Method and device for evaluating roadbed stability in seasonal frozen soil region and medium

Abstract

The invention provides a seasonal frozen earth region roadbed stability evaluation method, a seasonal frozen earth region roadbed stability evaluation device and a medium, and relates to the technical field of seasonal frozen earth region roadbed stability evaluation. Meanwhile, performing a freeze-thaw cycle experiment in a laboratory, and fitting to obtain a heat conduction coefficient and a moisture diffusion coefficient of the soil sample; inputting field data and experimental parameters into the finite element model, simulating hydrothermal transfer and mechanical behaviors under different seepage and drainage grating arrangement conditions, and analyzing dynamic change rules of a temperature field, a water content field and frost heaving stress; then, a hydrothermal dynamic behavior and rib-soil interface behavior prediction model is trained based on an LSTM network architecture, and hydrothermal changes and mechanical properties of the roadbed under different conditions are predicted; and finally, based on an output result of the prediction model, calculating a roadbed stability index by combining a comprehensive stability evaluation model.

Description

Technical Field

[0001] The invention relates to the technical field of roadbed stability evaluation in seasonal frozen soil areas, and in particular to a method, a device and a medium for evaluating roadbed stability in seasonal frozen soil areas. Background Art

[0002] Due to the influence of seasonal frozen soil, roadbed projects in seasonal frozen soil areas have long faced problems such as uneven settlement, cracking or water damage caused by freeze-thaw cycles. The evaluation of roadbed stability in seasonal frozen soil areas is a key link in ensuring road traffic safety and extending the service life of roads. The significance of the evaluation is to accurately monitor and predict the changes in the mechanical and physical properties of the roadbed during the freeze-thaw cycle, optimize the roadbed design, and prevent engineering quality accidents. Specific usage scenarios include the construction and maintenance of highways, railways, airport runways, etc. in frozen soil environments. The long-term stability of these projects is directly related to the reliability of regional transportation networks and economic development. However, due to the complex dynamic changes in temperature and moisture fields in frozen soil areas, and the close relationship between seepage and drainage performance and the thermal-water coupling of the soil, traditional empirical design methods are often difficult to meet actual needs.

[0003] In the existing technology, methods such as filling permeable materials in the roadbed, laying drainage pipes or widening the roadbed are usually used to solve the stability problems caused by freeze-thaw cycles, and relevant evaluations are carried out through finite element analysis, thermal-water coupling models and other methods. However, these methods have certain limitations: finite element analysis requires a large amount of field test data and model parameters, which is time-consuming and labor-intensive, and it is difficult to achieve real-time evaluation; although laying drainage pipes or permeable materials can alleviate the problem of concentrated distribution of moisture fields to a certain extent, there is a lack of comprehensive analysis of the coupling relationship between temperature fields and moisture fields, and the optimization effect is limited. In addition, due to the lack of precise control of the design parameters of seepage drainage grids, such as horizontal spacing, vertical spacing and material properties, the existing technology is difficult to fully adapt to the complex geological and climatic conditions in frozen areas.

[0004] The deficiency of existing technologies is that they cannot effectively solve the stability problems of roadbed projects in frozen areas caused by the thermal-water coupling effect. In particular, they lack the ability to conduct real-time analysis and optimize the design of the distribution laws of temperature and moisture fields in a dynamic environment. This limitation leads to insufficient adaptability of engineering design parameters and increases the risk of roadbed instability.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a method and device for evaluating roadbed stability in seasonally frozen areas to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The communication path selection method for device node data transmission includes the following specific steps:

[0009] Step 1: Collect soil samples to be tested at the target roadbed site, record the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress, bring the collected soil samples back to the laboratory, simulate freeze-thaw cycle experiments, and obtain the thermal conductivity and water diffusion coefficient of the soil samples through experimental fitting;

[0010] Step 2: Input the collected field observation data and experimental fitting results into the COMSOL finite element model, build a three-dimensional finite element model based on the field observation data and physical experimental parameters, and analyze the dynamic change laws of the temperature field, moisture content field and frost heave stress distribution by simulating the water heat transfer and mechanical behavior under different grid layout conditions;

[0011] Step 3: Use the finite element model to extract the dynamic characteristic data of temperature field, moisture content field and stress distribution, establish a hydrothermal behavior and reinforcement-soil interface behavior prediction model based on the LSTM network architecture, use the dynamic characteristic data as a training set, train the LSTM model, and predict the hydrothermal dynamic changes and mechanical behavior of the roadbed under different grid design conditions;

[0012] Step 4: Based on the output results of the prediction model and combined with the comprehensive stability evaluation model, the stability index of the frozen soil roadbed is calculated, and the design and layout of the seepage and drainage grilles are optimized according to the evaluation results.

[0013] Furthermore, the logic of collecting soil samples to be tested at the target roadbed site for testing and recording the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress is as follows: soil samples of different depths and different soil layer types are collected at the target roadbed site, and the collected soil samples are classified according to their layers and physical properties;

[0014] Temperature and humidity sensors and deformation monitoring equipment are deployed to obtain dynamic change data of the frozen soil layer. Multi-layer monitoring points are set up in the depth direction: 0.2m, 0.5m, and 1.0m to obtain the dynamic gradient of temperature and moisture content of the frozen soil layer. Thermocouple sensors are deployed to record the temperature changes on the roadbed surface and at different depths in real time. TDR (time domain reflectometer) sensors are deployed to record the dynamic changes in the moisture content of the roadbed soil. LVDT (linear variable differential transformer) displacement meters are deployed to monitor the frost heave and thaw deformation of the roadbed surface.

[0015] After preparation, the collected soil samples were made into cylindrical standard specimens with a diameter of 50 mm and a height of 100 mm. The prepared soil samples were placed in an environmental chamber with freezing and thawing control functions to simulate the temperature gradient and freeze-thaw cycle frequency according to the meteorological conditions of the target area. The temperature variation range of the simulated seasonal frozen soil in the target area was: -20°C to 5°C. Before the experiment, the initial moisture content of the soil sample was adjusted according to the actual state of the roadbed, and vertical pressure was applied by a loading test machine to simulate the overlying load conditions of the actual roadbed. Different freeze-thaw cycles were set according to engineering requirements: 10 times, 20 times, to simulate the effects of different freeze-thaw processes on the soil samples.

[0016] The method of bringing the collected soil samples back to the laboratory and simulating freeze-thaw cycle experiments to obtain the thermal conductivity coefficient and water diffusion coefficient of the soil samples through experimental fitting is as follows: measuring the thermal conductivity coefficient k of the soil samples at different water contents through steady-state or non-steady-state heat conduction experiments, studying the variation law of k with water content w, and fitting the empirical formula k=k(w); measuring the diffusion coefficient D of the soil samples under different temperature conditions through water diffusion experiments w , Research D w The law of change with temperature T, and fitting the empirical formula D w =D w (T).

[0017] Furthermore, the logic for calculating and collecting the on-site observation data is as follows: the temperature sensor records the temperature changes of different depths of the soil sample over time and outputs a set of time series data T(z,t T ), where z is the depth data, t T is the time data of temperature change. The temperature gradient is the core parameter of heat conduction and is used to calculate the heat flux in the soil layer. The formula based on the temperature depth gradient is:

[0018]

[0019] in, is the rate of change of temperature with depth, z i+1 and z i is the depth position of the adjacent sensor position, T(z i+1 ) and T(z i ) is the temperature value of the corresponding position, i is the index;

[0020] The moisture content sensor records the change of moisture content in the soil sample over time and outputs a set of time series data w(z, t w ), where z is the depth data, t w is the time data of moisture content change. The moisture content gradient describes the spatial change rate of water distribution in the soil sample. The formula for the moisture content depth gradient is:

[0021]

[0022] Among them, w(z i+1 ) and w(z i ) represent the depth z i+1 and z i The moisture content of i+1 and z i is the depth position of the adjacent sensor, It indicates the rate of change of water content with depth;

[0023] The trend of moisture content over time can be calculated by the time derivative of moisture content, based on the formula:

[0024]

[0025] Among them, w(t w,i+1 ) and w(t w，i ) indicates that at time t w，i+1 and t w，i The moisture content at that time, i represents the index; similarly, we can get It represents the rate of change of soil sample temperature over time;

[0026] The stress sensor records the frost heave force and thaw settlement force of the soil sample during the freeze-thaw cycle, and outputs a stress-time curve σ(τ), based on the formula:

[0027]

[0028] Among them, σ(τ) represents the instantaneous stress during the freeze-thaw cycle at time τ, F represents the measured force, and A represents the stress area;

[0029] The method for obtaining frost heave force data is to extract the maximum value in the frost heave stress curve, which is used to represent the maximum expansion force inside the soil sample during the freezing process. The formula is:

[0030] σ max =max(σ(τ))

[0031] Among them, σ max Indicates the peak value of frost heave force;

[0032] The method for obtaining the melting-settlement force data is to extract the minimum value in the melting-settlement stress curve, which is used to represent the settlement stress of the soil sample during the melting process. The formula is:

[0033] σ min =min(σ(τ))

[0034] Among them, σ min Indicates the peak melting and settling force.

