System and method for calculating and analyzing deformation stress of broken stone hardcore of road embankment

Abstract

The invention relates to the technical field of data analysis, in particular to a system and a method for calculating and analyzing deformation stress of a broken stone hardcore of a road embankment. A road embankment broken stone hardcore deformation stress calculation and analysis system comprises a data acquisition module and a broken stone hardcore deformation stress analysis module. The gradation characteristics of the broken stone hardcore, the characteristics of the pavement layer, the characteristics of the embankment fill stratum and the vehicle load condition vector are sent to the broken stone hardcore deformation stress analysis model for processing, and the corresponding deformation stress analysis image set is output, so that the calculation and analysis of the deformation stress of the broken stone hardcore are realized; traditional finite element analysis software is not needed any more, the limitation of the particle number of the traditional finite element analysis software is not needed, and the calculation and analysis efficiency of the deformation stress of the broken stone hardcore is improved; and the grading change of grading particle crushing is considered in the broken stone hardcore deformation stress analysis model, so that the accuracy of calculation and analysis of the broken stone hardcore deformation stress is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and particularly relates to a system and method for calculating and analyzing the deformation and stress of a crushed stone cushion for a road embankment. Background Art

[0002] The crushed stone cushion of a road embankment is crucial in road engineering. However, when traditional finite element analysis software is used to calculate and analyze the deformation and stress of the crushed stone cushion, especially when simulating a graded crushed stone cushion, in order to accurately reflect the actual particle gradation composition, it is necessary to model a large number of particles with different particle sizes, resulting in a sharp increase in the number of elements in the finite element model and an exponential growth in the calculation scale, exceeding the current computing power. This forces traditional methods to simplify the model. For example, the graded crushed stone cushion is regarded as a homogeneous material or the particle gradation is roughly simplified, ignoring the important influence of the discreteness of graded particles and the gradation characteristics on the deformation and stress characteristics, making the calculation results deviate greatly from the actual situation and difficult to accurately reflect the true stress and deformation state of the crushed stone cushion. Summary of the Invention

[0003] The present invention processes the gradation characteristics of the crushed stone cushion, the characteristics of the road surface layer, the characteristics of the embankment filling layer, and the vehicle load condition vector through a crushed stone cushion deformation and stress analysis model, and outputs a corresponding set of deformation and stress analysis images, realizing the calculation and analysis of the deformation and stress of the crushed stone cushion. There is no need to use traditional finite element analysis software, and there is no need to be limited by the number of particles in traditional finite element analysis software, improving the efficiency of calculating and analyzing the deformation and stress of the crushed stone cushion; and considering the gradation change of graded particle crushing in the crushed stone cushion deformation and stress analysis model, improving the accuracy of calculating and analyzing the deformation and stress of the crushed stone cushion.

[0004] The present invention provides a method for calculating and analyzing the deformation and stress of a crushed stone cushion for a road embankment, including: Obtaining the gradation characteristics of the crushed stone cushion; obtaining the set vehicle load condition vector; obtaining the characteristics of the road surface layer of the road surface layer and the characteristics of the embankment filling layer of the embankment filling layer; Forming the crushed stone cushion deformation and stress analysis data from the gradation characteristics of the crushed stone cushion, the characteristics of the road surface layer, the characteristics of the embankment filling layer, and the vehicle load condition vector, and sending the crushed stone cushion deformation and stress analysis data into the crushed stone cushion deformation and stress analysis model for processing, and outputting a corresponding set of deformation and stress analysis images, where the set of deformation and stress analysis images includes several stress analysis images and deformation analysis images; The deformation and stress analysis model of the gravel cushion is established based on a variational autoencoder, including an encoder layer, a sampling layer, a graded particle time series analysis layer, a decoder layer, and a deformation and stress analysis image set output layer. The encoder layer is used to encode the deformation and stress analysis data of the gravel cushion and output the corresponding mean vector and standard deviation vector; the sampling layer is used to sample the mean vector and standard deviation vector and output the corresponding latent vector; the graded particle time series analysis layer is used to analyze the time series changes of the graded characteristics under the vehicle load condition corresponding to the vehicle load condition vector and output the time series processed stress analysis characteristics; the decoder layer is used to perform a deconvolution operation on the time series processed stress analysis characteristics and output the corresponding deformation and stress analysis image set; the deformation and stress analysis image set output layer is used to output the deformation and stress analysis image set.

