Road surface wear resistance prediction method considering environment-load coupling effect
By constructing a multiphase coupled anti-wear simulation model and a convolutional bidirectional memory collaborative neural network, combined with the SHAP model, the problems of accuracy and real-time performance in predicting pavement anti-wear performance were solved, and high-precision pavement wear analysis and maintenance decision support were achieved.
Patent Information
- Application Number
- CN202511015515.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-23
- Publication Date
- 2025-11-07
AI Technical Summary
Existing technologies lack methods for predicting pavement wear resistance under environmental-load coupling effects, and cannot systematically analyze the quantitative relationship between multi-factor interaction mechanisms and main controlling factors, resulting in insufficient accuracy and real-time performance in pavement wear prediction.
A multiphase coupled anti-wear simulation model was constructed, and a convolutional bidirectional memory collaborative neural network architecture was used for performance prediction. The prediction results were analyzed through the SHAP model to explain the interaction of various factors and the main controlling factors.
It achieves high-precision, real-time prediction of pavement wear resistance performance, provides a basis for pavement maintenance, and improves the interpretability of predictions.
Smart Images
Figure CN120911271A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of pavement wear resistance prediction, and in particular to a pavement wear resistance prediction method considering environment-load coupling. BACKGROUND
[0002] Cement concrete pavement occupies an important position in the types of highway pavement in China, mainly due to its significant advantages in mechanical properties and use characteristics: high structural strength, excellent durability and good stability. However, during the use of cement concrete pavement, with the increase of the number of vehicles in China, the pavement is subjected to high-frequency tire wear of vehicles, and is also affected by complex environmental factors (temperature, freezing and thawing), which gradually reduces the surface safety performance of the pavement, leading to surface structure damage such as pavement polishing and spalling. Such diseases have adverse effects on the safety and smoothness of vehicle operation during the process, and seriously threaten the safety of vehicle operation. Accurate prediction of pavement wear resistance under different working conditions can provide support for the design of new cement concrete pavement and the development of detection schemes for existing pavement. Analyzing the interaction of various factors affecting pavement wear resistance and the main control factors can provide a reference for pavement maintenance decisions. The existing technology has two limitations: first, there is no wear resistance prediction method under the coupling of environment and load; second, it cannot systematically analyze the quantitative relationship between the multi-factor interaction mechanism and the main control factors.
[0003] Therefore, how to provide a pavement wear resistance prediction method with high precision, real-time performance and good interpretability is a problem to be solved at present. SUMMARY
[0004] In view of the problems existing in the prior art, the present application provides a pavement wear resistance prediction method considering environment-load coupling, to solve the technical problems of lack of analysis of the interaction of various factors and the main control factors in the prior art for pavement wear resistance prediction, and insufficient precision and real-time performance.
[0005] The present application provides a pavement wear resistance prediction method considering environment-load coupling, comprising:
[0006] S1, constructing a multi-phase coupling wear resistance simulation model comprising at least temperature field, freezing and thawing field, automobile tire and cement concrete pavement factors;
[0007] S2, using a convolution bidirectional memory collaborative neural network architecture to perform performance prediction on the multi-phase coupling wear resistance simulation model, and outputting a prediction result;
[0008] S3, based on the temperature field, the freezing and thawing field, the automobile tire and the cement concrete pavement factors, performing weight analysis on the prediction result by a SHAP model.
[0009] Optionally, the cement concrete pavement comprises:
[0010] 22-26 cm cement concrete surface layer, 2x20 cm cement stabilized macadam base and soil base, and four square foot concrete plates with a size of 5m x 5m are drawn by ANSYS Workbench, the inter-panel joints of the square foot concrete plates are connected by force transmission rods with a cross-sectional diameter of 30mm and a length of 500mm, the spacing is 300mm, the mesh module in Workbench is used to automatically divide the grid, and the grid size is set to a hexahedral solid grid of 5cm.
[0011] Optionally, the automobile tire comprises:
[0012] A radial tire is adopted, which is divided into an airtight layer, a body cord layer, a belt layer, a crown, a sidewall, and a bead, and the tire pressure of the automobile tire is set to 0.75MPa.