[0035] Furthermore, the logic of inputting the collected field observation data and experimental fitting results into the COMSOL finite element model to construct a three-dimensional finite element model based on the field observation data and physical experimental parameters is:

[0036] The field observation data and indoor experimental data collected in step 1 are interpolated and smoothed, and the processed data are imported into the material property module of COMSOL in the form of formula. The field observation data are used to set the boundary conditions of the temperature field and the moisture content field, and the indoor experimental data are used as the material property input of the model;

[0037] Set T(x, y, z, t T ) represents the temperature field in the soil sample, where x, y, and z are three-dimensional spatial coordinates, representing the position distribution, and represent the width of the roadbed, the longitudinal length of the roadbed, and the position in the depth direction, respectively, and t T The time variable representing the temperature, when z = 0 it corresponds to the roadbed surface, when z>0 it corresponds to the area below the roadbed;

[0038] The thermal conductivity coefficient k = k(w) and the diffusion coefficient D obtained from the indoor experiment w =D w The fitting formula of (T) is input into the material property module of COMSOL, and the Heat Transfer Module in COMSOL is used to simulate the temperature field T (x, y, z, t T ) dynamic changes, the formula based on the heat conduction equation is:

[0039]

[0040] Where ρ is the density of the soil sample, c is the specific heat capacity of the soil sample, k is the thermal conductivity coefficient, which depends on the water content w, and Q is the internal heat source term. Represents the temperature gradient vector data, that is is the rate of change of temperature along the x direction, is the rate of change of temperature along the y direction, is the rate of change of temperature along the z direction;

[0041] The Transport of Diluted Species Module is used to simulate water migration, which describes the internal moisture content field w (x, y, z, t w )'s spatial distribution and dynamic changes over time, the formula based on which the moisture migration equation is used is:

[0042]

[0043] Among them, J wis the water migration flux, the negative sign indicates that water migrates from the high water content area to the low water content area, D w is the water diffusion coefficient, i.e. D w =D w (T), is the water content gradient vector data, that is is the rate of change of moisture content along the x direction, is the rate of change of moisture content along the y direction, is the rate of change of moisture content along the z direction; w(x, y, z, t w ) is the moisture content field, where x, y, and z are three-dimensional spatial coordinates representing the position distribution, and t w Represents the dynamic change of moisture content over time.

[0044] Furthermore, the logic of setting the boundary conditions of the COMSOL finite element model is as follows: The method of setting the temperature boundary conditions is to use the boundary conditions consistent with the ambient temperature change on the roadbed surface, that is, z=0, and directly use the ambient temperature data collected on site as input:

[0045] T s (t T )=T e (t T )

[0046] Among them, T s (t T ) represents the surface temperature of the roadbed, T e (t T ) represents the collected ambient temperature, which changes with time t T Dynamic changes of

[0047] At the bottom of the roadbed, that is, z = z bottom : Assuming no heat is lost through the bottom boundary, set it to adiabatic boundary conditions:

[0048]

[0049] Among them, n represents the normal vector, which indicates the direction perpendicular to the bottom of the roadbed;

[0050] The method for setting the moisture boundary condition is to set the moisture loss boundary as follows: on the roadbed surface, i.e., z=0: considering the evaporation effect,

[0051] J w =f(E)

[0052] Among them, J w represents the water migration flux, f(E) is the evaporation rate function, which is obtained from literature or experimental data and is related to ambient temperature, humidity, and wind speed conditions; at the bottom of the roadbed, that is, z = zbottom : Assuming that water cannot be lost through the bottom of the roadbed, set it as an impermeable boundary condition: J w =0, this condition is used to simulate the assumption that the water flow at the bottom of the roadbed is blocked;

[0053] The method for setting the mechanical boundary conditions is to use the frost heave force and the thaw settlement force as the boundary conditions of the internal stress field, and obtain the maximum frost heave stress σ during freezing through experiments. max and the minimum melting stress σ min The boundary conditions of the input model, frost heave force and thaw settlement force are applied to the nodes on the top surface and side wall of the roadbed respectively to simulate the changes in the internal stress of the soil sample during the freeze-thaw cycle. This setting reflects the distribution law of the actual frost heave stress and analyzes its impact on the stability of the roadbed.

[0054] Furthermore, after the boundary conditions are set, numerical simulation is performed. The heat conduction equation and the water migration equation are coupled through Multiphysics Coupling in COMSOL to simulate the water-heat coupling behavior of the soil sample during the freeze-thaw cycle. The coupling relationship includes:

[0055] Temperature field T(x, y, z, t T ) by affecting the water diffusion coefficient D w (T) Realize the coupling relationship with water migration: D w =D w (T), diffusion coefficient D when temperature rises w Increase, resulting in faster water diffusion;

[0056] Water content field w(x, y, z, t w ) affects the thermal conductivity k and specific heat c of the soil sample, thus acting inversely on the temperature field: k = k(w), c = c(w). When the moisture content increases, both the thermal conductivity and specific heat capacity increase.

[0057] The logic of simulating the water heat transfer and mechanical behavior of different grid arrangements and analyzing the dynamic changes of temperature field and moisture field is as follows: setting multiple groups of simulation conditions to study the influence of different drainage grid arrangement schemes on water heat transfer and mechanical behavior: no grid scheme, as a benchmark comparison, is used to simulate the water heat behavior when there is no drainage grid; horizontal grid arrangement scheme is used to simulate the influence of horizontal grid distribution on water heat transfer; oblique grid arrangement scheme is used to simulate the influence of oblique grid distribution on the internal stability of the roadbed;

[0058] After the simulation is completed, the dynamic changes of three key physical fields are extracted and analyzed: Extract the three-dimensional temperature distribution T (x, y, z, t T), analyze the temperature gradient, freezing depth and freeze-thaw cycle range inside the soil sample, and compare the effects of different grid arrangements on the stability of the soil temperature field;

[0059] Extract the three-dimensional moisture content distribution w(x, y, z, t w ): Analyze the law of water migration, including the migration of water from the frozen area to the non-frozen area, and compare the effects of different grid arrangements on water dissipation capacity; extract the distribution of frost heave stress and thaw settlement stress from the stress field: the peak value of frost heave stress σ max and melting stress peak σ min ,By analyzing the overall stability of the roadbed, the anti-frost heave effect of different grid arrangement schemes was evaluated.

[0060] Furthermore, a prediction model of hydrothermal behavior and reinforcement-soil interface behavior was established based on the LSTM network architecture, with one to two layers of LSTM units, each layer including 50 to 100 LSTM units; the hidden layer used ReLU as the activation function, and the output layer used a linear activation function;

[0061] Obtain the data that affects the hydrothermal behavior in a time series distribution, that is, the time series data of the temperature field, moisture content field and stress distribution, as well as material parameters, including the thermal conductivity k, the water diffusion coefficient D w ; Extract dynamic features based on time series characteristics, including temperature gradient Moisture content gradient and its time derivative Input the feature matrix, use the dynamic feature data as the training set, and train the LSTM model;

[0062] Use Batch training to divide the entire training set into multiple small batches, and divide the data into small batches with a batch size of 64. In each training iteration, pass the input feature data into the LSTM model to calculate the output prediction value of the model. Use the mean square error as the loss function to measure the error between the model prediction value and the true value. Calculate the gradient of the loss function relative to the model parameters through the back propagation algorithm and update the model parameters.

[0063] During the training process, the hyperparameters of the LSTM model are set, and the hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the number of hidden layer neurons and the time step: the number of network layers is set to a two-layer network structure, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 64, the number of training times is set to 500, the number of hidden layer neurons is 32, and the time step is set to 10 time steps;

[0064] The model is gradually optimized and the root mean square error is used on the validation set to evaluate the model performance until the hydrothermal behavior prediction model is trained. The trained LSTM model can predict the hydrothermal behavior indicators under given time and space conditions: heat transfer rate Moisture flux and the behavior of the reinforcement-soil interface: the distribution of temperature gradient and moisture content gradient;

[0065] By comparing the prediction results with and without the seepage and drainage grid, the role of the grid in regulating temperature distribution and moisture content redistribution is quantified, and the regulation effect index R is calculated. T and R w , providing a basis for subsequent optimization design.

[0066] Furthermore, the R T is the temperature regulation coefficient, which is the relative reduction degree of freezing depth fluctuation or temperature gradient change under the condition of no grid and with grid, reflecting the regulation effect of drainage grid on temperature field distribution. The formula is:

[0067]

[0068] Where, ΔT no Indicates the maximum fluctuation range of temperature gradient under no grid condition, ΔT yes It indicates the maximum fluctuation range of temperature gradient under the condition of grid; the regulation effect is quantified by the fluctuation of freezing depth or the amplitude of temperature gradient change, and the formula is as follows:

[0069]

[0070] For each temperature field data under each condition, calculate the temperature gradient The spatial distribution of and record its maximum and minimum values;

[0071] Combining the simulation results without and with the grille, calculate the temperature regulation coefficient R T And analyze its physical meaning: When R T >0, indicating that the drainage grille can effectively reduce the temperature gradient fluctuation, and the regulation effect is better; when R T = 0, indicating that the grid has no effect on temperature distribution; when R T When <0, it means that the grid exacerbates the non-uniformity of the temperature gradient;

[0072] The R w is the moisture content adjustment coefficient, which is the relative reduction degree of moisture content gradient or cumulative water content under the condition of no grid and with grid, reflecting the adjustment effect of drainage grid on moisture content field distribution. The formula is:

[0073]

[0074] Among them, Δw no Indicates the fluctuation range of moisture content gradient or local moisture accumulation under no grid conditions, Δw yes It indicates the fluctuation range of moisture content gradient or local moisture accumulation under the condition of grid. The calculation formula of the fluctuation of moisture content gradient or local moisture accumulation is:

[0075]

[0076] For each condition of the moisture field data, calculate the moisture gradient spatial distribution of

[0077] When R w >0, it means that the drainage grid can effectively reduce the fluctuation of moisture content gradient and improve the uniformity of moisture distribution; when R w = 0, indicating that the grid has no effect on the moisture content distribution; when R w When <0, it means that the grid aggravates the unevenness of moisture distribution;

[0078] The logic of calculating the comprehensive stability evaluation model based on the machine learning analysis results of step 3 is: the adjustment effect index R T and R w , respectively reflect the role of the grid in regulating the temperature gradient fluctuation and the moisture content gradient fluctuation, and quantify the contribution of the grid to the environmental regulation. They reflect the improvement of the grid on the non-uniformity of the temperature field and the moisture content field. Therefore, R T and R w It is a key indicator that directly reflects the thermal stability and water stability of the roadbed; R T >0 and R w >0 indicates that the grid has a positive regulation effect, while R T <0 or R w <0 means that the grille design is improper, which exacerbates the unevenness;

[0079] In order to comprehensively evaluate the thermal-water coupled stability of the roadbed, based on R T and R w The comprehensive roadbed stability score is calculated based on the formula:

