of the sample.

[0036] Asphalt mixture adopts a balanced design method, and its goal is to balance the performance of asphalt mixture, that is, according to the needs of different application scenarios, a demand matrix of asphalt mixture performance under specific use conditions is proposed, and suitable asphalt is selected and mixed with minerals of different particle sizes. The preferred range of the asphalt mixture oil-stone ratio in this embodiment is 4.3-4.9%, and the preferred value is 4.4% determined by the Marshall test. The initial mixing temperature of the sample is 140-150 degrees Celsius. Each layer of the sample is initially compacted, and then another layer is spread until the sample preparation is completed. The thickness of the sample spread each time does not exceed 20cm. The cooling rate of the spread asphalt mixture does not exceed 3 degrees Celsius per minute. The mineral material of this scheme is preferably a mixture of limestone aggregate and limestone powder.

[0037] The road surface structure is exposed to the air environment and will be affected by the combined effects of accumulated water, air, sunlight and vehicle load. Rutting can cause serious damage to asphalt pavement, especially long and steep slopes, so this application pays special attention to the relevant parameters of the number of times the load acts on the road surface. Since the speed, load, wheelbase or wheelbase of different vehicles vary greatly, it is difficult to count and directly accumulate them separately, so a virtual benchmark is used to unify the status of all vehicles and convert them into the number of times the load acts under the virtual benchmark, which is convenient for subsequent accumulation and calculation. The virtual benchmark is preferably a speed of 10m / s, a wave speed of 1 time / second, a ground pressure of 100,000Pa-1MPa, and a single load duration of 0.1 second.

[0038] Step 2: Carry out a load test on the pavement area where the sample is located. During the time period when the specified rutting depth is reached, record the variables including the equivalent load conversion value borne by the asphalt mixture during each load test, each load time, slope angle, daily average temperature, pore water pressure, sample length, and sample moisture content to obtain the axle load times x1 and pore water action times x2.

[0039] The axial load times x1 and interstitial water action times x2 obtained in step 2 are constructed by respectively constructing the following relationship: Where n0 is the wave speed; p is the ground pressure in MPa; υ is the vehicle speed in m / s; t1 is the duration of each load; P E is the equivalent load conversion value of each load, unit is MPa; θ is the slope angle, the value range is [-45°, 45°]; T is the daily average temperature; l is a non-zero constant; P N is the pore water pressure, in megapascals; t2 is the action time of the pore water pressure; L is the length of the sample; S is the adjustment coefficient; γ% is the percentage of water content in the sample, and the percentage of water content in the sample is greater than zero. It can be seen that the axle load times x1 in the above formula are related to the parameters such as the number of ground loading times, vehicle speed, equivalent load, slope and daily average temperature related to vehicle driving, because factors such as temperature, driving speed and load duration have a significant impact on the permanent deformation of asphalt mixtures; the pore water action times x2 are related to the pore water pressure, water pressure action time and water content of asphalt mixtures. When the air is humid, the sample is immersed in water, or the voids inside the sample are filled with water, it will accelerate the peeling of the material on the surface or inside of the mixture. In order to avoid the numerator taking the value of 0 when the daily average temperature is 0 degrees Celsius or 60 degrees Celsius, resulting in the axle load times x1 being zero, a non-zero constant l is introduced to avoid this situation. When the same vehicle passes the sample, it may only affect the axle load times, or it may also affect the axle load times and the pore water action times at the same time.

[0040] The adjustment coefficient s is not fixed. Its value is related to the average monthly precipitation at the sample location. If the average monthly precipitation is less than 30, the value range of s is relatively low, i.e. (0.8, 1); if the average monthly precipitation is higher than 30, the value range of s is relatively high, i.e. [1, 1.2).

[0041] Step 3: According to the axle load times x1 and the pore water action times x2, the cumulative load times X is obtained, and the relationship between the cumulative load times X and the rutting depth is fitted.

[0042] Let the cumulative load times X = λ1∑x1+λ2∑x2, where λ1 and λ2 are weights, and the values ​​of λ1 and λ2 are both (0, 1]. The size of λ1 and λ2 represents which of the axle load and pore water effect has the main influence on rutting and which has the secondary influence. The axle load times x1 and the pore water effect times x2 can be accumulated according to a fixed period, such as every day, every week or every month as a cumulative period; the values ​​of λ1 and λ2 are positively correlated with the cumulative use time of the sample.