[0013] Optionally, the temperature field comprises:
[0014] The range is -30℃ / m-60℃ / m, the temperature gradient of the cement concrete pavement is calculated according to the one-dimensional heat conduction theory, and the one-dimensional heat conduction theory formula is represented as:
[0015]
[0016] In the formula, a is the temperature coefficient, T is the air temperature, z is the thickness of the cement concrete pavement, t is the time, and the thickness of the cement concrete pavement is divided into layers according to every 2-4cm, and each layer corresponds to the calculation of the temperature gradient according to the one-dimensional heat conduction theory formula.
[0017] Optionally, the freeze-thaw field comprises:
[0018] According to the indoor rapid freeze-thaw test, the mechanical parameter curve of the target cement concrete surface layer material changes with the number of indoor freeze-thaw times, the mechanical parameters include: elastic modulus, tensile strength, and compressive strength, and the maximum freeze-thaw times of the freeze-thaw test is 300 times.
[0019] Optionally, the convolution bidirectional memory collaborative neural network architecture comprises:
[0020] The convolutional neural network and the bidirectional long short-term memory neural network input the original data of the multiphase coupling anti-wear simulation model into the convolutional neural network for feature extraction, and input the extracted features into the bidirectional long short-term memory neural network to output the prediction result.
[0021] Optionally, the original data of the multiphase coupling anti-wear simulation model is input into the convolutional neural network for feature extraction, comprising:
[0022] S201, missing value and outlier processing is performed on the original data;
[0023] S202, data standardization processing is performed on the original data;
[0024] S203, the original data is divided into a training set and a test set respectively;
[0025] S204, the training set is input into the convolutional neural network for training, and a trained convolutional neural network is obtained;
[0026] S205, the test set is input into the pooling layer of the convolutional neural network for feature compression, feature extraction, and the features are input into the bidirectional long short-term memory neural network.
[0027] Optionally, the SHAP model comprises:
[0028] The Shap value of the SHAP model for analyzing the temperature field, the freeze-thaw field, the automobile tire and the cement concrete pavement factors is represented as:
[0029]
[0030] In the formula: N is the set of all features, |S| is a feature subset, i is any feature in the temperature field, the freeze-thaw field, the automobile tire and the cement concrete pavement, and φ i (f) is the Shap value of any feature i.
[0031] Compared with the prior art, the present application has the following advantages:
[0032] A prediction analysis method fusing multi-physical field coupling and explainable AI is proposed, road surface wear resistance data is obtained through a multi-factor coupling wear resistance simulation model of a temperature field, a freeze-thaw field, an automobile tire and a cement concrete pavement, wear resistance performance prediction is performed based on model data through a CNN-BiLSTM neural network, and finally the interaction and the main control factor of each factor in the prediction model are analyzed through a SHAP model explanation tool, thereby providing a basis for cement concrete pavement maintenance and operation, the multi-phase coupling wear resistance simulation model has the characteristics of high precision and real-time, and the SHAP model realizes explainability. BRIEF DESCRIPTION OF DRAWINGS
[0033] The drawings incorporated into the specification and forming a part thereof, show embodiments consistent with the present application, and together with the specification serve to explain the principles of the present application.
[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, for those skilled in the art, other drawings can also be obtained from these drawings without any creative effort.
[0035] Figure 1 A flowchart of the method of the present application is shown in the figure.
[0036] Figure 2 A general structure diagram of the CNN-BiLSTM prediction model provided in the embodiments of the present application is shown in the figure.
[0037] Figure 3 A CNN-BiLSTM neural network prediction result diagram provided in the embodiments of the present application is shown in the figure.
[0038] Figures 4(a) and (b) are interaction and main control factor analysis diagrams of various factors in the road surface wear resistance prediction model provided in the embodiments of the present application.
[0039] Figure 5 A prediction parameter diagram of different types of machine learning algorithms provided in the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION
[0040] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the following will describe the technical solutions in the embodiments of the present application clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort are within the scope of protection of the present application. The function units with the same and similar structures and functions in the embodiments of the present application have the same and similar structures and functions.