[0080] S = α·(1-R T )+β·(1-R w )

[0081] Among them, S is the comprehensive stability score, (1-R T ) represents the residual inhomogeneity that cannot be covered by the temperature regulation effect, (1-R w)It represents the residual non-uniformity that the moisture content adjustment effect fails to cover. α is the importance weight of temperature adjustment, and β is the importance weight of moisture content adjustment. The values of α and β range from 0.4 to 0.6, and α + β = 1, which is adjusted according to whether the actual environment in the frozen soil area is heat-dominated or water-dominated;

[0082] According to the comprehensive evaluation score S, the subgrade stability is divided into the following levels:

[0083] The subgrade stability is extremely high, that is, when S ≤ 0.2: The hydrothermal coupling behavior of the subgrade is highly uniform, with almost no significant non-uniformity, and it is applicable to high-standard frozen soil subgrade projects;

[0084] The subgrade stability is relatively high, that is, when 0.2 < S ≤ 0.4: The hydrothermal coupling behavior of the subgrade is relatively uniform, and the non-uniformity has no significant impact on the engineering performance;

[0085] The subgrade stability is average, that is, when 0.4 < S ≤ 0.6: The subgrade shows a certain degree of non-uniformity, which will affect the long-term stability;

[0086] The subgrade stability is poor, that is, when S > 0.6: The hydrothermal coupling behavior of the subgrade is significantly non-uniform, and the design needs to be optimized to improve the stability;

[0087] Combined with the evaluation result of S, optimize the design parameters of the drainage and drainage grille to further improve the stability of the subgrade, and at the same time consider the economy of the construction cost; Optimization variables: Grille horizontal spacing, the layout distance of the drainage and drainage grille in the horizontal direction. The smaller the horizontal spacing, the stronger the drainage and drainage capacity, and the improvement of R T and R w , and the better the uniform distribution effect of moisture and heat, but the construction cost increases;

[0088] Grille vertical spacing, the layout distance of the drainage and drainage grille in the vertical direction. The smaller the vertical spacing, the more significant the hydrothermal regulation effect of the grille on the deep soil layer, and the further improvement of R T and R w , and the on-site geological conditions and construction feasibility need to be comprehensively considered;

[0089] Grille material properties, including the permeability coefficient, which directly affects the moisture migration ability. The higher the permeability coefficient, the stronger the moisture migration ability, and then the change of R w and compressive strength. The higher the compressive strength, the better the durability; Constraint conditions: The comprehensive stability score S ≤ 0.2, and the project cost is less than or equal to the budget to ensure cost control;

[0090] The optimization design needs to adjust the horizontal spacing, vertical spacing and material properties, the initial material permeability coefficient and compressive strength. Use the LSTM model in step 3 to input the design parameters, predict the distribution of temperature gradient and moisture content gradient, and calculate R T 、Rw and comprehensive stability score S; if S>0.2, it means that the roadbed stability has not reached the target, and the design parameters need to be adjusted, such as reducing the horizontal or vertical spacing, or improving the water permeability of the material, re-entering the prediction model, and updating R T and R w ; By dynamically adjusting the parameters, S is iteratively calculated until the constraints are met, that is, S≤0.2 and the cost is within the budget.

[0091] The present invention also provides a device for evaluating the stability of a roadbed in a seasonally frozen soil region. The system is used to implement the above-mentioned method for evaluating the stability of a roadbed in a seasonally frozen soil region, and specifically includes:

[0092] The data acquisition module is used to collect soil samples to be tested at the target roadbed site, record the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress, bring the collected soil samples back to the laboratory, simulate freeze-thaw cycle experiments, and obtain the thermal conductivity and water diffusion coefficient of the soil samples through experimental fitting;

[0093] Three-dimensional model construction is used to input the collected field observation data and experimental fitting results into the COMSOL finite element model, build a three-dimensional finite element model based on field observation data and physical experimental parameters, and analyze the dynamic changes of temperature field, moisture content field and frost heave stress distribution by simulating water heat transfer and mechanical behavior under different grid layout conditions;

[0094] The model building module is used to extract the dynamic characteristic data of temperature field, moisture content field and stress distribution using the finite element model, establish the hydrothermal behavior and reinforcement-soil interface behavior prediction model based on the LSTM network architecture, use the dynamic characteristic data as the training set, train the LSTM model, and predict the hydrothermal dynamic changes and mechanical behavior of the roadbed under different grid design conditions:

[0095] The stability judgment module is used to calculate the stability index of the frozen soil roadbed based on the output results of the prediction model and combined with the comprehensive stability evaluation model, and optimize the design and layout of the seepage and drainage grilles according to the evaluation results.

[0096] The present invention further provides a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, each step of the above-mentioned method for evaluating roadbed stability in seasonally frozen soil areas is implemented.

[0097] Compared with the prior art, the present invention has the following beneficial effects:

[0098] The present invention collects soil samples to be tested at the target roadbed site, records the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress, and combines laboratory simulated freeze-thaw cycle experiments to fit the thermal conductivity and moisture diffusion coefficient of the soil samples. This process can accurately quantify the thermal-water coupling characteristics of the soil under freeze-thaw cycle conditions based on real on-site data and experimental analysis, providing high-precision input data for subsequent simulation analysis, thereby effectively solving the problem of insufficient parameter acquisition in the prior art.

[0099] The field observation data and experimental fitting parameters are input into the COMSOL finite element model to construct a three-dimensional finite element model. By simulating the water-heat transfer and mechanical behavior under different seepage and drainage grid layout conditions, the distribution laws of temperature field, moisture content field and frost heave stress are dynamically analyzed. The dynamic feature data extracted by the finite element model is used to establish a prediction model for water-heat behavior and reinforcement-soil interface behavior based on the LSTM (long short-term memory) network architecture. With the help of the LSTM network's time series analysis ability for dynamic data, it is possible to comprehensively reflect the water-heat dynamic changes and mechanical behavior of the roadbed under different grid design conditions through training and prediction, thereby achieving high-precision prediction of roadbed performance and overcoming the shortcomings of existing technologies in real-time and prediction accuracy.

[0100] Based on the output results of the prediction model and combined with the comprehensive stability evaluation model, the present invention calculates the stability index of the frozen soil roadbed and optimizes the design and layout of the drainage grilles accordingly; a full-process solution from data acquisition, model construction, behavior prediction to design optimization is constructed. By combining field data with experimental parameters, finite element simulation with deep learning prediction, and comprehensive evaluation with design optimization, the shortcomings of the existing technology in data accuracy, dynamic prediction capability and scientific design optimization are successfully solved, thereby improving the effect of roadbed stability evaluation and optimization design in seasonal frozen soil areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0102] Figure 2 It is a schematic diagram of the overall device module of the present invention. DETAILED DESCRIPTION

[0103] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0104] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0105] Example:

[0106] See also Figure 1 , the present invention provides a technical solution:

[0107] A method for evaluating roadbed stability in seasonally frozen soil areas, comprising the following specific steps:

[0108] Step 1: Collect soil samples to be tested at the target roadbed site, record the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress, bring the collected soil samples back to the laboratory, simulate freeze-thaw cycle experiments, and obtain the thermal conductivity and water diffusion coefficient of the soil samples through experimental fitting;

[0109] The logic of collecting soil samples to be tested at the target roadbed site for testing and recording the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress is as follows: collecting soil samples of different depths and different soil layer types at the target roadbed site, and classifying the collected soil samples according to their layers and physical properties;

[0110] Temperature and humidity sensors and deformation monitoring equipment are deployed to obtain dynamic change data of the frozen soil layer. Multi-layer monitoring points are set up in the depth direction: 0.2m, 0.5m, and 1.0m to obtain the dynamic gradient of temperature and moisture content of the frozen soil layer. Thermocouple sensors are deployed to record the temperature changes on the roadbed surface and at different depths in real time. TDR (time domain reflectometer) sensors are deployed to record the dynamic changes in the moisture content of the roadbed soil. LVDT (linear variable differential transformer) displacement meters are deployed to monitor the frost heave and thaw deformation of the roadbed surface.

[0111] After preparation, the collected soil samples were made into cylindrical standard specimens with a diameter of 50 mm and a height of 100 mm. The prepared soil samples were placed in an environmental chamber with freezing and thawing control functions to simulate the temperature gradient and freeze-thaw cycle frequency according to the meteorological conditions of the target area. The temperature variation range of the simulated seasonal frozen soil in the target area was: -20°C to 5°C. Before the experiment, the initial moisture content of the soil sample was adjusted according to the actual state of the roadbed, and vertical pressure was applied by a loading test machine to simulate the overlying load conditions of the actual roadbed. Different freeze-thaw cycles were set according to engineering requirements: 10 times, 20 times, to simulate the effects of different freeze-thaw processes on the soil samples.

[0112] The method of bringing the collected soil samples back to the laboratory and simulating freeze-thaw cycle experiments to obtain the thermal conductivity coefficient and water diffusion coefficient of the soil samples through experimental fitting is as follows: measuring the thermal conductivity coefficient k of the soil samples at different water contents through steady-state or non-steady-state heat conduction experiments, studying the variation law of k with water content w, and fitting the empirical formula k=k(w); measuring the diffusion coefficient D of the soil samples under different temperature conditions through water diffusion experiments w , Research D w The law of change with temperature T, and fitting the empirical formula D w =D w (T);

[0113] The steady-state heat conduction experiment refers to the method of measuring the heat conduction coefficient k when thermal equilibrium is reached, which is suitable for heat conduction behavior under a stable environment; the unsteady-state heat conduction experiment calculates the heat conduction coefficient k by measuring the temperature response of the soil sample under dynamic heating conditions;

[0114] The logic for calculating and collecting field observation data is as follows: the temperature sensor records the temperature changes at different depths of the soil sample over time and outputs a set of time series data T(z, t T ), where z is the depth data, t T is the time data of temperature change. The temperature gradient is the core parameter of heat conduction and is used to calculate the heat flux in the soil layer. The formula based on the temperature depth gradient is:

[0115]

[0116] in, is the rate of change of temperature with depth, z i+1 and z i is the depth position of the adjacent sensor position, T(z i+1 ) and T(z i ) is the temperature value of the corresponding position, i is the index;

[0117] It indicates the rate of change of temperature with depth, that is, the temperature depth gradient, which reflects the direction and intensity of heat transfer in the soil layer. The temperature gradient is the driving force of heat conduction. The larger the temperature gradient, the more significant the temperature difference and the faster the heat transfer rate. In roadbed design, it is used to calculate heat flux.