[0043] After obtaining the cumulative load times, the relationship between the cumulative load times X and the rutting depth is further fitted, and the rutting depth is set to f(X), and we have

[0044] Where k1 is the first threshold; k2 is the second threshold; a, b, c, d, E and Fi N is the fitting parameter, i=1, 2, 3, 4, 5, 6. The reason for setting different first and second thresholds is that the interior of the sample is not completely compacted and there is a certain amount of compressed space. After exceeding the first threshold, it means that the work done by the load on the sample makes the interior of the sample compacted to a certain extent. Then, by further doing work on the sample until the second threshold k2 is reached, the size of k2 reflects the sample's ability to resist rutting. The smaller the value, the worse the ability to resist rutting or water damage.

[0045] Specifically, the first threshold k1 corresponds to the cumulative load times of 94% density of asphalt mixture; the second threshold k2 corresponds to the cumulative load times of 96% density of asphalt mixture; the calculation formula of density G is: Where m0 is the mass of asphalt mixture; A is the oil-stone ratio; V1 is the volume of asphalt before mixing; V2 is the volume of mineral aggregate before mixing; ρ0 is the theoretical maximum density of asphalt mixture.

[0046] Step 4: Build a neural network model, optimize the fitting parameters in the relationship between the cumulative load times X and the rutting depth, until the iteration termination condition or the maximum number of iterations is reached, and output the trained fitting parameters to substitute into the relationship between the cumulative load times X and the rutting depth;

[0047] In step 4, the neural network model is constructed to construct a neural network with at least three hidden layers. The activation function of the neural network model is the sigmoidal function. In order to accelerate the learning rate, a variable learning rate can be used to accelerate learning. For example, a fixed initial learning rate of 1 is used for the first Y iterations, and the learning rate is gradually reduced by exponential decay from the Y+1th iteration. The learning rate is U*η y , U is the initial learning rate, which is a real number between (0, 1); η is the decay coefficient, which is a real number between (0, 1); y is the decay rate, which is proportional to the number of iterations. The expression of the decay rate is y=y0e kN , y0 is the initial decay rate; k is a constant, the value is -0.05; N is the cumulative number of iterations, N>Y; until the maximum number of iterations is reached. In this embodiment, let Y≤20 times and N≤40 times.

[0048] Step 5: Use the relationship between the cumulative load times X and the rutting depth after optimizing the fitting parameters, and predict the rutting depth under the corresponding axle load times and pore water action times according to the input parameters. That is, by inputting the cumulative load times X and environmental conditions, the rutting depth of the asphalt mixture under the specified axle load times and pore water action can be obtained. Thus, a better prediction model for predicting the performance of asphalt mixture is obtained.

[0049] In addition, the present invention also provides an asphalt mixture performance prediction system, comprising:

[0050] The sample forming module is used to prepare asphalt mixture samples and preheat and insulate the samples;

[0051] The load test module is used to perform a load test on the pavement area where the specimen is located until the specified rutting depth is reached;

[0052] The data recording module is used to record the equivalent load conversion value of the asphalt mixture, each load time, slope angle, daily average temperature, pore water pressure, sample length, and sample water content during each load test within the time period of reaching the specified rutting depth; obtain the axle load times x1 and pore water action times x2;

[0053] The depth fitting module is used to obtain the axle load times x1 and the pore water action times x2 according to the contents of each variable obtained by the data recording module, and further obtain the cumulative load times X; fit the relationship between the cumulative load times X and the rutting depth;

[0054] Training module: construct a neural network model, optimize the fitting parameters in the relationship between the cumulative load times X and the rutting depth, until the iteration termination condition or the maximum number of iterations is reached, and output the trained fitting parameters to substitute into the relationship between the cumulative load times X and the rutting depth;

[0055] The performance prediction module uses the relationship between the cumulative load times X and the rutting depth after optimizing the fitting parameters, and predicts the rutting depth under the corresponding axle load times and pore water action times according to the input parameters.

[0056] The present invention also provides a computer-readable storage medium. When a computer program stored in the storage medium is run on a computer, the computer is enabled to execute the above-mentioned asphalt mixture performance prediction method.

[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting asphalt mixture performance, characterized in that: The following steps are involved: Step 1: Prepare asphalt mixture samples and preheat and insulate the samples; Step 2: Carry out a load test on the pavement area where the sample is located. During the time period when the specified rutting depth is reached, record the variables including the equivalent load conversion value borne by the asphalt mixture during each load test, each load time, slope angle, daily average temperature, pore water pressure, sample length, and sample water content, and obtain the axle load times x1 and pore water action times x2; Step 3: According to the axle load times x1 and the pore water action times x2, the cumulative load times X is obtained, and the relationship between the cumulative load times X and the rutting depth is fitted; Step 4: Build a neural network model, optimize the fitting parameters in the relationship between the cumulative load times X and the rutting depth, until the iteration termination condition or the maximum number of iterations is reached, and output the trained fitting parameters to substitute into the relationship between the cumulative load times X and the rutting depth; Step 5: Use the relationship between the cumulative load times X and the rutting depth after optimizing the fitting parameters to predict the rutting depth under the corresponding axle load times and pore water action times according to the input parameters.

