Method and system for predicting service life of environment-friendly asphalt pavement

Abstract

The invention discloses a method and a system for predicting the service life of an environment-friendly asphalt pavement, and belongs to the technical field of road engineering. The method comprises the following steps: S1, acquiring an original radar signal of an asphalt pavement, and obtaining pavement state information based on the original radar signal; and S2, constructing a prediction model, and predicting the road surface state information based on the prediction model to obtain a prediction result. According to the method, the radar signal is obtained by adopting the three-dimensional radar, the real-time pavement interlayer bonding state is obtained based on the radar signal, the pavement state grade prediction is performed on the pavement interlayer bonding state based on the constructed prediction model, and the asphalt pavement life prediction is performed based on the pavement state grade and the fatigue performance evaluation model. And certain reference data is provided for scientific maintenance and efficient operation of the pavement.

Description

Technical Field

[0001] The invention belongs to the technical field of road engineering, and in particular relates to a method and system for predicting the life of an environmentally friendly asphalt pavement. Background Art

[0002] Asphalt pavement performance prediction is the basis for making decisions on asphalt pavement maintenance and management, and is also the basis for transportation infrastructure planning, construction and resource allocation. Accurate and reasonable prediction of asphalt pavement performance is of guiding significance for scientific maintenance and efficient operation of highways.

[0003] At present, the pavement performance prediction models based on state degradation can be divided into deterministic models, statistical models and random models. Among them, the deterministic model describes the degradation path as a linear or nonlinear regression equation; the statistical model regards the degradation rate as a random variable and estimates the reliability by fitting the data to a probability distribution function; however, due to the changing characteristics of factors such as materials, environment, traffic load and maintenance measures, there is a great deal of uncertainty in the degradation of road performance, which makes it difficult to quantify the pavement degradation parameters, time-consuming and labor-intensive statistics, and difficult to update later. The random model believes that data is a macroscopic manifestation of the combined effects of various influencing factors on road performance. By fully developing and mining data information and extracting useful information, it can effectively describe the law of pavement performance degradation, which is also applicable to the characteristics of "small samples, little information, and uncertainty" of asphalt pavements in China today. The Markov process is a currently recognized random model and has been fully applied to the prediction of asphalt pavement performance. However, as a discrete random process, the Markov process cannot describe the continuous characteristics of pavement degradation. Summary of the invention

[0004] The present invention aims to solve the deficiencies of the prior art and proposes an environmentally friendly asphalt pavement life prediction method and system. By adopting a three-dimensional radar, radar signals are obtained, and the pavement condition grade of the radar signals is predicted based on the constructed prediction model. The asphalt pavement life is predicted based on the pavement condition grade and fatigue performance evaluation model.

[0005] To achieve the above object, the present invention provides the following scheme: a method for predicting the life of an environmentally friendly asphalt pavement, comprising the following steps:

[0006] S1. Collecting original radar signals of an asphalt pavement, and obtaining pavement status information based on the original radar signals;

[0007] S2. Construct a prediction model, and predict the road surface status information based on the prediction model to obtain a prediction result.

[0008] Further preferably, S1 includes:

[0009] S1.1. collecting the original radar signal from the asphalt road surface through a three-dimensional radar, and preprocessing the original radar signal to obtain a processed signal;

[0010] S1.2. Based on the processed signal, an image of the interlayer adhesion state of the asphalt pavement is obtained, and based on the interlayer adhesion state image, the pavement state information is obtained.

[0011] Further preferably, the method for the prediction model to predict the road surface state information includes:

[0012] The road surface state information is input into the prediction model and passed through a 3×3 convolution layer to obtain a preliminary feature matrix;

[0013] Afterwards, the preliminary feature matrix is ​​passed through the intermediate layer to obtain a weighted relationship between each channel of the preliminary feature matrix and each spatial position;

[0014] After that, it is outputted through the middle layer and connected to the 1×1 convolutional layer for feature extraction to obtain the second feature matrix;

[0015] The second feature matrix is ​​processed by global average pooling and then passes through two fully connected layers to obtain the final output, and the size of the final output is 1×1×6.

[0016] Further preferably, the middle layer includes 16 MBConv modules, and the last 5 MBConv modules include a channel attention module and a position attention module;

[0017] The channel attention module is used to obtain the feature dependency between channels of the preliminary feature matrix;

[0018] The position attention module is used to obtain the spatial feature dependencies of the preliminary feature matrix.