[0118] T(z i+1 )-T(z i ) means that the greater the temperature difference between two points, the greater the temperature gradient; z i+1 -z i The smaller the depth spacing, the more accurate the gradient calculation, but if the spacing is too small, temperature differences may be difficult to detect;

[0119] The moisture content sensor records the change of moisture content in the soil sample over time and outputs a set of time series data w(z, t w ), where z is the depth data, t w is the time data of moisture content change. The moisture content gradient describes the spatial change rate of water distribution in the soil sample. The formula for the moisture content depth gradient is:

[0120]

[0121] Among them, w(z i+1 ) and w(z i ) represent the depth z i+1 and z i The moisture content of i+1 and z i is the depth position of the adjacent sensor;

[0122] It indicates the rate of change of moisture content with depth, reflecting the distribution law of water in the soil layer. The larger the value, the stronger the unevenness of water distribution and the more significant the water migration. The moisture content gradient is the driving force of water migration and is used to calculate water flux. w(z i+1 )-w(z i ) means that the greater the difference in moisture content between the two points, the greater the gradient; z i+1 -z i The smaller the depth spacing, the more accurate the gradient calculation;

[0123] The trend of moisture content over time can be calculated by the time derivative of moisture content, based on the formula:

[0124]

[0125] Among them, w(t w,i+1 ) and w(t w，i ) indicates that at time t w,i+1 and tw，i The moisture content at that time, i represents the index;

[0126] It indicates the rate of change of moisture content over time, reflecting the dynamic change of moisture content in soil samples. It is used to analyze the moisture migration behavior of soil layers at different time stages. The larger the value, the more drastic the moisture change, which may be related to changes in ambient temperature or external water supply conditions. w,i+1 -t w，i The shorter the time interval, the more accurate the trend of change

[0127] The stress sensor records the frost heave force and thaw settlement force of the soil sample during the freeze-thaw cycle, and outputs a stress-time curve σ(τ), based on the formula:

[0128]

[0129] Among them, σ(τ) represents the instantaneous stress during the freeze-thaw cycle at time τ, F represents the measured force, and A represents the stress area. The larger the measured force F is, the greater the stress is, and the smaller the stress area A is, the greater the stress per unit area is.

[0130] The method for obtaining frost heave force data is to extract the maximum value in the frost heave stress curve, which is used to represent the maximum expansion force inside the soil sample during the freezing process. The formula is:

[0131] σ max =max(σ(τ))

[0132] Among them, σ max Indicates the peak value of frost heave force;

[0133] The method for obtaining the melting-settlement force data is to extract the minimum value in the melting-settlement stress curve, which is used to represent the settlement stress of the soil sample during the melting process. The formula is:

[0134] σ min =min(σ(τ))

[0135] Among them, σ min Indicates the peak melting and sinking force;

[0136] σ(τ) represents the instantaneous stress of soil sample during freezing and thawing. Frost heave and thaw settlement are important parameters affecting roadbed stability, which directly affect the deformation and bearing capacity of roadbed. max The larger the value, the stronger the frost heave effect and the higher the risk of structural damage; min The smaller it is, the more significant the melting and sedimentation effect is.

[0137] Step 2: Input the collected field observation data and experimental fitting results into the COMSOL finite element model, build a three-dimensional finite element model based on the field observation data and physical experimental parameters, and analyze the dynamic change laws of the temperature field, moisture content field and frost heave stress distribution by simulating the water heat transfer and mechanical behavior under different grid layout conditions;

[0138] The logic of inputting the collected field observation data and experimental fitting results into the COMSOL finite element model to construct a three-dimensional finite element model based on the field observation data and physical experimental parameters is as follows:

[0139] The field observation data and indoor experimental data collected in step 1 are interpolated and smoothed, and the processed data are imported into the material property module of COMSOL in the form of formula. The field observation data are used to set the boundary conditions of the temperature field and the moisture content field, and the indoor experimental data are used as the material property input of the model;

[0140] The field observation data is used to reflect the boundary conditions of the subgrade temperature field and moisture field in the real natural environment, ensuring that the model has authentic input conditions. The indoor experimental data is used to obtain the thermal conductivity coefficient k and moisture diffusion coefficient D through indoor experiments. w Parameters such as these provide key physical properties of the soil sample, ensuring that the model is sufficiently accurate in describing the material properties;

[0141] Interpolation and smoothing Interpolate and smooth discrete experimental or field data to make them more suitable for input into the finite element model and avoid numerical oscillation or calculation errors;

[0142] Set T(x, y, z, t T ) represents the temperature field in the soil sample, where x, y, and z are three-dimensional spatial coordinates, representing the position distribution, and represent the width of the roadbed, the longitudinal length of the roadbed, and the position in the depth direction, respectively, and t T represents the time variable of temperature. When z=0, it corresponds to the roadbed surface, and when z>0, it corresponds to the area below the roadbed. x represents the coordinate position of the soil sample in the horizontal direction, that is, the width direction of the roadbed. y represents the coordinate position of the soil sample in the other axis of the horizontal direction, that is, the longitudinal direction of the roadbed, which is the coordinate position along the road. z represents the coordinate position of the soil sample in the vertical direction, that is, the depth direction.

[0143] The thermal conductivity coefficient k = k(w) and the diffusion coefficient D obtained from the indoor experiment w =D w The fitting formula of (T) is input into the material property module of COMSOL, and the Heat Transfer Module in COMSOL is used to simulate the temperature field T (x, y, z, t T ) dynamic changes, the formula based on the heat conduction equation is:

[0144]

[0145] Where ρ is the density of the soil sample, c is the specific heat capacity of the soil sample, k is the thermal conductivity coefficient, which depends on the water content w, and Q is the internal heat source term. Represents the temperature gradient vector data, that is is the rate of change of temperature along the x direction, is the rate of change of temperature along the y direction, is the rate of change of temperature along the z direction;

[0146] It indicates the rate of change of soil sample temperature over time, reflecting how the soil layer dynamically responds to external heat input or output. It is used to simulate the dynamic change of temperature during the freeze-thaw cycle and predict the freezing depth and heat propagation law. The larger the value, the faster the temperature changes, and the more significant the impact of heat conduction or ambient temperature changes on the soil layer.

[0147] Describes the contribution of heat conduction to temperature change. The thermal conductivity coefficient k depends on the moisture content w. The higher the moisture content, the stronger the thermal conductivity of the soil sample. The larger k is, the faster the heat conduction and the temperature change rate. The larger the value, the greater the internal heat source term Q, which indicates the heat generation or absorption inside the soil sample. The larger the value, the more significant the increase in heat inside the soil sample and the greater the temperature change rate.

[0148] The Transport of Diluted Species Module is used to simulate water migration, which describes the internal moisture content field w (x, y, z, t w )'s spatial distribution and dynamic changes over time, the formula based on which the moisture migration equation is used is:

[0149]

[0150] Among them, J w is the water migration flux, the negative sign indicates that water migrates from the high water content area to the low water content area, D w is the water diffusion coefficient, i.e. D w =D w (T), is the water content gradient vector data, that is is the rate of change of moisture content along the x direction, is the rate of change of moisture content along the y direction, is the rate of change of moisture content along the z direction; w(x, y, z, t w ) is the moisture content field, where x, y, and z are three-dimensional spatial coordinates representing the position distribution, and tw Indicates the dynamic change of moisture content over time;

[0151] It indicates the rate of change of soil moisture content over time, reflecting the dynamic migration of water during freezing and thawing. It is used to evaluate the impact of freeze-thaw cycles on soil moisture distribution and predict the migration law of water from frozen areas to non-frozen areas. The larger the value, the more drastic the moisture change and the higher the moisture migration rate. is the divergence of the water migration flux, indicating the net outflow of water in a certain area. The larger the outflow, The larger it is, the faster the water content decreases;

[0152] J w It indicates the migration rate of water through a unit area. The negative sign indicates that water migrates from a high-water content area to a low-water content area. It is used to calculate the direction and intensity of water flow inside the soil sample and analyze the drainage capacity of the roadbed. The larger the value, the faster the water migration and the larger the water gradient or diffusion coefficient. The water diffusion coefficient D w , depends on temperature T, when the temperature increases D w Increase, the rate of water migration increases;

[0153] The logic of setting the boundary conditions of the COMSOL finite element model is as follows: The method of setting the temperature boundary conditions is to use the boundary conditions consistent with the ambient temperature changes on the roadbed surface, that is, z=0, and directly use the ambient temperature data collected on site as input:

[0154] T s (t T )=T e (t T )

[0155] Among them, T s (t T ) represents the surface temperature of the roadbed, T e (t T ) represents the collected ambient temperature, which changes with time t T Dynamic changes of

[0156] Directly use the ambient temperature as the surface boundary condition to ensure that the model input is consistent with the actual working conditions. The higher the value, the more significant the impact of the external temperature on the roadbed surface.

[0157] At the bottom of the roadbed, z = z bottom : Assuming no heat is lost through the bottom boundary, set it to adiabatic boundary conditions:

[0158]

[0159] Among them, n represents the normal vector, which indicates the direction perpendicular to the bottom of the roadbed; the physical meaning of the adiabatic boundary condition is to prevent heat from being transferred through the bottom boundary, simulating the engineering assumption that the bottom of the roadbed is insulated; a value of 0 means that there is no heat exchange at the bottom, and the temperature change is only driven by heat transfer from the top and side walls;

[0160] The method for setting the moisture boundary condition is to set the moisture loss boundary as follows: on the roadbed surface, i.e., z=0: considering the evaporation effect,

[0161] J w =f(E)

[0162] Among them, J w represents the water migration flux, f(E) is the evaporation rate function, which is obtained from literature or experimental data and is related to ambient temperature, humidity, and wind speed conditions. The evaporation rate function f(E) is used to describe the water loss boundary and simulate the actual evaporation conditions. The greater the evaporation rate, the faster the water loss, and the more significant the impact on the water distribution of the roadbed.