2. The asphalt mixture performance prediction method according to claim 1, characterized in that: The asphalt mixture sample prepared in step 1 is designed according to the asphalt-to-stone ratio of 4.3-4.9%, and the average density of the mineral aggregate is 2.78 g / cm 3 ; Filling in a rigid metal box with a density of 88%-90% to make a sample, the sample is prepared by keeping warm at 60 degrees Celsius for more than 2 hours.

3. The asphalt mixture performance prediction method according to claim 2, characterized in that: In step 2, the equivalent load conversion value of the asphalt mixture, the load time of each load, the slope angle, the daily average temperature, the pore water pressure, the sample length, and the sample water content are recorded for each load test, and the axle load times x1 and the pore water action times x2 are obtained, which are respectively constructed as follows: Where n0 is the wave speed; p is the ground pressure in MPa; υ is the vehicle speed in m / s; t1 is the duration of each load; P E is the equivalent load conversion value of each load, unit is MPa; θ is the slope angle; T is the daily average temperature; l is a non-zero constant; P N is the pore water pressure, unit is MPa; t2 is the action time of the pore water pressure; L is the length of the sample; s is the adjustment coefficient; γ% is the moisture content of the sample, and the moisture content of the sample is greater than zero.

4. The asphalt mixture performance prediction method according to claim 3, characterized in that: The value of the adjustment coefficient S is related to the average monthly precipitation at the sample location. If the average monthly precipitation is less than 30, the value of S is (0.8, 1); if the average monthly precipitation is greater than 30, the value of S is [1, 1.2).

5. The asphalt mixture performance prediction method according to claim 3, characterized in that: As described in step 3, the cumulative load times X is obtained according to the axial load times x1 and the pore water action times x2, which is to set X = λ1Σx1+λ2Σx2, where λ1 and λ2 are weights, and the values ​​of λ1 and λ2 are both (0, 1].

6. The asphalt mixture performance prediction method according to claim 5, characterized in that: The relationship between the cumulative load times X and the rutting depth is fitted by setting the rutting depth to f(X), Where k1 is the first threshold; k2 is the second threshold; a, b, c, d, E and F i N are fitting parameters, i=1,2,3,4,5,6.

7. The asphalt mixture performance prediction method according to claim 6, characterized in that: The first threshold k1 corresponds to the cumulative load times of 94% density of asphalt mixture; the second threshold k2 corresponds to the cumulative load times of 96% density of asphalt mixture; the calculation formula of density G is: Where m0 is the mass of asphalt mixture; A is the oil-stone ratio; V1 is the volume of asphalt before mixing; V2 is the volume of mineral aggregate before mixing; ρ0 is the theoretical maximum density of asphalt mixture.

8. The asphalt mixture performance prediction method according to claim 1, characterized in that: The neural network model constructed in step 4 is to construct a neural network with at least three hidden layer structures, and the activation function of the neural network model is the sigmoidal function.

9. An asphalt mixture performance prediction system, characterized in that: include: The sample forming module is used to prepare asphalt mixture samples and preheat and insulate the samples; The load test module is used to perform a load test on the pavement area where the specimen is located until the specified rutting depth is reached; The data recording module is used to record the equivalent load conversion value of the asphalt mixture, each load time, slope angle, daily average temperature, pore water pressure, sample length, and sample water content during each load test within the time period of reaching the specified rutting depth; obtain the axle load times x1 and pore water action times x2; The depth fitting module is used to obtain the axle load times x1 and the pore water action times x2 according to the contents of each variable obtained by the data recording module, and further obtain the cumulative load times X; fit the relationship between the cumulative load times X and the rutting depth; Training modules; A neural network model is constructed to optimize the fitting parameters in the relationship between the cumulative number of loads X and the rutting depth until the iteration termination condition or the maximum number of iterations is reached, and the trained fitting parameters are output and substituted into the relationship between the cumulative number of loads X and the rutting depth; The performance prediction module uses the relationship between the cumulative load times X and the rutting depth after optimizing the fitting parameters, and predicts the rutting depth under the corresponding axle load times and pore water action times according to the input parameters.

10. A computer-readable storage medium, characterized in that: When the computer program stored in the storage medium is executed on a computer, the computer is enabled to execute the asphalt mixture performance prediction method according to any one of claims 1 to 8.

Images