[0005] As a preferred aspect, the deformation and stress analysis data of the gravel cushion is sent into the deformation and stress analysis model of the gravel cushion for processing, and the corresponding deformation and stress analysis image set is output. The specific steps are as follows: The deformation and stress analysis data of the gravel cushion is sent into the encoder layer for fully connected processing, and the corresponding mean vector E and standard deviation vector Q are output; In the sampling layer, the latent vector Z is calculated through the following formula: Z = E + Qε, where ε is the random noise sampled from the standard normal distribution; In the graded particle time series analysis layer, there are N LSTM units. The latent vector Z is processed sequentially through N LSTM units, and the time series processed stress analysis characteristics are output; The decoder layer includes several parallel channels. The time series processed stress analysis characteristics are sent into each channel respectively to perform deconvolution operations, and the outputs of all channels are combined to form the deformation and stress analysis image set; The deformation and stress analysis image set output layer is used to output the deformation and stress analysis image set.

[0006] As a preferred aspect, the deformation and stress analysis model of the gravel cushion is trained, including the following steps: Obtain a number of training samples for the analysis of the deformation and force of the gravel cushion. The training samples for the analysis of the deformation and force of the gravel cushion include the data for the analysis of the deformation and force of the gravel cushion. Label the training samples for the analysis of the deformation and force of the gravel cushion through a sequence of deformation and force analysis image sets. The sequence of deformation and force analysis image sets includes a number of deformation and force analysis image sets that change over time. Combine all the labeled training samples for the analysis of the deformation and force of the gravel cushion to form a training set for the analysis of the deformation and force of the gravel cushion. Train the model for the analysis of the deformation and force of the gravel cushion using the training set for the analysis of the deformation and force of the gravel cushion, calculate the loss value, and determine whether the loss value is within a preset range. If the loss value is within the preset range, output the trained model for the analysis of the deformation and force of the gravel cushion; otherwise, continue to train the model for the analysis of the deformation and force of the gravel cushion using the training set for the analysis of the deformation and force of the gravel cushion.

[0007] As a preferred aspect, it also includes simulating and optimizing the hyperparameters in the model for the analysis of the deformation and force of the gravel cushion, specifically including the following steps: Construct a number of hyperparameter sets for the model for the analysis of the deformation and force of the gravel cushion. Each hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion includes the initialized hyperparameters for the model for the analysis of the deformation and force of the gravel cushion; Combine all the hyperparameter sets for the model for the analysis of the deformation and force of the gravel cushion to form a population set, and set the maximum number of iterations; Calculate the fitness value corresponding to each hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion; Iteratively update the population set through a swarm optimization algorithm; Until the number of iterations reaches the maximum number of iterations, output the hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion corresponding to the maximum fitness value to set the hyperparameters in the model for the analysis of the deformation and force of the gravel cushion, and complete the simulation and optimization of the hyperparameters in the model for the analysis of the deformation and force of the gravel cushion.

[0008] As a preferred aspect, calculate the fitness value corresponding to the hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion. The specific calculation method is as follows: Set the hyperparameters in the model for the analysis of the deformation and force of the gravel cushion through the hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion, and then train the model for the analysis of the deformation and force of the gravel cushion using the training set for the analysis of the deformation and force of the gravel cushion. Use the accuracy of the trained model for the analysis of the deformation and force of the gravel cushion as the fitness value corresponding to the hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion.

[0009] As a preferred aspect, the swarm optimization algorithm uses the sparrow search algorithm.