[0163] At the bottom of the roadbed, z = z bottom : Assuming that water cannot be lost through the bottom of the roadbed, set it as an impermeable boundary condition: J w =0, this condition is used to simulate the assumption that the water flow at the bottom of the roadbed is blocked, preventing the water from escaping through the bottom boundary and ensuring that the water migrates only from the surface and side walls;

[0164] The method for setting the mechanical boundary conditions is to use the frost heave force and the thaw settlement force as the boundary conditions of the internal stress field, and obtain the maximum frost heave stress σ during freezing through experiments. max and the minimum melting stress σ min Input model, boundary conditions of frost heave force and thaw settlement force are applied to nodes on the top surface and side wall of the roadbed respectively to simulate the changes of internal stress of soil samples during freeze-thaw cycles. This setting reflects the distribution law of actual frost heave stress and analyzes its influence on roadbed stability.

[0165] The maximum expansion stress σ generated on the top surface of the roadbed during the simulated freezing process max , reflects the expansion effect of the frozen area. The larger the value, the stronger the frost heave effect and the lower the roadbed stability. The minimum settlement stress σ generated on the top surface of the roadbed during the simulated melting process min , reflects the settlement characteristics of the soil sample. The smaller the value, the more significant the melting settlement effect and the worse the roadbed stability.

[0166] After the boundary conditions are set, numerical simulation is performed. In COMSOL, the heat conduction equation and the water migration equation are coupled through Multiphysics Coupling to simulate the water-heat coupling behavior of the soil sample during the freeze-thaw cycle. The coupling relationship includes:

[0167] Temperature field T(x, y, z, t T ) by affecting the water diffusion coefficient D w (T) Realize the coupling relationship with water migration: D w =D w (T), diffusion coefficient D when temperature rises w Increase, resulting in faster water diffusion;

[0168] Water content field w(x, y, z, t w ) affects the thermal conductivity k and specific heat c of the soil sample, thus acting inversely on the temperature field: k = k(w), c = c(w). When the moisture content increases, both the thermal conductivity and specific heat capacity increase.

[0169] Temperature field affects the moisture migration coefficient D w , and the moisture content in turn affects the thermal conductivity k and specific heat c. This two-way coupling simulates the complex interaction of the hydrothermal behavior of the roadbed during the freeze-thaw cycle.

[0170] The logic of simulating the water heat transfer and mechanical behavior of different grid arrangements and analyzing the dynamic changes of temperature field and moisture field is as follows: setting multiple groups of simulation conditions to study the influence of different drainage grid arrangement schemes on water heat transfer and mechanical behavior: no grid scheme, as a benchmark comparison, is used to simulate the water heat behavior when there is no drainage grid; horizontal grid arrangement scheme is used to simulate the influence of horizontal grid distribution on water heat transfer; oblique grid arrangement scheme is used to simulate the influence of oblique grid distribution on the internal stability of the roadbed;

[0171] After the simulation is completed, the dynamic changes of three key physical fields are extracted and analyzed: Extract the three-dimensional temperature distribution T (x, y, z, t T ), analyze the temperature gradient, freezing depth and freeze-thaw cycle range inside the soil sample, and compare the effects of different grid arrangements on the stability of the soil temperature field;

[0172] Extract the three-dimensional moisture content distribution w(x, y, z, t w ): Analyze the law of water migration, including the migration of water from the frozen area to the non-frozen area, and compare the effects of different grid arrangements on water dissipation capacity; extract the distribution of frost heave stress and thaw settlement stress from the stress field: the peak value of frost heave stress σ max and melting stress peak σ min ,By analyzing the overall stability of the roadbed, the anti-frost heave effect of different grid arrangement schemes was evaluated.

[0173] Step 3: Use the finite element model to extract the dynamic characteristic data of temperature field, moisture content field and stress distribution, establish a hydrothermal behavior and reinforcement-soil interface behavior prediction model based on the LSTM network architecture, use the dynamic characteristic data as a training set, train the LSTM model, and predict the hydrothermal dynamic changes and mechanical behavior of the roadbed under different grid design conditions;

[0174] Based on the LSTM network architecture, a prediction model for hydrothermal behavior and reinforcement-soil interface behavior is established. One to two layers of LSTM units are set, and each layer includes 50 to 100 LSTM units. The hidden layer uses ReLU as the activation function, and the output layer uses a linear activation function.

[0175] Obtain the data that affects the hydrothermal behavior in a time series distribution, that is, the time series data of the temperature field, moisture content field and stress distribution, as well as material parameters, including the thermal conductivity k, the water diffusion coefficient D w ; Extract dynamic features based on time series characteristics, including temperature depth gradient Water content depth gradient and its time derivative Input the feature matrix, use the dynamic feature data as the training set, and train the LSTM model;

[0176] Use Batch training to divide the entire training set into multiple small batches, and divide the data into small batches with a batch size of 64. In each training iteration, pass the input feature data into the LSTM model to calculate the output prediction value of the model. Use the mean square error as the loss function to measure the error between the model prediction value and the true value. Calculate the gradient of the loss function relative to the model parameters through the back propagation algorithm and update the model parameters.

[0177] During the training process, the hyperparameters of the LSTM model are set, and the hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the number of hidden layer neurons and the time step: the number of network layers is set to a two-layer network structure, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 64, the number of training times is set to 500, the number of hidden layer neurons is 32, and the time step is set to 10 time steps;

[0178] The model is gradually optimized and the root mean square error is used on the validation set to evaluate the model performance until the hydrothermal behavior prediction model is trained. The trained LSTM model can predict the hydrothermal behavior indicators under given time and space conditions: heat transfer rate Moisture flux and the behavior of the reinforcement-soil interface: the distribution of temperature gradient and moisture content gradient;

[0179] q represents the heat transfer rate per unit area per unit time, which is used to analyze the freezing depth and the heat dissipation performance of the soil. The larger its value, the higher the efficiency of heat transfer through the soil. k is the thermal conductivity coefficient, which depends on the water content w. The larger k is, the higher the heat transfer rate q is. is the temperature gradient. The larger the gradient, the higher the heat transfer rate q.

[0180] J w It indicates the water flow rate per unit area per unit time, and is used to analyze the law of water migration and its impact on soil stability. The larger the value, the more significant the water flow effect. w is the diffusion coefficient, which depends on the temperature T. As the temperature increases, D w Increase, moisture flux J w Increase; is the moisture content gradient. The larger the gradient, the higher the moisture flux J w The higher;

[0181] By comparing the prediction results with and without the seepage and drainage grid, the role of the grid in regulating temperature distribution and moisture content redistribution is quantified, and the regulation effect index R is calculated. T and R w , providing a basis for subsequent optimization design;

[0182] The R T is the temperature regulation coefficient, which is the relative reduction degree of freezing depth fluctuation or temperature gradient change under the condition of no grid and with grid, reflecting the regulation effect of drainage grid on temperature field distribution. The formula is:

[0183]

[0184] Where, ΔT no It indicates the maximum fluctuation range of temperature gradient under the condition of no grid. The larger the range, the greater the temperature difference inside the soil. yes It indicates the maximum fluctuation range of temperature gradient under the condition of grid. The smaller it is, the better the grid adjustment effect is. T It reflects the regulating effect of the drainage grid on the temperature field distribution and quantifies the relative reduction degree of freezing depth fluctuation or temperature gradient change;

[0185] ΔT no The larger the value, the stronger the inhomogeneity of the temperature field itself; R T The larger the absolute value of ΔT yes The smaller the value, the more significant the adjustment effect of the grille. T The bigger;

[0186] The regulation effect is quantified by the fluctuation of freezing depth or the amplitude of temperature gradient change, as follows:

[0187]

[0188] For each temperature field data under each condition, calculate the temperature gradient The spatial distribution of and record its maximum and minimum values;

[0189] Combining the simulation results without and with the grille, calculate the temperature regulation coefficient R T And analyze its physical meaning: When R T >0, indicating that the drainage grille can effectively reduce the temperature gradient fluctuation, and the regulation effect is better; when R T = 0, indicating that the grid has no effect on temperature distribution; when R T When <0, it means that the grid exacerbates the non-uniformity of the temperature gradient;

[0190] The R w is the moisture content adjustment coefficient, which is the relative reduction degree of moisture content gradient or cumulative water content under the condition of no grid and with grid, reflecting the adjustment effect of drainage grid on moisture content field distribution. The formula is:

[0191]

[0192] Among them, Δw no It indicates the fluctuation range of moisture content gradient or local moisture accumulation under the condition of no grid. The larger the value, the stronger the unevenness of moisture migration. yes It indicates the fluctuation range of moisture content gradient or local moisture accumulation under the condition of grid. The smaller it is, the better the grid adjustment effect is. w It reflects the regulating effect of the drainage grid on the water content field distribution and quantifies the relative reduction of gradient fluctuation or local water accumulation;

[0193] Δw no The larger the value, the stronger the inhomogeneity of the moisture content field; R w The larger the absolute value of Δw yes The smaller the value, the more significant the adjustment effect of the grille. w The bigger;

[0194] The calculation formula for the moisture content gradient or the fluctuation of the local moisture accumulation is:

[0195]

[0196] For each condition of the moisture field data, calculate the moisture gradient spatial distribution of

[0197] When R w >0, it means that the drainage grid can effectively reduce the fluctuation of moisture content gradient and improve the uniformity of moisture distribution; when R w= 0, indicating that the grid has no effect on the moisture content distribution; when R w When <0, it means that the grid aggravates the unevenness of moisture distribution.