[0019] The present invention also provides an environmentally friendly asphalt pavement life prediction system, comprising: a data acquisition system and a prediction system;

[0020] The data acquisition system is used to collect original radar signals of asphalt pavement, and obtain pavement status information based on the original radar signals;

[0021] The prediction system is used to construct a prediction model, and predict the road surface state information based on the prediction model to obtain a prediction result.

[0022] Further preferably, the data acquisition system comprises: an acquisition module and a processing module;

[0023] The acquisition module acquires the original radar signal from the asphalt road surface through a three-dimensional radar;

[0024] The processing module is used to preprocess the original radar signal to obtain a processed signal; based on the processed signal, obtain an asphalt pavement interlayer bonding state image, and obtain the pavement state information based on the interlayer bonding state image.

[0025] Further preferably, in the prediction system, the method in which the prediction model predicts the road surface state information includes:

[0026] The road surface state information is input into the prediction model and passed through a 3×3 convolution layer to obtain a preliminary feature matrix;

[0027] Afterwards, the preliminary feature matrix is ​​passed through the intermediate layer to obtain a weighted relationship between each channel of the preliminary feature matrix and each spatial position;

[0028] After that, it is outputted through the middle layer and connected to the 1×1 convolutional layer for feature extraction to obtain the second feature matrix;

[0029] The second feature matrix is ​​processed by global average pooling and then passes through two fully connected layers to obtain the final output, and the size of the final output is 1×1×6.

[0030] Further preferably, the middle layer includes 16 MBConv modules, and the last 5 MBConv modules include a channel attention module and a position attention module;

[0031] The channel attention module is used to obtain the feature dependency between channels of the preliminary feature matrix;

[0032] The position attention module is used to obtain the spatial feature dependencies of the preliminary feature matrix.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The present invention uses a three-dimensional radar to obtain radar signals, and based on the radar signals, obtains the real-time pavement interlayer bonding state, and based on the constructed prediction model, predicts the pavement state grade of the pavement interlayer bonding state, and predicts the asphalt pavement life based on the pavement state grade and fatigue performance evaluation model, providing certain reference data for scientific pavement maintenance and efficient operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. 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 labor.

[0036] Figure 1The present invention is a flowchart of an environmentally friendly asphalt pavement life prediction method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in 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.

[0038] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] Embodiment 1:

[0040] like Figure 1 As shown, this embodiment provides an environmentally friendly asphalt pavement life prediction method, comprising the following steps:

[0041] S1. Collecting original radar signals of asphalt pavement, and obtaining pavement status information based on the original radar signals.

[0042] Specifically, S1 includes:

[0043] S1.1. Collect original radar signals from the asphalt pavement through a three-dimensional radar, and pre-process the original radar signals to obtain processed signals.

[0044] The collected original radar signal may be subject to electromagnetic interference from various aspects. Therefore, this embodiment performs external signal source interference suppression and background noise removal processing on the original radar signal.

[0045] Among them, the method of suppressing interference from external signal sources includes:

[0046]

[0047] In the formula, D(f m ) indicates the frequency is f m The frequency domain signal, is the target signal, When the difference between the frequency of the interference signal and the frequency domain signal exceeds the threshold, the interference signal of the external signal source will be eliminated.

[0048] Methods for background noise removal include:

[0049]

[0050] Where x(i,j) is the original radar signal; N is the number of sampling points.

[0051] Afterwards, the preprocessed original radar signal is converted from a frequency domain signal to a time domain signal using discrete Fourier transform to obtain a processed signal.

[0052] S1.2. Based on the processed signal, an image of the interlayer adhesion state of the asphalt pavement is obtained, and pavement state information is obtained based on the interlayer adhesion state image.

[0053] According to the processed signal, the amplitude of the phase axis between the asphalt layer and the base layer is extracted to obtain the interlayer bonding state image of the asphalt pavement. The interlayer contact state is evaluated by the amplitude strength of the phase axis at the layer interface of the asphalt pavement: the larger the phase axis amplitude, the worse the interlayer contact state; the smaller the phase axis amplitude, the better the interlayer contact state.

[0054] The obtained asphalt pavement interlayer bonding state image is scored by manual labeling, and the asphalt pavement interlayer bonding state image with the score is used as the pavement state information. The pavement state information is preprocessed by cropping to obtain an image of 224 pixels × 224 pixels.