[0041] Referring to Figure 1 The present application provides a road surface wear resistance performance prediction method considering environmental load coupling, comprising:
[0042] S1, constructing a multi-phase coupling wear resistance simulation model comprising at least temperature field, freeze-thaw field, automobile tire and cement concrete pavement factors;
[0043] S2, using a convolution bidirectional memory collaborative neural network architecture to perform performance prediction on the multi-phase coupling wear resistance simulation model, and outputting a prediction result;
[0044] S3, based on the temperature field, freeze-thaw field, automobile tire and cement concrete pavement factors, performing weight analysis on the prediction result by a SHAP model.
[0045] In this embodiment, S1, the construction of a multi-phase coupling anti-wear simulation model at least includes temperature field, freeze-thaw field, automobile tire and cement concrete pavement factors.
[0046] First, a multi-phase coupling anti-wear simulation model is constructed, and the cement concrete pavement structure is: 22-26 cm cement concrete surface layer, 2x20 cm cement stabilized macadam base and soil base layer, the temperature field range is: -30℃ / m-60℃ / m, the freeze-thaw field range is: 0-300 times of indoor rapid freeze-thaw, the cyclic wear load is: >10 million times, and the model accuracy is greater than 90%. The pavement anti-wear performance data is the average pavement structure depth under the coupling action of environmental load, and the value is (0-1.1 mm). 1.1 indicates that the pavement anti-wear performance is intact, and 0 indicates that the pavement loses the anti-wear capacity.
[0047] Specifically,
[0048] Referring to Figure 2 , the automobile tire adopts a radial tire. First, a two-dimensional cross-section model of the tire is drawn in ANSYS Workbench, and then the two-dimensional model is rotated by 360° to obtain a three-dimensional finite element model of the tire. Secondly, the mesh is automatically divided through the Mesh module in Workbench, and the model accuracy is controlled by adjusting the mesh size. Finally, the file with the mesh is output in the form of.K file and imported into LS-DYNA. The materials of each layer of the tire are assigned in LS-DYNA, and the closed cavity is processed to ensure the air tightness. The structure of the automobile tire mainly includes an air tight layer, a body cord layer, a belt layer, a crown, a side, and a bead part. The tire pressure is applied by the control volume method, and the tire pressure range is 0.75 MPa.
[0049] The cement concrete pavement structure is 26 cm cement concrete surface layer, 2x20 cm cement stabilized macadam base and soil base layer. Four square foot concrete plates with a size of 5m x 5m are drawn in ANSYS Workbench. The joints between the plates are connected by dowel bars with a cross-sectional diameter of 30 mm and a length of 500 mm. The spacing is 300 mm. The mesh is automatically divided by the Mesh module in Workbench, and the mesh size is set to 5 cm hexahedral solid mesh.
[0050] The temperature field data is the temperature gradient of the cement concrete pavement obtained after calculation according to the meteorological data of different regions by one-dimensional heat conduction theory.
[0051] The one-dimensional heat conduction theory formula is:
[0052]
[0053] In the formula, a is the temperature coefficient (0.003-0.005 m 2h); T is air temperature, z is road surface thickness; t is time.
[0054] The temperature field in the environment-tire-pavement multiphase coupling anti-wear simulation model is applied in the following manner: the cement concrete surface layer is divided into one layer with a thickness of 2-4 cm, a temperature is added to each layer, and the temperature of each layer is set according to the calculation results of the one-dimensional heat conduction theory formula in different regions. A stable temperature gradient is formed along the thickness direction of the road surface after different temperatures are input to each layer.
[0055] The freeze-thaw field data is obtained from the target cement concrete surface layer material mechanics parameter curve varying with the number of indoor freeze-thaw cycles (maximum freeze-thaw 300 times) obtained from the indoor rapid freeze-thaw test. The mechanics parameters include elastic modulus, tensile strength, and compressive strength.
[0056] S2, a convolutional bidirectional memory collaborative neural network architecture is used to perform performance prediction on the multiphase coupling anti-wear simulation model, and a prediction result is output.