[0198] Step 4: Based on the output results of the prediction model and combined with the comprehensive stability evaluation model, calculate the stability index of the frozen soil roadbed, and optimize the design and layout of the seepage and drainage grilles according to the evaluation results;

[0199] The logic of calculating the comprehensive stability evaluation model based on the machine learning analysis results of step 3 is: the adjustment effect index R T and R w , respectively reflect the role of the grid in regulating the temperature gradient fluctuation and the moisture content gradient fluctuation, and quantify the contribution of the grid to the environmental regulation. They reflect the improvement of the grid on the non-uniformity of the temperature field and the moisture content field. Therefore, R T and R w It is a key indicator that directly reflects the thermal stability and water stability of the roadbed; R T >0 and R w >0 indicates that the grid has a positive regulation effect, while R T <0 or R w <0 means that the grille design is improper, which exacerbates the unevenness;

[0200] In order to comprehensively evaluate the thermal-water coupled stability of the roadbed, based on R T and R w The comprehensive roadbed stability score is calculated based on the formula:

[0201] S=α·(1-R T )+β·(1-R w )

[0202] Among them, S is the comprehensive stability score, (1-R T ) represents the residual inhomogeneity that cannot be covered by the temperature regulation effect, (1-R w ) represents the residual heterogeneity that cannot be covered by the moisture content adjustment effect, α is the importance weight of temperature adjustment, β is the importance weight of moisture content adjustment, the value range of α and β is 0.4-0.6, and α+β=1, which is adjusted according to whether the actual environment in the frozen soil area is heat-dominated or water-dominated;

[0203] S reflects the overall hydrothermal coupling stability of the roadbed. It is a key indicator for comprehensively evaluating the effects of temperature regulation and moisture content regulation. It is used to quantify the overall optimization effect of the grid and guide the design and construction of the roadbed. The larger the value, the more residual inhomogeneity in the temperature and moisture content fields, and the more unstable the hydrothermal coupling behavior of the roadbed. The smaller the value, the less residual inhomogeneity in the temperature and moisture content fields, and the more stable the roadbed.

[0204] RT The larger the temperature regulation coefficient, the better the temperature gradient fluctuation regulation effect. T The larger the value, the larger (1-R T ) is smaller, the smaller the comprehensive stability score S is, and the more stable the roadbed is; R w The larger the moisture content adjustment coefficient is, the better the moisture content gradient fluctuation adjustment effect is. w The larger the value, the larger (1-R w ) is smaller, the smaller the comprehensive stability score S is, and the more stable the roadbed is;

[0205] α and β represent the importance weights of temperature regulation and moisture content regulation, respectively. The larger α is, the higher the proportion of temperature regulation in S; the larger β is, the higher the proportion of moisture content regulation in S;

[0206] If the temperature field fluctuation in the frozen soil area has a more significant impact on the stability of the roadbed, such as in high-cold areas, where temperature changes cause large fluctuations in the freezing depth, the weight of α should be increased, and the importance of adjusting the temperature should be given priority; if the gradient fluctuation of soil moisture content has a more significant impact on the stability of the roadbed, such as in humid areas or areas with abundant groundwater, where local moisture accumulation causes frost heave or thaw settlement, the weight of β should be increased, and the importance of adjusting the moisture content should be given priority;

[0207] If ΔT>Δw, it means that temperature change is the main influencing factor, and consider setting α=0.6 and β=0.4; if Δw>ΔT: it means that moisture content change is the main influencing factor, and consider setting α=0.4 and β=0.6; based on engineering experience or on-site investigation, confirm the order of importance between adjusting temperature and adjusting moisture content;

[0208] According to the comprehensive evaluation score S, the roadbed stability is divided into the following levels:

[0209] The roadbed stability is extremely high, that is, when S≤0.2: the water-heat coupling behavior of the roadbed is highly uniform, with almost no significant inhomogeneity, and is suitable for high-standard frozen soil roadbed projects;

[0210] The roadbed stability is high, that is, when 0.2<S≤0.4: the water-thermal coupling behavior of the roadbed is relatively uniform, and the heterogeneity has no significant effect on the engineering performance;

[0211] The roadbed stability is average, that is, when 0.4<S≤0.6: the roadbed shows a certain degree of unevenness, which will affect the long-term stability;

[0212] The roadbed stability is poor, that is, when S>0.6: the water-thermal coupling behavior of the roadbed is significantly uneven, and the design needs to be optimized to improve stability;

[0213] Combined with the evaluation results of S, the design parameters of the seepage and drainage grilles are optimized to further improve the stability of the roadbed while considering the economic efficiency of the construction cost; the optimization variables are: the horizontal spacing of the grilles, the arrangement distance of the seepage and drainage grilles in the horizontal direction. The smaller the horizontal spacing, the stronger the seepage and drainage capacity, and the higher the R T and R w , the more evenly the moisture and heat are distributed, the better, but the construction cost increases;

[0214] The vertical spacing of the grid is the vertical arrangement distance of the drainage grid. The smaller the vertical spacing, the more significant the water and heat regulation effect of the grid on the deep soil, further improving R T and R w , it is necessary to comprehensively consider the on-site geological conditions and construction feasibility;

[0215] The properties of the grid material, including the water permeability coefficient, directly affect the water migration capacity. The higher the water permeability coefficient, the stronger the water migration capacity, which in turn changes the R w and compressive strength. The higher the compressive strength, the better the durability. Constraints: comprehensive stability score S≤0.2, engineering cost is less than or equal to the budget, ensuring cost controllability;

[0216] The optimization design requires adjusting the horizontal spacing, vertical spacing and material properties, the initial material permeability coefficient and compressive strength, using the LSTM model in step 3 to input the design parameters, predict the distribution of temperature gradient and moisture content gradient, and calculate R T , R w and comprehensive stability score S; if S>0.2, it means that the roadbed stability has not reached the target, and the design parameters need to be adjusted, such as reducing the horizontal or vertical spacing, or improving the water permeability of the material, re-entering the prediction model, and updating R T and R w ; By dynamically adjusting the parameters, S is iteratively calculated until the constraints are met, that is, S≤0.2 and the cost is within the budget.

[0217] See also Figure 2 The present invention also provides a device for evaluating the stability of a roadbed in a seasonally frozen soil region. The system is used to implement the above-mentioned method for evaluating the stability of a roadbed in a seasonally frozen soil region, and specifically includes:

[0218] The data acquisition module is used to collect soil samples to be tested at the target roadbed site, record the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress, bring the collected soil samples back to the laboratory, simulate freeze-thaw cycle experiments, and obtain the thermal conductivity and water diffusion coefficient of the soil samples through experimental fitting;

[0219] Three-dimensional model construction is used to input the collected field observation data and experimental fitting results into the COMSOL finite element model, build a three-dimensional finite element model based on field observation data and physical experimental parameters, and analyze the dynamic changes of temperature field, moisture content field and frost heave stress distribution by simulating water heat transfer and mechanical behavior under different grid layout conditions;

[0220] The model building module is used to extract the dynamic characteristic data of temperature field, moisture content field and stress distribution using the finite element model, establish the hydrothermal behavior and reinforcement-soil interface behavior prediction model based on the LSTM network architecture, use the dynamic characteristic data as the training set, train the LSTM model, and predict the hydrothermal dynamic changes and mechanical behavior of the roadbed under different grid design conditions:

[0221] The stability judgment module is used to calculate the stability index of the frozen soil roadbed based on the output results of the prediction model and combined with the comprehensive stability evaluation model, and optimize the design and layout of the seepage and drainage grilles according to the evaluation results.

[0222] In order to solve the above technical problems, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for evaluating roadbed stability in seasonally frozen soil areas are implemented.

[0223] Finally, the present invention further provides a device, which is used to implement the above-mentioned method for evaluating roadbed stability in seasonally frozen areas, comprising:

[0224] A processor, and a memory connected to the processor;

[0225] The memory is used to store a computer program, and the computer program is used to at least execute the above-mentioned method for evaluating roadbed stability in seasonally frozen soil areas;

[0226] The processor is used to call and execute the computer program in the memory.

[0227] The present invention further provides a storage medium, which stores a computer program. When the computer program is executed by a processor, each step of the above-mentioned method for evaluating the stability of a roadbed in a seasonally frozen soil area is implemented. The method includes: dynamically calculating a comprehensive stability score by real-time analysis of the temperature field and moisture content field distribution of the soil medium, combined with the horizontal spacing, vertical spacing and material performance parameters of the seepage and drainage grilles; further adjusting the design parameters according to the optimization target to reduce the non-uniformity of the temperature gradient and the moisture content gradient; and finally achieving effective regulation of the thermal-water coupling behavior of the roadbed and improving the overall stability of the roadbed.

[0228] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0229] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0230] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0231] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for evaluating roadbed stability in seasonally frozen areas, characterized in that: The specific steps include: Step 1: Collect soil samples to be tested at the target roadbed site, record the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress, bring the collected soil samples back to the laboratory, simulate freeze-thaw cycle experiments, and obtain the thermal conductivity and water diffusion coefficient of the soil samples through experimental fitting; Step 2: Input the collected field observation data and experimental fitting results into the COMSOL finite element model, build a three-dimensional finite element model based on the field observation data and physical experimental parameters, and analyze the dynamic change laws of the temperature field, moisture content field and frost heave stress distribution by simulating the water heat transfer and mechanical behavior under different grid layout conditions; Step 3: Use the finite element model to extract the dynamic characteristic data of temperature field, moisture content field and stress distribution, establish a hydrothermal behavior and reinforcement-soil interface behavior prediction model based on the LSTM network architecture, use the dynamic characteristic data as a training set, train the LSTM model, and predict the hydrothermal dynamic changes and mechanical behavior of the roadbed under different grid design conditions; Step 4: Based on the output results of the prediction model and combined with the comprehensive stability evaluation model, the stability index of the frozen soil roadbed is calculated, and the design and layout of the seepage and drainage grilles are optimized according to the evaluation results.