[0055] The scoring is divided into: 100 - road surface condition is level 6, 90-100 - road surface condition is level 5, 80-90 - road surface condition is level 4, 70-80 - road surface condition is level 3, 60-70 - road surface condition is level 2, 0-60 - road surface condition is level 1.

[0056] S2. Construct a prediction model, predict the road surface status information based on the prediction model, and obtain a prediction result.

[0057] Specifically, the prediction model is mainly composed of a 3×3 convolution layer, an intermediate layer and an output layer. The 224×224×3 image is input into the prediction model, and after the 3×3 convolution layer, a preliminary feature matrix of size 112×112×32 is obtained. Afterwards, the preliminary feature matrix is ​​passed through the intermediate layer to obtain a weighted relationship between each channel and each spatial position of the preliminary feature matrix; in this embodiment, the intermediate layer includes 16 MBConv modules, and the last 5 MBConv modules include a channel attention module and a position attention module; the channel attention module is used to obtain the feature dependency between the channels of the preliminary feature matrix; the position attention module is used to obtain the spatial feature dependency of the preliminary feature matrix. Among them, the MBConv module includes a 1×1 convolution layer, connected to the BN layer and the SiLU activation function, and a new feature matrix is ​​generated by calculating the linear combination function of each channel of the input feature matrix. It also includes a layer of depth-separable convolution operation, and then connects the BN layer and the SiLU activation function. It is then divided into two branches, one of which passes through global average pooling, two fully connected layers and a Sigmoid activation function in turn. The other branch does not perform any operation, and multiplies the main elements of the results of the two branches. The multiplied result is output after passing through a 1×1 convolution layer. In the channel attention module, the input matrix first undergoes two parallel pooling operations to extract features and analyze weight relationships, and then passes through two fully connected layers, SiLU and Sigmoid activation functions to generate two new feature matrices; the two generated feature matrices are added by element-by-element summation, and the channel attention matrix is ​​obtained after passing through the Sigmoid activation function. The channel attention matrix and the input matrix are multiplied element-by-element to obtain a fusion matrix; the fusion matrix is ​​input into the position attention module, and the two results are concatenated after two parallel pooling operations to obtain concatenated features. After that, the concatenated features are sequentially passed through a 7×7 convolution kernel, a BN layer, and a Sigmoid activation function to obtain a position attention matrix.

[0058] After that, the output of the intermediate layer is connected to the 1×1 convolutional layer for feature extraction to obtain the second feature matrix; the second feature matrix is ​​processed by global average pooling and then passes through two fully connected layers to obtain the final output, and the final output size is 1×1×6.

[0059] The pavement condition grade is obtained through the above prediction model, and then an estimation model between the pavement condition grade and life is constructed. The estimation model includes:

[0060] Y=0.001e 0.0056Es ×1367.8e -0.004Eb ×(9×10 -6 μ+6×10 -5 )×802.55P -1.459 ,

[0061] Where Es is the modulus of the asphalt layer; Eb is the modulus of the base layer; μ is the friction coefficient between the surface layer and the base layer; and P is the load.

[0062] Among them, the friction coefficient between the surface layer and the base layer is related to the pavement condition level; when the pavement condition is level 6, the friction coefficient is 1; when the pavement condition is level 5, the friction coefficient is 0.9; when the pavement condition is level 4, the friction coefficient is 0.7; when the pavement condition is level 3, the friction coefficient is 0.4; when the pavement condition is level 2, the friction coefficient is 0.2; when the pavement condition is level 1, the friction coefficient is 0.1.

[0063] Embodiment 2:

[0064] This embodiment provides an environmentally friendly asphalt pavement life prediction system, including: a data acquisition system and a prediction system.

[0065] The data acquisition system is used to collect the original radar signals of the asphalt pavement and obtain the pavement status information based on the original radar signals.

[0066] Specifically, the data acquisition system includes: an acquisition module and a processing module; the acquisition module collects original radar signals from the asphalt pavement through a three-dimensional radar. The processing module is used to pre-process the original radar signals to obtain processed signals; based on the processed signals, an image of the interlayer bonding state of the asphalt pavement is obtained, and the pavement state information is obtained based on the image of the interlayer bonding state.