[0010] The present invention also provides a system for calculating and analyzing the deformation and force of a gravel cushion for a road embankment, including: A data acquisition module, configured to obtain the grading characteristics of the crushed stone cushion; obtain the set vehicle load condition vector; obtain the pavement layer characteristics of the pavement layer and the embankment fill layer characteristics of the embankment fill layer; A crushed stone cushion deformation and stress analysis module, configured to form the crushed stone cushion deformation and stress analysis data from the grading characteristics of the crushed stone cushion, the pavement layer characteristics, the embankment fill layer characteristics and the vehicle load condition vector, and send the crushed stone cushion deformation and stress analysis data into a crushed stone cushion deformation and stress analysis model for processing, and output the corresponding deformation and stress analysis image set, where the deformation and stress analysis image set includes a plurality of stress analysis images and deformation analysis images; The crushed stone cushion deformation and stress analysis model is established based on a variational autoencoder, and includes an encoder layer, a sampling layer, a grading particle time series analysis layer, a decoder layer and a deformation and stress analysis image set output layer. The encoder layer is configured to perform encoding processing on the crushed stone cushion deformation and stress analysis data and output the corresponding mean vector and standard deviation vector; the sampling layer is configured to perform sampling operations on the mean vector and the standard deviation vector and output the corresponding latent vector; the grading particle time series analysis layer is configured to analyze the time series changes of the grading characteristics under the vehicle load conditions corresponding to the vehicle load condition vector and output the time series processed stress analysis characteristics; the decoder layer is configured to perform deconvolution operations on the time series processed stress analysis characteristics and output the corresponding deformation and stress analysis image set; the deformation and stress analysis image set output layer is configured to output the deformation and stress analysis image set.

[0011] The present invention has the following advantages: By sending the grading characteristics of the crushed stone cushion, the pavement layer characteristics, the embankment fill layer characteristics and the vehicle load condition vector into the crushed stone cushion deformation and stress analysis model for processing and outputting the corresponding deformation and stress analysis image set, the present invention realizes the calculation and analysis of the deformation and stress of the crushed stone cushion, no longer needs to use traditional finite element analysis software, and is not limited by the number of particles of traditional finite element analysis software, improving the efficiency of the calculation and analysis of the deformation and stress of the crushed stone cushion; and considering the grading change of the grading particle breakage in the crushed stone cushion deformation and stress analysis model, improving the accuracy of the calculation and analysis of the deformation and stress of the crushed stone cushion. Description of the Drawings

[0012] Figure 1 It is a schematic structural diagram of a road embankment crushed stone cushion deformation and stress calculation and analysis system adopted in an embodiment of the present invention. Detailed Embodiments