2. A method for evaluating roadbed stability in seasonally frozen soil areas according to claim 1, characterized in that: The logic of collecting soil samples to be tested at the target roadbed site and recording the dynamic changes of ambient temperature, initial soil moisture content and frost heave stress is as follows: soil samples of different depths and different soil layer types are collected at the target roadbed site, and the collected soil samples are classified according to their layers and physical properties; Temperature and humidity sensors and deformation monitoring equipment are deployed to obtain dynamic change data of the frozen soil layer. Multi-layer monitoring points are set up in the depth direction: 0.2m, 0.5m, and 1.0m to obtain the dynamic gradient of temperature and moisture content of the frozen soil layer. Thermocouple sensors are deployed to record the temperature changes on the roadbed surface and at different depths in real time. TDR (time domain reflectometer) sensors are deployed to record the dynamic changes in the moisture content of the roadbed soil. LVDT (linear variable differential transformer) displacement meters are deployed to monitor the frost heave and thaw deformation of the roadbed surface. After preparation, the collected soil samples were made into cylindrical standard specimens with a diameter of 50 mm and a height of 100 mm. The prepared soil samples were placed in an environmental chamber with freezing and thawing control functions to simulate the temperature gradient and freeze-thaw cycle frequency according to the meteorological conditions of the target area; the seasonal frozen soil temperature variation range in the simulated target area was -20°C to 5°C. Before the experiment, the initial moisture content of the soil sample was adjusted according to the actual state of the roadbed, and vertical pressure was applied through a loading test machine to simulate the overlying load conditions of the actual roadbed; different freeze-thaw cycles were set according to engineering requirements: 10 times, 20 times, to simulate the impact of different freeze-thaw processes on the soil samples; The collected soil samples were brought back to the laboratory to simulate freeze-thaw cycle experiments. The method of fitting the thermal conductivity coefficient and water diffusion coefficient of the soil samples was as follows: the thermal conductivity coefficient k of the soil samples at different water contents was measured through steady-state or non-steady-state heat conduction experiments, the variation law of k with water content w was studied, and the empirical formula k = k (w) was fitted; the diffusion coefficient D of the soil samples under different temperature conditions was measured through water diffusion experiments w , Research D w The law of change with temperature T, and fitting the empirical formula D w =D w (T).

3. A method for evaluating roadbed stability in seasonally frozen soil areas according to claim 2, characterized in that: The logic for calculating and collecting field observation data is as follows: the temperature sensor records the temperature changes at different depths of the soil sample over time and outputs a set of time series data T(z,t T ), where z is the depth data, t T is the time data of temperature change. The temperature gradient is the core parameter of heat conduction and is used to calculate the heat flux in the soil layer. The formula based on the temperature depth gradient is: in, is the rate of change of temperature with depth, z i+1 and z i is the depth position of the adjacent sensor position, T(z i+1 ) and T(z i ) is the temperature value of the corresponding position, i is the index; The moisture content sensor records the change of moisture content in the soil sample over time and outputs a set of time series data w(z,t w ), where z is the depth data, t w is the time data of moisture content change. The moisture content gradient describes the spatial change rate of water distribution in the soil sample. The formula for the moisture content depth gradient is: Among them, w(z i+1 ) and w(z i ) represent the depth z i+1 and z i The moisture content; i+1 and z i is the depth position of the adjacent sensor, It indicates the rate of change of water content with depth; The trend of moisture content over time can be calculated by the time derivative of moisture content, based on the formula: Among them, w(t w,i+1 ) and w(t w,i ) indicates that at time t w,i+1 and t w,i The moisture content at that time, i represents the index; similarly, we can get It represents the rate of change of soil sample temperature over time; The stress sensor records the frost heave force and thaw settlement force of the soil sample during the freeze-thaw cycle, and outputs the stress-time curve σ(τ), based on the formula: Among them, σ(τ) represents the instantaneous stress during the freeze-thaw cycle at time τ, F represents the measured force, and a represents the stress area; The method for obtaining frost heave force data is to extract the maximum value in the frost heave stress curve, which is used to represent the maximum expansion force inside the soil sample during the freezing process. The formula is: s max =max(σ(τ)) Among them, σ max Indicates the peak value of frost heave force; The method for obtaining the melting settlement force data is to extract the minimum value in the melting settlement stress curve, which is used to represent the settlement stress of the soil sample during the melting process. The formula is: s min =min(σ(τ)) Among them, σ min Indicates the peak melting and settling force.

4. A method for evaluating roadbed stability in seasonally frozen soil areas according to claim 3, characterized in that: The collected field observation data and experimental fitting results are input into the COMSOL finite element model. The logic of constructing a three-dimensional finite element model based on field observation data and physical experimental parameters is as follows: The field observation data and indoor experimental data collected in step 1 are interpolated and smoothed, and the processed data are imported into the material property module of COMSOL in the form of formula. The field observation data are used to set the boundary conditions of the temperature field and the moisture content field, and the indoor experimental data are used as the material property input of the model; Set T(x,y,z,t T ) represents the temperature field in the soil sample, where x, y, and z are three-dimensional spatial coordinates, representing the position distribution, and represent the width of the roadbed, the longitudinal length of the roadbed, and the position in the depth direction, respectively, and t T The time variable representing the temperature, when z = 0 it corresponds to the roadbed surface, when z>0 it corresponds to the area below the roadbed; The thermal conductivity coefficient k = k(w) and the diffusion coefficient D obtained from the indoor experiment w =D w The fitting formula of (T) is input into the material property module of COMSOL, and the Heat Transfer Module in COMSOL is used to simulate the temperature field T (x, y, z, t T ) dynamic changes, the formula based on the heat conduction equation is: Where ρ is the density of the soil sample, c is the specific heat capacity of the soil sample, k is the thermal conductivity coefficient, which depends on the water content w, and Q is the internal heat source term. Represents the temperature gradient vector data, that is is the rate of change of temperature along the x direction, is the rate of change of temperature along the y direction, is the rate of change of temperature along the z direction; The Transport of Diluted Species Module is used to simulate water migration, which describes the internal moisture content field w(x, y, z, t w )'s spatial distribution and dynamic changes over time, the formula based on which the moisture migration equation is used is: Among them, J w is the water migration flux, the negative sign indicates that water migrates from the high water content area to the low water content area, D w is the water diffusion coefficient, i.e. D w =D w (T), is the water content gradient vector data, that is is the rate of change of moisture content along the x direction, is the rate of change of moisture content along the y direction, is the rate of change of moisture content along the z direction; w(x,y,z,t w ) is the moisture content field, where x, y, z are three-dimensional spatial coordinates representing the location distribution, t w Represents the dynamic change of moisture content over time.

5. A method for evaluating roadbed stability in seasonally frozen soil areas according to claim 4, characterized in that: The logic of setting the boundary conditions of the COMSOL finite element model is as follows: The method of setting the temperature boundary conditions is, on the roadbed surface, that is, z = 0: adopt boundary conditions consistent with the ambient temperature changes, and directly use the ambient temperature data collected on site as input: T s (t T )=T e (t T ) Among them, T s (t T ) represents the surface temperature of the roadbed, T e (t T ) represents the collected ambient temperature, which changes with time t T Dynamic changes of At the bottom of the roadbed, that is, z = z bottom : Assuming no heat is lost through the bottom boundary, set it to adiabatic boundary conditions: Among them, n represents the normal vector, which indicates the direction perpendicular to the bottom of the roadbed; The method of setting the moisture boundary condition is to set the moisture loss boundary on the roadbed surface, that is, z = 0: Considering the evaporation effect, the moisture loss boundary is set as: J w =f(E) Among them, J w represents the water migration flux, f(E) is the evaporation rate function, which is obtained from literature or experimental data and is related to ambient temperature, humidity, and wind speed conditions; at the bottom of the roadbed, that is, z = z bottom : Assuming that water cannot be lost through the bottom of the roadbed, set it as an impermeable boundary condition: J w =0, this condition is used to simulate the assumption that the water flow at the bottom of the roadbed is blocked; The method of setting mechanical boundary conditions is to use the frost heave force and thaw settlement force as the boundary conditions of the internal stress field, and obtain the maximum frost heave stress σ during freezing through experiments. max and the minimum melting stress σ min The boundary conditions of the input model, frost heave force and thaw settlement force are applied to the nodes on the top surface and side wall of the roadbed respectively to simulate the changes in the internal stress of the soil sample during the freeze-thaw cycle. This setting reflects the distribution law of the actual frost heave stress and analyzes its impact on the stability of the roadbed.

6. A method for evaluating roadbed stability in seasonally frozen soil areas according to claim 5, characterized in that: After the boundary conditions are set, numerical simulation is performed. In COMSOL, the heat conduction equation and the water migration equation are coupled through Multiphysics Coupling to simulate the water-heat coupling behavior of the soil sample during the freeze-thaw cycle. The coupling relationship includes: Temperature field T(x,y,z,t T ) by affecting the water diffusion coefficient D w (T) Realize the coupling relationship with water migration: D w =D w (T), diffusion coefficient D when temperature increases w Increase, resulting in faster water diffusion; Water content field w(x,y,z,t w ) affects the thermal conductivity k and specific heat c of the soil sample, thus acting inversely on the temperature field: k = k(w), c = c(w). When the moisture content increases, both the thermal conductivity and specific heat capacity increase. The logic of simulating the water heat transfer and mechanical behavior of different grid arrangements and analyzing the dynamic changes of temperature field and moisture field is as follows: setting multiple groups of simulation conditions to study the influence of different drainage grid arrangement schemes on water heat transfer and mechanical behavior: the no-grid scheme, as a benchmark comparison, is used to simulate the water heat behavior when there is no drainage grid; the horizontal grid arrangement scheme is used to simulate the influence of the horizontal distribution of the grid on water heat transfer; the oblique grid arrangement scheme is used to simulate the influence of the oblique distribution of the grid on the internal stability of the roadbed; After the simulation is completed, the dynamic changes of three key physical fields are extracted and analyzed: Extract the three-dimensional temperature distribution T (x, y, z, t T ), analyze the temperature gradient, freezing depth and freeze-thaw cycle range inside the soil sample, and compare the effects of different grid arrangements on the stability of the soil temperature field; Extract the three-dimensional moisture content distribution w(x,y,z,t w ): Analyze the law of water migration, including the migration of water from the frozen area to the non-frozen area, and compare the effects of different grid arrangements on water dissipation capacity; extract the distribution of frost heave stress and thaw settlement stress from the stress field: the peak value of frost heave stress σ max and melting stress peak σ min ,By analyzing the overall stability of the roadbed, the anti-frost heave effect of different grid arrangement schemes was evaluated.