[0067] The collected original radar signal may be subject to electromagnetic interference from various aspects. Therefore, in this embodiment, the processing module first performs external signal source interference suppression and background noise removal processing on the original radar signal.

[0068] Among them, the method of suppressing interference from external signal sources includes:

[0069]

[0070] In the formula, D(f m ) indicates the frequency is f m The frequency domain signal, is the target signal, When the difference between the frequency of the interference signal and the frequency domain signal exceeds the threshold, the interference signal of the external signal source will be eliminated.

[0071] Methods for background noise removal include:

[0072]

[0073] Where x(i,j) is the original radar signal; N is the number of sampling points.

[0074] Afterwards, the preprocessed original radar signal is converted from a frequency domain signal to a time domain signal using discrete Fourier transform to obtain a processed signal.

[0075] Finally, according to the processed signal, the amplitude of the phase axis between the asphalt layer and the base layer is extracted to obtain the interlayer bonding state image of the asphalt pavement. The interlayer contact state is evaluated by the amplitude intensity of the phase axis at the layer interface of the asphalt pavement: the larger the phase axis amplitude, the worse the interlayer contact state; the smaller the phase axis amplitude, the better the interlayer contact state.

[0076] The obtained asphalt pavement interlayer bonding state image is scored by manual labeling, and the asphalt pavement interlayer bonding state image with the score is used as the pavement state information. The pavement state information is preprocessed by cropping to obtain an image of 224 pixels × 224 pixels.

[0077] The scoring is divided into: 100 - road surface condition is level 6, 90-100 - road surface condition is level 5, 80-90 - road surface condition is level 4, 70-80 - road surface condition is level 3, 60-70 - road surface condition is level 2, 0-60 - road surface condition is level 1.

[0078] The prediction system is used to build a prediction model, predict the road surface status information based on the prediction model, and obtain the prediction result.

[0079] In this embodiment, the constructed prediction model is mainly composed of a 3×3 convolution layer, an intermediate layer and an output layer. The 224×224×3 image is input into the prediction model, and after the 3×3 convolution layer, a preliminary feature matrix of size 112×112×32 is obtained. Afterwards, the preliminary feature matrix is ​​passed through the intermediate layer to obtain a weighted relationship between each channel and each spatial position of the preliminary feature matrix; in this embodiment, the intermediate layer includes 16 MBConv modules, and the last 5 MBConv modules include a channel attention module and a position attention module; the channel attention module is used to obtain the feature dependency between the channels of the preliminary feature matrix; the position attention module is used to obtain the spatial feature dependency of the preliminary feature matrix. Among them, the MBConv module includes a 1×1 convolution layer, connected to the BN layer and the SiLU activation function, and a new feature matrix is ​​generated by calculating the linear combination function of each channel of the input feature matrix. It also includes a layer of depth separable convolution operation, and then connects the BN layer and the SiLU activation function. It is then divided into two branches, one of which passes through global average pooling, two fully connected layers and Sigmoid activation function in turn. The other branch does not perform any operation, and multiplies the main elements of the results of the two branches. The multiplied result is output after passing through a 1×1 convolution layer. In the channel attention module, the input matrix first undergoes two parallel pooling operations to extract features and analyze weight relationships, and then passes through two fully connected layers, SiLU and Sigmoid activation functions to generate two new feature matrices; the two generated feature matrices are added by element-by-element summation, and the channel attention matrix is ​​obtained after passing through the Sigmoid activation function. The channel attention matrix and the input matrix are multiplied element-by-element to obtain a fusion matrix; the fusion matrix is ​​input into the position attention module, and the two results are concatenated after two parallel pooling operations to obtain concatenated features. After that, the concatenated features are sequentially passed through a 7×7 convolution kernel, a BN layer, and a Sigmoid activation function to obtain a position attention matrix.

[0080] After that, the output of the intermediate layer is connected to the 1×1 convolutional layer for feature extraction to obtain the second feature matrix; the second feature matrix is ​​processed by global average pooling and then passes through two fully connected layers to obtain the final output, and the final output size is 1×1×6.

[0081] The pavement condition grade is obtained through the above prediction model, and then an estimation model between the pavement condition grade and life is constructed. The estimation model includes:

[0082] Y=0.001e 0.0056Es ×1367.8e -0.004Eb ×(9×10 -6 μ+6×10 -5 )×802.55P -1.459 ,

[0083] Where Es is the modulus of the asphalt layer; Eb is the modulus of the base layer; μ is the friction coefficient between the surface layer and the base layer; and P is the load.