[0013] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0014] Embodiment 1, a method for calculating and analyzing the deformation and stress of a road embankment crushed stone cushion, includes: Obtain the gradation characteristics of the crushed stone cushion. Here, the gradation characteristics refer to the corresponding characteristic information after configuring the crushed stone cushion of the road embankment into a gradation model, generally including the thickness of the crushed stone cushion, particle shape, particle size distribution, fractal dimension, porosity ratio, and generation method, etc. These gradation characteristics can represent the corresponding gradation model; obtain the set vehicle load condition vector. Here, the vehicle load condition vector characterizes the load conditions of the vehicle during the execution of the deformation and stress analysis, generally including wheel load pressure and action frequency, etc.; and considering that the road embankment will not only include the crushed stone cushion, but generally also include the pavement layer above the crushed stone cushion and the embankment filling layer below the crushed stone cushion, so it is also necessary to obtain the pavement layer characteristics of the pavement layer and the embankment filling layer characteristics of the embankment filling layer. The pavement layer characteristics generally include the thickness of the pavement layer, pavement layer modulus, and generation method, etc. The embankment filling layer characteristics generally include the thickness of the embankment filling layer, embankment filling layer modulus, and generation method. Here, the pavement layer modulus and embankment filling layer modulus refer to the corresponding stiffness data and shear strength data, etc.; Combine the gradation characteristics of the crushed stone cushion, pavement layer characteristics, embankment filling layer characteristics, and vehicle load condition vector to form the deformation and stress analysis data of the crushed stone cushion, and send the deformation and stress analysis data of the crushed stone cushion into the deformation and stress analysis model of the crushed stone cushion for processing, and output the corresponding deformation and stress analysis image set. The deformation and stress analysis image set includes several stress analysis images and deformation analysis images. The stress analysis images are generally vertical direction stress images, horizontal direction stress images, and medium shear stress images, etc. The deformation analysis images are generally vertical displacement images and horizontal displacement images, etc.; The deformation and stress situation of the crushed stone cushion can be directly reflected through the deformation and stress analysis image set, thereby providing assistance for the deformation and stress calculation and analysis of the crushed stone cushion of the road embankment, and not being limited by the number of particles of traditional finite element analysis software; The deformation and stress analysis model of the crushed stone cushion is established based on the variational autoencoder, including an encoder layer, a sampling layer, a grading particle time series analysis layer, a decoder layer, and an output layer of the deformation and stress analysis image set. The encoder layer is used to encode the deformation and stress analysis data of the crushed stone cushion, and output the corresponding mean vector and standard deviation vector. Through the mean vector and standard deviation vector, the deformation and stress analysis data of the crushed stone cushion can be mapped into the latent space, and the Gaussian distribution of the deformation and stress analysis data of the crushed stone cushion mapped in the latent space is represented by the corresponding mean vector and standard deviation vector. The sampling layer is used to sample the mean vector and standard deviation vector, and output the corresponding latent vector. Here, the latent vector can reflect a sample that conforms to the Gaussian distribution in the latent space, and is used for the subsequent construction of the deformation and stress analysis image set. The grading particle time series analysis layer is used to analyze the time series changes of the grading characteristics under the vehicle load condition vector corresponding to the vehicle load condition, and output the time series processed stress analysis characteristics. Under the vehicle load condition, the grading model corresponding to the crushed stone cushion may undergo grading changes due to the crushing of grading particles, which in turn leads to changes in the deformation and stress of the crushed stone cushion. Therefore, the grading particle time series analysis layer is used to fit the grading changes caused by the crushing of grading particles, thereby improving the accuracy of the deformation and stress analysis of the crushed stone cushion. The decoder layer is used to perform deconvolution operations on the time series processed stress analysis characteristics, and output the corresponding deformation and stress analysis image set. It should be noted here that there are several parallel channels in the decoder layer, and each channel performs deconvolution operations. Only the deconvolution parameters corresponding to each channel change with the tasks corresponding to the channels. Each channel here corresponds to the generation of a stress analysis image or a deformation analysis image. The output layer of the deformation and stress analysis image set is used to output the deformation and stress analysis image set; In this application, by sending the grading characteristics, pavement layer characteristics, embankment fill layer characteristics, and vehicle load condition vector of the crushed stone cushion into the deformation and stress analysis model of the crushed stone cushion for processing, and outputting the corresponding deformation and stress analysis image set, the calculation and analysis of the deformation and stress of the crushed stone cushion are realized. There is no need to use traditional finite element analysis software, and there is no need to be limited by the number of particles in the traditional finite element analysis software, which improves the efficiency of the calculation and analysis of the deformation and stress of the crushed stone cushion; and the grading changes caused by the crushing of grading particles are considered in the deformation and stress analysis model of the crushed stone cushion, which improves the accuracy of the calculation and analysis of the deformation and stress of the crushed stone cushion.