[0057] A convolutional bidirectional memory collaborative neural network (CNN-BiLSTM) is constructed: the neural network is composed of a feature extraction module and a prediction module. In the first stage, a convolutional neural network (CNN) is used to construct a feature extractor to effectively extract and process the wear original data. The data enters the convolutional layer through the input layer, and the features of the data are extracted through this layer. After data extraction, data preprocessing is performed. Specifically, data cleaning: processing missing values (interpolation or deletion), abnormal values (smoothing or rejection); standardization: scaling features of different dimensions and ranges to [0, 1]; data set division: dividing the preprocessed data into a training set (70%-80%) and a test set (30%-20%); the processed data is compressed through the pooling layer to extract the main features, and then multiple features are output from the pooling layer and input into the BiLSTM model.
[0058] In the second stage, the data optimized by CNN is input into the BiLSTM neural network to predict the pavement anti-wear performance. The BiLSTM model can process forward and backward time series at the same time, and the size of the hidden layer is 50, so that the input feature information can be more comprehensively extracted. Through the calculation of the previous layer, the obtained feature space is mapped to the sample space, and the non-linear expression ability of the network is increased through the activation function (linear) to improve the network prediction performance. After a series of training of CNN and BiLSTM, the prediction result is obtained.
[0059] Referring to Figure 3, the anti-wear performance of the pavement is predicted according to data of an environment-tire-pavement multi-phase coupling wear model, the anti-wear model sets influencing factors as load, wear action times (40,000-1,200,000 times), freeze-thaw (25 times, 50 times, 75 times, 100 times, 150 times, 200 times), temperature gradient (-30℃ / m-60℃ / m), and the average pavement structure depth under the wear action is taken as an output index, the data is obtained from the model data, 75% of the training set is selected, and the prediction result of 25% of the test set data is obtained, the prediction model can well capture the change trend of the data, has high prediction accuracy, the model determination coefficient R2 is 0.98, R 2 represents the explanation ability of the model to the target variable, R 2 The closer to 1, the stronger the prediction ability of the model.
[0060] S3, based on the temperature field, freeze-thaw field, automobile tire and cement concrete pavement factors, the prediction result is analyzed by the SHAP model.
[0061] The prediction result is brought into the SHAP model, the interaction and main control factor of each factor in the neural network prediction model are analyzed, and the index is the Shap value, and the Shap value φ i (f) The calculation formula is:
[0062]
[0063] In the formula: N-all feature set; |S| - feature subset, φ i (f) is the Shap value of any feature i.
[0064] Referring to FIGS. 4(a) and 4(b), the interaction and main control factor between the wear load, wear action times, freeze-thaw and temperature gradient in the prediction model are analyzed by the SHAP model explanation tool, the interaction and main control factor result between each factor, the result shows that under the current working condition, the factor that most affects the pavement anti-wear action is freeze-thaw, followed by action times, and temperature has the least effect on the wear performance, the freeze-thaw-action times interaction synergistic action dominates the pavement wear performance, and the coupling synergistic action of temperature and other factors is not obvious, and in the design of cement concrete pavement in northern areas, attention should be paid to anti-freeze-thaw design.
[0065] In another embodiment, referring to Figure 5 The convolutional bidirectional memory collaborative neural network architecture (CNN-BiLSTM) is compared with typical machine learning algorithms (BP neural network, support vector machine (SVM), CNN neural network, and random forest algorithm (RF)) in stability and accuracy, and the working conditions are the same as in the previous embodiment. According to R 2The different machine learning algorithms are ranked in terms of the prediction ability of the road surface wear performance as CNN-BiLSTM>CNN>RF>SVM>BP, the mean absolute error MAE is the average of the absolute error between the predicted value and the true value, the smaller the MAE, the smaller the prediction error of the model, in terms of prediction error, CNN-BiLSTM has the smallest error, from the two indicators, CNN-BiLSTM has the highest prediction accuracy, which shows that the fused neural network has an advantage in performance prediction.
[0066] It should be noted that, in this document, relational terms such as "first" and "second", and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.