[0084] Among them, the friction coefficient between the surface layer and the base layer is related to the pavement condition level; when the pavement condition is level 6, the friction coefficient is 1; when the pavement condition is level 5, the friction coefficient is 0.9; when the pavement condition is level 4, the friction coefficient is 0.7; when the pavement condition is level 3, the friction coefficient is 0.4; when the pavement condition is level 2, the friction coefficient is 0.2; when the pavement condition is level 1, the friction coefficient is 0.1.

[0085] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for predicting the life of an environmentally friendly asphalt pavement, characterized in that: The following steps are involved: S1. Collecting original radar signals of an asphalt pavement, and obtaining pavement status information based on the original radar signals; S2. Construct a prediction model, and predict the road surface status information based on the prediction model to obtain a prediction result.

2. According to claim 1, the method for predicting the life of an environmentally friendly asphalt pavement is characterized in that: S1 includes: S1.

1. collecting the original radar signal from the asphalt road surface through a three-dimensional radar, and preprocessing the original radar signal to obtain a processed signal; S1.

2. Based on the processed signal, an image of the interlayer adhesion state of the asphalt pavement is obtained, and based on the interlayer adhesion state image, the pavement state information is obtained.

3. The method for predicting the life of an environmentally friendly asphalt pavement according to claim 1, characterized in that: The method for predicting the road surface state information by the prediction model includes: The road surface state information is input into the prediction model and passed through a 3×3 convolution layer to obtain a preliminary feature matrix; Afterwards, the preliminary feature matrix is ​​passed through the intermediate layer to obtain a weighted relationship between each channel of the preliminary feature matrix and each spatial position; After that, it is outputted through the middle layer and connected to the 1×1 convolutional layer for feature extraction to obtain the second feature matrix; The second feature matrix is ​​processed by global average pooling and then passes through two fully connected layers to obtain the final output, and the size of the final output is 1×1×6.

4. The method for predicting the life of an environmentally friendly asphalt pavement according to claim 3, characterized in that: The middle layer includes 16 MBConv modules, and the last 5 MBConv modules include a channel attention module and a position attention module; The channel attention module is used to obtain the feature dependency between channels of the preliminary feature matrix; The position attention module is used to obtain the spatial feature dependencies of the preliminary feature matrix.

5. An environmentally friendly asphalt pavement life prediction system, the system is used to implement the method according to any one of claims 1 to 4, characterized in that: include: Data collection system and prediction system; The data acquisition system is used to collect original radar signals of asphalt pavement, and obtain pavement status information based on the original radar signals; The prediction system is used to construct a prediction model, and predict the road surface state information based on the prediction model to obtain a prediction result.

6. The environmentally friendly asphalt pavement life prediction system according to claim 5, characterized in that: The data acquisition system comprises: an acquisition module and a processing module; The acquisition module acquires the original radar signal from the asphalt road surface through a three-dimensional radar; The processing module is used to preprocess the original radar signal to obtain a processed signal; based on the processed signal, obtain an asphalt pavement interlayer bonding state image, and obtain the pavement state information based on the interlayer bonding state image.

7. The environmentally friendly asphalt pavement life prediction system according to claim 5, characterized in that: In the prediction system, the method for the prediction model to predict the road surface state information includes: The road surface state information is input into the prediction model and passed through a 3×3 convolution layer to obtain a preliminary feature matrix; Afterwards, the preliminary feature matrix is ​​passed through the intermediate layer to obtain a weighted relationship between each channel of the preliminary feature matrix and each spatial position; After that, it is outputted through the middle layer and connected to the 1×1 convolutional layer for feature extraction to obtain the second feature matrix; The second feature matrix is ​​processed by global average pooling and then passes through two fully connected layers to obtain the final output, and the size of the final output is 1×1×6.

8. The environmentally friendly asphalt pavement life prediction system according to claim 7, characterized in that: The middle layer includes 16 MBConv modules, and the last 5 MBConv modules include a channel attention module and a position attention module; The channel attention module is used to obtain the feature dependency between channels of the preliminary feature matrix; The position attention module is used to obtain the spatial feature dependencies of the preliminary feature matrix.

Images