[0015] Sending the deformation and stress analysis data of the crushed stone cushion into the deformation and stress analysis model of the crushed stone cushion for processing, and outputting the corresponding deformation and stress analysis image set, specifically includes the following steps: Send the deformation force analysis data of the gravel cushion layer into the encoder layer for fully connected processing, and output the corresponding mean vector E and standard deviation vector Q. It should be noted that the mean vector E and the standard deviation vector Q have the same dimension, and are consistent with the dimension of the latent space. The dimension of the latent space is set by the developer, usually 10; In the sampling layer, the latent vector Z is calculated through the following formula: Z = E + Qε, where ε is the random noise sampled from the standard normal distribution; In the graded particle time series analysis layer, there are N LSTM units. The latent vector Z is processed through N LSTM units in sequence to output the time series processed force analysis features. It should be added that if it is to observe the time series changes of the deformation force analysis image set, the output of each LSTM unit can also be regarded as a time series processed force analysis feature and subsequent processing can be performed; The decoder layer includes several parallel channels. The time series processed force analysis features are sent into each channel respectively to perform deconvolution operations, and the outputs of all channels are combined to form a deformation force analysis image set; The deformation force analysis image set output layer is used to output the deformation force analysis image set.

[0016] Train the gravel cushion layer deformation force analysis model, including the following steps: Obtain a number of training samples for the analysis of the deformation and force of the gravel cushion. The training samples for the analysis of the deformation and force of the gravel cushion include the data for the analysis of the deformation and force of the gravel cushion. Here, the data for the analysis of the deformation and force of the gravel cushion is obtained through the actual calculation and analysis of the deformation and force of the gravel cushion performed by finite element analysis software. The finite element analysis software here can be PFC2D or FLAC software. And since it is the construction of training samples, the number of graded particles can be increased without considering the calculation efficiency. Label the training samples for the analysis of the deformation and force of the gravel cushion through the image set of the analysis of the deformation and force. It should be noted that the image set of the analysis of the deformation and force here is output by the finite element analysis software, and a dynamic crushing algorithm for graded particles can be embedded in the finite element analysis software, triggering crushing based on the contact force and energy threshold between particles, and updating the fractal dimension of the grading after crushing, so that the image set of the analysis of the deformation and force is extended to a sequence of image sets of the analysis of the deformation and force. The sequence of image sets of the analysis of the deformation and force includes several image sets of the analysis of the deformation and force that change with time. Combine all the labeled training samples for the analysis of the deformation and force of the gravel cushion into a training set for the analysis of the deformation and force of the gravel cushion. Train the model for the analysis of the deformation and force of the gravel cushion through the training set for the analysis of the deformation and force of the gravel cushion. During the training, the output of each LSTM unit is regarded as a time-series processed force analysis feature, and subsequent processing is performed, calculate the loss value, and determine whether the loss value is within the preset range. The preset range is determined by the developer. If the loss value is within the preset range, output the trained model for the analysis of the deformation and force of the gravel cushion; otherwise, continue to train the model for the analysis of the deformation and force of the gravel cushion through the training set for the analysis of the deformation and force of the gravel cushion.

[0017] To further improve the accuracy of the model for the analysis of the deformation and force of the gravel cushion, it also includes simulating and optimizing the hyperparameters in the model for the analysis of the deformation and force of the gravel cushion. The specific steps are as follows: Construct several hyperparameter sets for the model for the analysis of the deformation and force of the gravel cushion. The hyperparameter sets for the model for the analysis of the deformation and force of the gravel cushion include various initialized hyperparameters for the model for the analysis of the deformation and force of the gravel cushion, such as the number of fully connected layers of the encoder, the number N of LSTM units, and the learning rate, etc.; Combine all the hyperparameter sets for the model for the analysis of the deformation and force of the gravel cushion into a population set, and set the maximum number of iterations; Calculate the fitness value corresponding to each hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion. The specific calculation method is as follows: Set the hyperparameters in the model for the analysis of the deformation and force of the gravel cushion through the hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion, and then train the model for the analysis of the deformation and force of the gravel cushion through the training set for the analysis of the deformation and force of the gravel cushion. Use the accuracy of the trained model for the analysis of the deformation and force of the gravel cushion as the fitness value corresponding to the hyperparameter set for the model for the analysis of the deformation and force of the gravel cushion; Iteratively update the population set through the sparrow search algorithm; Until the number of iterations reaches the maximum number of iterations, output the set of hyperparameters of the crushed stone cushion deformation and stress analysis model corresponding to the maximum fitness value, set the hyperparameters in the crushed stone cushion deformation and stress analysis model, and complete the simulation optimization of the hyperparameters in the crushed stone cushion deformation and stress analysis model.