7. A method for evaluating roadbed stability in seasonally frozen soil areas according to claim 6, characterized in that: Based on the LSTM network architecture, a prediction model for hydrothermal behavior and reinforcement-soil interface behavior is established. One to two layers of LSTM units are set, and each layer includes 50 to 100 LSTM units. The hidden layer uses ReLU as the activation function, and the output layer uses a linear activation function. Obtain the data that affects the hydrothermal behavior in a time series distribution, that is, the time series data of the temperature field, moisture content field and stress distribution, as well as material parameters, including the thermal conductivity k, the water diffusion coefficient D w ; Extract dynamic features based on time series characteristics, including temperature gradient Moisture content gradient and its time derivative Input the feature matrix, use the dynamic feature data as the training set, and train the LSTM model; Use Batch batch training to divide the entire training set into multiple small batches, and divide the data into small batches with a batch size of 64; In each training iteration, the input feature data is passed into the LSTM model to calculate the output prediction value of the model; the mean square error is used as the loss function to measure the error between the model prediction value and the true value; the gradient of the loss function relative to the model parameters is calculated through the back propagation algorithm, and the model parameters are updated; During the training process, set the hyperparameters of the LSTM model. The hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the number of neurons in the hidden layer, and the time step. Among them, the number of network layers is set to a two-layer network structure, the number of iterations is set to 200, the learning rate is set to 0.001, the batch size is set to 64, the number of training times is set to 500, the number of neurons in the hidden layer is 32, and the time step is set to 10 time steps. The model is gradually optimized and the root mean square error is used on the validation set to evaluate the model performance until the hydrothermal behavior prediction model is trained. The trained LSTM model can predict the hydrothermal behavior indicators under given time and space conditions: heat transfer rate Moisture flux and the behavior of the reinforcement-soil interface: the distribution of temperature gradient and moisture content gradient; By comparing the prediction results with and without the seepage and drainage grid, the role of the grid in regulating temperature distribution and moisture content redistribution is quantified, and the regulation effect index R is calculated. T and R w , providing a basis for subsequent optimization design.

8. A method for evaluating roadbed stability in seasonally frozen soil areas according to claim 7, characterized in that: R T is the temperature regulation coefficient, which is the relative reduction degree of freezing depth fluctuation or temperature gradient change under the condition of no grid and with grid, reflecting the regulation effect of drainage grid on temperature field distribution. The formula is: Where, ΔT no Indicates the maximum fluctuation range of temperature gradient under no grid condition, ΔT yes It indicates the maximum fluctuation range of temperature gradient under the condition of grid; the regulation effect is quantified by the fluctuation of freezing depth or the amplitude of temperature gradient change, and the formula is as follows: For each temperature field data under each condition, calculate the temperature gradient The spatial distribution of and record its maximum and minimum values; Combining the simulation results without and with the grille, calculate the temperature regulation coefficient R T And analyze its physical meaning: When R T >0, indicating that the drainage grille can effectively reduce the temperature gradient fluctuation, and the regulation effect is better; when R T = 0, indicating that the grid has no effect on temperature distribution; when R T When <0, it means that the grid exacerbates the non-uniformity of the temperature gradient; R w is the moisture content adjustment coefficient, which is the relative reduction degree of moisture content gradient or cumulative water content under the condition of no grid and with grid, reflecting the adjustment effect of drainage grid on moisture content field distribution. The formula is: Among them, Δw no Indicates the fluctuation range of moisture content gradient or local moisture accumulation under no grid conditions, Δw yes It indicates the fluctuation range of moisture content gradient or local moisture accumulation under the condition of grid. The calculation formula of the fluctuation of moisture content gradient or local moisture accumulation is: For each condition of the moisture field data, calculate the moisture gradient spatial distribution of When R w >0, it means that the drainage grid can effectively reduce the fluctuation of moisture content gradient and improve the uniformity of moisture distribution; when R w = 0, indicating that the grid has no effect on the moisture content distribution; when R w When <0, it means that the grid exacerbates the unevenness of moisture distribution; Based on the machine learning analysis results of step 3, the logic of the comprehensive stability evaluation model is calculated as follows: T and R w , respectively reflect the role of the grid in regulating the temperature gradient fluctuation and the moisture content gradient fluctuation, and quantify the contribution of the grid to the environmental regulation. They reflect the improvement of the grid on the non-uniformity of the temperature field and the moisture content field. Therefore, R T and R w It is a key indicator that directly reflects the thermal stability and water stability of the roadbed; R T >0 and R w >0 indicates that the grid has a positive regulation effect, while R T <0 or R w <0 means that the grille design is improper, which exacerbates the unevenness; In order to comprehensively evaluate the thermal-water coupled stability of the roadbed, based on R T and R w The comprehensive roadbed stability score is calculated based on the formula: S=α·(1-R T )+β·(1-R w ) Among them, S is the comprehensive stability score, (1-R T ) represents the residual inhomogeneity that cannot be covered by the temperature regulation effect, (1-R w ) represents the residual heterogeneity that cannot be covered by the moisture content adjustment effect, α is the importance weight of temperature adjustment, β is the importance weight of moisture content adjustment, the value range of α and β is 0.4-0.6, and α+β=1, which is adjusted according to whether the actual environment in the frozen soil area is heat-dominated or water-dominated; According to the comprehensive evaluation score S, the subgrade stability is divided into the following levels: The subgrade stability is extremely high, that is, when S ≤ 0.2: The hydrothermal coupling behavior of the subgrade is highly uniform, with almost no significant non-uniformity, and it is suitable for high-standard frozen soil subgrade engineering. The subgrade stability is relatively high, that is, when 0.2 < S ≤ 0.4: The hydrothermal coupling behavior of the subgrade is relatively uniform, and the non-uniformity has no significant impact on the engineering performance. The subgrade stability is average, that is, when 0.4 < S ≤ 0.6: The subgrade shows a certain degree of non-uniformity, which will affect the long-term stability. The subgrade stability is poor, that is, when S > 0.6: The hydrothermal coupling behavior of the subgrade is significantly non-uniform, and the design needs to be optimized to improve the stability. Combined with the evaluation results of S, the design parameters of the seepage and drainage grilles are optimized to further improve the stability of the roadbed while considering the economic efficiency of the construction cost; the optimization variables are: the horizontal spacing of the grilles, the arrangement distance of the seepage and drainage grilles in the horizontal direction. The smaller the horizontal spacing, the stronger the seepage and drainage capacity, and the better the R T and R w , the more evenly the moisture and heat are distributed, the better, but the construction cost increases; The vertical spacing of the grid is the vertical arrangement distance of the drainage grid. The smaller the vertical spacing, the more significant the water and heat regulation effect of the grid on the deep soil, further improving R T and R w , it is necessary to comprehensively consider the on-site geological conditions and construction feasibility; The properties of the grid material, including the water permeability coefficient, directly affect the water migration capacity. The higher the water permeability coefficient, the stronger the water migration capacity, which in turn changes the R w and compressive strength. The higher the compressive strength, the better the durability. Constraints: comprehensive stability score S≤0.2, engineering cost is less than or equal to the budget, ensuring cost controllability; The optimization design requires adjusting the horizontal spacing, vertical spacing and material properties, the initial material permeability coefficient and compressive strength, using the LSTM model in step 3 to input the design parameters, predict the distribution of temperature gradient and moisture content gradient, and calculate R T , R w and comprehensive stability score S; if S>0.2, it means that the roadbed stability has not reached the target, and the design parameters need to be adjusted, such as reducing the horizontal or vertical spacing, or improving the permeability of the material, re-entering the prediction model, and updating R T and R w ; By dynamically adjusting the parameters, S is iteratively calculated until the constraints are met, that is, S≤0.2 and the cost is within the budget.

9. A device for evaluating roadbed stability in seasonally frozen areas, characterized in that: The device is used to execute a method for evaluating the subgrade stability in a seasonal frozen soil area according to any one of claims 1-8, including: A data acquisition module, which is used to detect by collecting the soil samples to be detected at the target subgrade site, record the dynamic changes of the ambient temperature, the initial moisture content of the soil body, and the frost heave stress, bring the collected soil samples back to the laboratory, simulate the freeze-thaw cycle experiment, and obtain the thermal conductivity coefficient and the moisture diffusion coefficient of the soil samples through experimental fitting. Three-dimensional model construction, which is used to input the collected on-site observation data and experimental fitting results into the COMSOL finite element model, construct a three-dimensional finite element model based on the on-site observation data and physical experiment parameters, and analyze the dynamic change laws of the temperature field, moisture content field, and frost heave stress distribution by simulating the hydrothermal transfer and mechanical behavior under different grille layout conditions. A model establishment module, which is used to extract the dynamic characteristic data of the temperature field, moisture content field, and stress distribution by using the finite element model, establish a prediction model for the hydrothermal behavior and the soil-reinforcement interface behavior based on the LSTM network architecture, use the dynamic characteristic data as the training set, train the LSTM model, and predict the hydrothermal dynamic changes and mechanical behavior of the subgrade under different grille design conditions. A stability judgment module, which is used to calculate the stability index of the frozen soil subgrade based on the output result of the prediction model and in combination with the comprehensive stability evaluation model, and optimize the design and layout of the drainage grille according to the evaluation result.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, it implements a method for evaluating the subgrade stability in a seasonal frozen soil area according to any one of claims 1-8.

Images