[0067] The above description is merely that of a specific implementation of the present application, enabling a person skilled in the art to understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will accord with the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for predicting the anti-wear performance of a road surface taking into account the environmental-load coupling, characterized by, The application relates to a method for predicting the performance of a cement concrete pavement, and belongs to the field of cement concrete pavement performance prediction. The method comprises the following steps: S1, constructing a multi-phase coupling anti-wear simulation model comprising temperature field, freeze-thaw field, automobile tire and cement concrete pavement factors; S2, using a convolution bidirectional memory collaborative neural network architecture to perform performance prediction on the multi-phase coupling anti-wear simulation model, and outputting a prediction result; 2. The method for predicting the anti-wear performance of a pavement considering environmental-load coupling according to claim 1, wherein S3, based on the temperature field, freeze-thaw field, automobile tire and cement concrete pavement factors, performing weight analysis on the prediction result through a SHAP model. The cement concrete pavement comprises:
3. The method of claim 1, wherein the method is characterized by, 22-26 cm cement concrete surface layer, 2*20 cm cement stabilized macadam base layer and soil base layer, and four square full-size concrete plates with a size of 5m*5m are drawn through ANSYS Workbench, the inter-plate joints of the square full-size concrete plates are connected through force transmission rods, the cross section diameter of the force transmission rods is 30mm, the length is 500mm, the interval is 300mm, the mesh module in Workbench is used to automatically divide the grid, and the grid size is set to be a hexahedral solid grid with a size of 5cm. The automobile tire comprises:
4. The method of claim 1, wherein the method is characterized by, A radial tire is used, and the structure is divided into an airtight layer, a body ply, a belt layer, a crown, a sidewall and a bead, the tire pressure of the automobile tire is set to be 0.75MPa. The temperature field comprises: The range is -30 DEG C / m-60 DEG C / m, the temperature gradient of the cement concrete pavement is calculated according to regional meteorological data through a one-dimensional heat conduction theory, and the one-dimensional heat conduction theory formula is expressed as:
5. The method of claim 1, wherein the method of predicting the pavement wear resistance performance considering the environmental-load coupling is characterized by, In the formula, a is a temperature conduction coefficient, T is air temperature, z is the thickness of the cement concrete pavement, and t is time, and the thickness of the cement concrete pavement is divided into layers according to every 2-4cm, and the temperature gradient of each layer is calculated according to the one-dimensional heat conduction theory formula. The freeze-thaw field comprises:
6. The method of claim 1, wherein the method of predicting the pavement wear resistance performance considering the environmental-load coupling is characterized by, According to the indoor rapid freeze-thaw test, a target cement concrete surface layer material mechanical parameter change curve with indoor freeze-thaw times is obtained, the mechanical parameters include an elastic modulus, a tensile strength and a compressive strength, and the maximum freeze-thaw times of the freeze-thaw test is 300 times. The convolution bidirectional memory collaborative neural network architecture comprises:
7. The method of claim 6, wherein the method is characterized by, A convolution neural network and a bidirectional long short-term memory neural network, original data of the multi-phase coupling anti-wear simulation model are input into the convolution neural network for feature extraction, and the extracted features are input into the bidirectional long short-term memory neural network to output a prediction result. The original data of the multi-phase coupling anti-wear simulation model are input into the convolution neural network for feature extraction, and the method comprises the following steps: S201, performing missing value and abnormal value processing on the original data; S202, performing data standardization processing on the original data; S203, dividing the original data into a training set and a test set respectively; S204, inputting the training set into the convolution neural network for training to obtain the trained convolution neural network; 8. The method of claim 1, wherein the method is characterized by, S205, inputting the test set into a pooling layer of the convolution neural network for feature compression, extracting features, and inputting the features into the bidirectional long short-term memory neural network. The SHAP model comprises: The Shap value in the SHAP model for analyzing the temperature field, freeze-thaw field, automobile tire and cement concrete pavement factors is expressed as: where: N is the set of all features, |S| is the subset of features, i is any of the features of temperature field, freeze-thaw field, automobile tire, and cement concrete pavement, φ i (f) is the Shap value for any feature i.