[0018] Example 2, a system for calculating and analyzing the deformation and stress of a crushed stone cushion for a road embankment, see Figure 1 , including: A data acquisition module for obtaining the grading characteristics of the crushed stone cushion. Here, the grading characteristics refer to the characteristic information corresponding to the grading model configured for the crushed stone cushion of the road embankment, generally including the thickness of the crushed stone cushion, particle shape, particle size distribution, fractal dimension, porosity ratio, and generation method, etc. These grading characteristics can represent the corresponding grading model; obtain the set vehicle load condition vector. Here, the vehicle load condition vector characterizes the load conditions of the vehicle during the deformation and stress analysis, generally including wheel load pressure and action frequency, etc.; and considering that the road embankment does not only include the crushed stone cushion, it generally also includes the road surface layer above the crushed stone cushion and the embankment fill layer below the crushed stone cushion. Therefore, it is also necessary to obtain the road surface layer characteristics of the road surface layer and the embankment fill layer characteristics of the embankment fill layer. The road surface layer characteristics generally include the thickness of the road surface layer, the modulus of the road surface layer, and the generation method, etc. The embankment fill layer characteristics generally include the thickness of the embankment fill layer, the modulus of the embankment fill layer, and the generation method. Here, the modulus of the road surface layer and the modulus of the embankment fill layer refer to the corresponding stiffness data and shear strength data, etc.; A crushed stone cushion deformation and stress analysis module for forming the crushed stone cushion deformation and stress analysis data from the grading characteristics of the crushed stone cushion, the road surface layer characteristics, the embankment fill layer characteristics, and the vehicle load condition vector, and sending the crushed stone cushion deformation and stress analysis data into the crushed stone cushion deformation and stress analysis model for processing, and outputting the corresponding deformation and stress analysis image set. The deformation and stress analysis image set includes several stress analysis images and deformation analysis images. The stress analysis images are generally vertical direction stress images, horizontal direction stress images, and medium shear stress images, etc. The deformation analysis images are generally vertical displacement images and horizontal displacement images, etc.; through the deformation and stress analysis image set, the deformation and stress conditions of the crushed stone cushion can be directly reflected, thereby providing assistance for the deformation and stress calculation and analysis of the crushed stone cushion of the road embankment, and not being limited by the number of particles of traditional finite element analysis software; The deformation and stress analysis model of the gravel cushion is established based on the variational autoencoder, including an encoder layer, a sampling layer, a grading particle time series analysis layer, a decoder layer, and a deformation and stress analysis image set output layer. The encoder layer is used to encode the deformation and stress analysis data of the gravel cushion, output the corresponding mean vector and standard deviation vector, and map the deformation and stress analysis data of the gravel cushion into the latent space through the mean vector and standard deviation vector. Moreover, the Gaussian distribution of the deformation and stress analysis data of the gravel cushion mapped in the latent space is represented by the corresponding mean vector and standard deviation vector. The sampling layer is used to sample the mean vector and standard deviation vector and output the corresponding latent vector. The latent vector here can reflect a sample that conforms to the Gaussian distribution in the latent space and is used for the subsequent construction of the deformation and stress analysis image set. The grading particle time series analysis layer is used to analyze the time series changes of the grading characteristics under the vehicle load condition corresponding to the vehicle load condition vector and output the time series processed stress analysis characteristics. Under the vehicle load condition, the grading model corresponding to the gravel cushion may undergo grading changes due to the crushing of grading particles, which in turn leads to changes in the deformation and stress of the gravel cushion. Therefore, the grading particle time series analysis layer is used to fit the grading changes caused by the crushing of grading particles, thereby improving the accuracy of the deformation and stress analysis of the gravel cushion. The decoder layer is used to perform a transposed convolution operation on the time series processed stress analysis characteristics and output the corresponding deformation and stress analysis image set. It should be noted here that there are several parallel channels in the decoder layer, and each channel performs a transposed convolution operation. Only the transposed convolution parameters corresponding to each channel change with the task corresponding to the channel. Each channel here corresponds to the generation of a stress analysis image or a deformation analysis image. The deformation and stress analysis image set output layer is used to output the deformation and stress analysis image set.

[0019] It should be understood that those of ordinary skill in the art can make improvements or transformations according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention. The parts not described in detail in this specification belong to the prior art well known to those skilled in the art.

Claims

1. A calculation and analysis method for the deformation and force of a crushed stone cushion on a road embankment, characterized in that, Including: Compose the grading characteristics of the gravel cushion layer, the pavement layer characteristics, the embankment filling soil layer characteristics, and the vehicle load condition vector into the gravel cushion layer deformation and stress analysis data, and send the gravel cushion layer deformation and stress analysis data into the gravel cushion layer deformation and stress analysis model for processing, and output the corresponding deformation and stress analysis image set. The deformation and stress analysis image set includes several stress analysis images and deformation analysis images; The gravel cushion layer deformation and stress analysis model is established based on the variational autoencoder, and includes an encoder layer, a sampling layer, a grading particle time series analysis layer, a decoder layer, and a deformation and stress analysis image set output layer. The encoder layer is used to perform encoding processing on the gravel cushion layer deformation and stress analysis data and output the corresponding mean vector and standard deviation vector; the sampling layer is used to perform sampling operations on the mean vector and the standard deviation vector and output the corresponding latent vector; the grading particle time series analysis layer is used to analyze the time series changes of the grading characteristics under the vehicle load conditions corresponding to the vehicle load condition vector and output the time series processed stress analysis characteristics; the decoder layer is used to perform deconvolution operations on the time series processed stress analysis characteristics and output the corresponding deformation and stress analysis image set; The deformation and stress analysis image set output layer is used to output the deformation and stress analysis image set.

2. The deformation and stress calculation and analysis method of a crushed stone cushion for a road embankment according to claim 1, characterized in that, Sending the gravel cushion layer deformation and stress analysis data into the gravel cushion layer deformation and stress analysis model for processing and outputting the corresponding deformation and stress analysis image set specifically includes the following steps: Send the gravel cushion layer deformation and stress analysis data into the encoder layer for fully connected processing and output the corresponding mean vector E and standard deviation vector Q; In the sampling layer, calculate the latent vector Z through the following formula, Z = E + Qε, where ε is the random noise sampled from the standard normal distribution; In the grading particle time series analysis layer, there are N LSTM units. The latent vector Z is sequentially processed through N LSTM units to output the time series processed stress analysis characteristics; The decoder layer includes several parallel channels. The time series processed stress analysis characteristics are respectively sent into each channel to perform deconvolution operations, and the outputs of all channels are combined into a deformation and stress analysis image set; The deformation and stress analysis image set output layer is used to output the deformation and stress analysis image set.

3. A method for calculating and analyzing the deformation and stress of a crushed stone cushion for a road embankment according to claim 2, characterized in that, Training the gravel cushion layer deformation and stress analysis model includes the following steps: Obtain several gravel cushion layer deformation and stress analysis training samples. The gravel cushion layer deformation and stress analysis training samples include the gravel cushion layer deformation and stress analysis data. The gravel cushion layer deformation and stress analysis training samples are labeled through the deformation and stress analysis image set sequence. The deformation and stress analysis image set sequence includes several deformation and stress analysis image sets that change with time. All the labeled gravel cushion layer deformation and stress analysis training samples are composed into a gravel cushion layer deformation and stress analysis training set. The gravel cushion layer deformation and stress analysis model is trained through the gravel cushion layer deformation and stress analysis training set, calculate the loss value, and judge whether the loss value is within the preset range. If the loss value is within the preset range, output the trained gravel cushion layer deformation and stress analysis model; otherwise, continue to train the gravel cushion layer deformation and stress analysis model through the gravel cushion layer deformation and stress analysis training set.

4. A method for calculating and analyzing the deformation and stress of a crushed stone cushion for a road embankment according to claim 3, characterized in that It also includes simulating and optimizing the hyperparameters in the deformation and stress analysis model of the gravel cushion layer, specifically including the following steps: Construct several sets of hyperparameters for the deformation and stress analysis model of the gravel cushion layer. Each set of hyperparameters for the deformation and stress analysis model of the gravel cushion layer includes initialized hyperparameters for the deformation and stress analysis model of the gravel cushion layer. Combine all sets of hyperparameters for the deformation and stress analysis model of the gravel cushion layer into a population set and set the maximum number of iterations; Calculate the fitness value corresponding to each set of hyperparameters for the deformation and stress analysis model of the gravel cushion layer; Iteratively update the population set through a swarm optimization algorithm; Until the number of iterations reaches the maximum number of iterations, output the set of hyperparameters for the deformation and stress analysis model of the gravel cushion layer corresponding to the maximum fitness value, and set the hyperparameters in the deformation and stress analysis model of the gravel cushion layer to complete the simulation and optimization of the hyperparameters in the deformation and stress analysis model of the gravel cushion layer.

5. A method for calculating and analyzing the deformation and stress of a crushed stone cushion for a road embankment according to claim 4, characterized in that, Calculate the fitness value corresponding to the set of hyperparameters for the deformation and stress analysis model of the gravel cushion layer. The specific calculation method is as follows: Set the hyperparameters in the deformation and stress analysis model of the gravel cushion layer through the set of hyperparameters for the deformation and stress analysis model of the gravel cushion layer, and then train the deformation and stress analysis model of the gravel cushion layer with the deformation and stress analysis training set of the gravel cushion layer. Use the accuracy of the deformation and stress analysis model of the gravel cushion layer after training as the fitness value corresponding to the set of hyperparameters for the deformation and stress analysis model of the gravel cushion layer.

6. The deformation and stress calculation and analysis method of a crushed stone cushion for a road embankment according to claim 5, characterized in that, The swarm optimization algorithm uses the sparrow search algorithm.

7. A calculation and analysis system for the deformation and force of a crushed stone cushion in a road embankment, characterized in that, The system applies the method for calculating and analyzing the deformation and stress of the gravel cushion layer of the road embankment described in any one of claims 1-6 above, including: A data acquisition module, which is used to obtain the gradation characteristics of the gravel cushion layer; obtain the set vehicle load condition vector; obtain the pavement layer characteristics of the pavement layer and the embankment fill layer characteristics of the embankment fill layer; A deformation and stress analysis module of the gravel cushion layer, which is used to form the deformation and stress analysis data of the gravel cushion layer by combining the gradation characteristics, pavement layer characteristics, embankment fill layer characteristics and vehicle load condition vector of the gravel cushion layer, and send the deformation and stress analysis data of the gravel cushion layer into the deformation and stress analysis model of the gravel cushion layer for processing, and output the corresponding deformation and stress analysis image set. The deformation and stress analysis image set includes several stress analysis images and deformation analysis images; The deformation and stress analysis model of the gravel cushion layer is established based on a variational autoencoder, and includes an encoder layer, a sampling layer, a gradation particle time series analysis layer, a decoder layer and a deformation and stress analysis image set output layer. The encoder layer is used to encode the deformation and stress analysis data of the gravel cushion layer and output the corresponding mean vector and standard deviation vector; the sampling layer is used to sample the mean vector and standard deviation vector and output the corresponding latent vector; the gradation particle time series analysis layer is used to analyze the time series changes of the gradation characteristics under the vehicle load conditions corresponding to the vehicle load condition vector and output the time series processed stress analysis characteristics; the decoder layer is used to perform a deconvolution operation on the time series processed stress analysis characteristics and output the corresponding deformation and stress analysis image set; the deformation and stress analysis image set output layer is used to output the deformation and stress analysis image set.

Images